DayVectors

* 20 articles including: Joe Biden (series); generative AI arrives (series); Norway 3SM cruise missile | automated air refueling | smart 57-mm shell; COVID resistance; bump-stock ban going nowhere | synthetic couples | Farage strikes out; Chang'e 6 Moon probe | PREFIRE infrared environmental CubeSat; Psyche asteroid probe; upside-down flag | Spout poll on Gaza | using MS Copilot; nano-shapes | improved desalinizer | post-quantum encryption; current solar activity & auroras; Ukraine peace summit | Netanyahu under pressure | Palmer Report; WASP-107B puff planet | sinking cities on US East Coast; digital Estonia; Putin's economic quandary | Trump found guilty | Gaza peace proposal.

[FRI 28 JUN 24] JOE BIDEN (32)

* JOE BIDEN (32): Joe was, for once, in no rush to get on the move again; haste would only prolong the recovery. He gradually began to get some sun, take walks, use an archery set Jimmy had given him, and hit golf balls at a driving range. He didn't take calls, leading the news media publicly speculating that the second operation had left him in a vegetative state.

The only person who got through the wall was Ted Kennedy, who characteristically refused to take NO for an answer. He showed up with a framed etching of an Irish stag, signed: "To My Chairman" -- and also brought a swimsuit for use with the Station's pool. Kennedy did not talk business, instead saying only that Joe was missed, and everyone was waiting for him to come back to Congress. Joe found the visit appropriately encouraging. Later, Joe found out that Kennedy had informally taken charge of Joe's staff, with the two staffs collaborating to make sure everything got done in both their domains.

During this time, Joe had to deal with a family problem. Beau didn't like to drink but Hunter did, and he started to experiment with drugs -- to be busted by the police for possession of cocaine. Almost nothing was said about the matter, with Hunter's criminal record expunged, and it's not clear exactly when the incident happened. Some of Joe's critics later saw something suspicious about the way Hunter's criminal record was dropped, but they had nothing they could prove one way or another. After the bust, Hunter straightened up -- for a time.

The incident of course had to have greatly distressed Joe and Jill. Joe never touched a drop, knowing that alcoholism ran in the family, and discouraged the kids from drinking. Valerie never had much interest in alcohol; Jimmy drank for a time as an adult, until he found out he was allergic to alcohol -- his immune system would attack it, not merely preventing him from getting drunk, but putting him in the hospital. One stint in the hospital was enough for Jimmy, and he never drank again. Frank, however, drank from when he was in high school and later suffered bouts of alcoholism for about a decade, when he finally went straight for good. Jill drank for a time but never had a problem with booze, and ultimately quit.

In any case, Joe finally made a public appearance at the end of August, speaking at the annual Delaware Sussex County Jamboree, telling an audience of hundreds, who were chanting his name:

QUOTE:

END_QUOTE

A few days later, Joe was talking with Jimmy in the kitchen, when he got a call from Mike Dukakis -- who was battling George Herbert Walker Bush, Reagan's vice president, in the ending stretch of the presidential election. Bush was pulling ahead, and Dukakis wanted to bring back John Sasso, the staffer who had been fired after leaking Joe's gaffe on the Kinnock story to the press. Dukakis wanted to know if that would cause any problems. Joe didn't have any difficulties with the matter; whatever helped Dukakis beat Bush was good.

When Dukakis announced the re-instatement to the press, he said that he had talked about doing so with Joe Biden, who "could not have been more gracious." Sasso then issued an apologetic statement:

QUOTE:

END_QUOTE

That was the effective end of the matter. Joe's medical emergency would have derailed his campaign in any case, if it didn't kill him; he could try again later. The "plagiarism" charge would be resurrected occasionally later, but it no longer carried any weight.

On 7 September 1988, after almost seven months, Joe took the morning Amtrak train from Wilmington to Washington DC, accompanied by the Biden clan. Railroad staff put up balloons and signs to welcome him back, and the engineer gave an extended toot of the horn as a salute.

Back in Congress, his staff gave him a surprise party, and his Senate peers dropped by his office to greet him warmly. Near the end of the day, the Senate passed a resolution as a welcome. Joe spoke to the assembly:

QUOTE:

END_QUOTE

Of course, he didn't stop there; that wasn't like Joe. However, he had changed in one significant way: he was no longer a young man in a hurry. Joe remained focused on his work, of course, but he didn't have the same sense of urgency. That was reserved for things that actually were urgent.

That didn't mean Joe was no longer serious about his political career. In fact he realized that, in his effort to balance his family and his work, he hadn't been serious enough about it:

QUOTE:

END_QUOTE

[TO BE CONTINUED]

[THU 27 JUN 24] WINGS & WEAPONS

* WINGS & WEAPONS: As discussed in an article from THEDRIVE.com ("Supersonic Naval Cruise Missile For Norway, Germany In The Works" by Joseph Trevethick, 29 November 2023), the Kongsberg Defense company of Norway has developed the "Naval Strike Missile (NSM)", an antiship / land-attack cruise missile, popular with NATO forces, and went to a derivative "Joint Strike Missile (JSM)", which is entering service. Now, boosted by an agreement between Norway and Germany, Kongsberg is working on the "SuperSonic Strike Missile (3SM)" AKA "Tyrfing", a magical sword from Norse mythology.

The only thing released concerning the project so far is imagery of a digital model, revealing a spikelike rocket with cruciform tailfins, strakes on each side of the fuselage, and a large underslung engine intake that suggests the missile uses ramjet propulsion, presumably with rocket-boosted take-off. Range, speed, and warhead size are not known. It will presumably have a guidance system like that of the NSM, featuring GPS-INS navigation and terrain-following, along with an imaging infrared targeting seeker. There's no announced schedule for introduction to service.

* As discussed in an article from NEWATLAS.com ("Autonomous Air-To-Air Refueling System Will Radically Boost Endurance" by Loz Blain, 26 November 2023), European aerospace giant Airbus has an "innovation division" named "UpNext", which works on leading-edge tech for future Airbus products. One of its latest production is a next-generation autonomous inflight-refueling system, focused on keeping drones in the air on a long-term basis. Drones that can stay on-station, anywhere in uncontested airspace, are useful for surveillance, communications relay, electronic warfare, or combat patrolling.

The UpNext team accordingly working on a program called "Auto'Mate", envisioned as a pushbutton, automated solution for inflight refueling. The idea is that the Auto'Mate system will take control of the receiver aircraft from the tanker when it's several kilometers away, and guide it into formation with the tanker plane using "AI-based relative navigation and co-operative control technologies."

At this point, with the two aircraft flying together under a single system's control, the refueling probe is deployed and hooked up to the receiver aircraft, fuel is transferred, and control is handed back to the receiver aircraft so it can continue its mission.

Preliminary tests so far have featured an Airbus A310 Multi-Role Tanker-Transport (MRTT) aircraft refueling Airbus / Dornier DT-25 target drones used as receivers. The refueling system uses a set of cameras, high-precision GPS and Lidar, along with covert, secure multi-node communication channels and built-in collision avoidance gear. There's no schedule on when Auto'Mate will be introduced to service, however.

* As discussed in an article from NEWATLAS.com ("Northrop Grumman To Build One-Shot-One-Kill Shell For US Navy" by David Szondy, 08 October 2023), Northrop Grumman has been awarded a US Navy contract to develop a new self-guided 57-millimeter artillery shell for the Navy's Mk110 Naval Gun Mount medium-sized gun used on Littoral Combat Ships. The shell will have a miniaturized seeker system -- exactly what kind of seekers was not described -- and an aft-mounted directional system. In addition, the fuze can self-select for proximity or point-detonation mode for maximum effect.

[WED 26 JUN 24] GEN-AI ARRIVES (20)

* GEN-AI ARRIVES (20): Probing an AI mind with chain-of-thought reasoning is a very interesting idea, but late in 2023 Sam Bowman -- Sam Bowman, a computer scientist at New York University and Anthropic -- and his colleagues demonstrated that chain-of-thought explanations are not necessarily faithful indicators of what an LLM is really doing.

The researchers first intentionally biased their study models by, for example, giving them a series of multiple-choice questions for which the answer was always option "A". The team then asked a final test question. The models usually answered "A" -- which may not have been correct -- but when asked why they chose "A", they almost never said it was out of habit. Instead, they cooked up some "logic" to justify their response. Humans often do the same thing, consciously or unconsciously.

This phenomenon is similar to the implicit social bias that sometimes makes recruiters hire candidates who look or act like them, even while they proclaim that the successful applicant was simply the most qualified for the job. Bowman's paper shows similar social bias in LLMs.

However, Sandra Wachter -- who studies technology regulation at the Oxford Internet Institute, part of the University of Oxford in the UK -- says that doesn't really prove that chain-of-thought reasoning is useless. It's not useless to ask humans to justify their reasonings, even if they generate hallucinations in response. Discussions with chatbots just have to be taken with a grain of salt "in the same way that when you're talking to a human you have some healthy distrust."

David Bau says it's "a little weird" to study LMMs the same way we study humans -- but though there are limits to the parallels, the behavior of the two overlaps in surprising ways. Many papers in the past few years have applied human questionnaires and experiments to LLMs, measuring the machine equivalents of personality, reasoning, bias, moral values, creativity, emotions, obedience, and "theory of mind" -- an understanding of the thoughts, opinions and beliefs of others or oneself. In many cases, machines reproduce human behavior; in other situations, they don't. For instance, Hagendorff, Bau, and Bowman each note that LLMs are more suggestible than humans, with their behavior morphing drastically depending on how a question is phrased. Thilo Hagendorff says:

QUOTE:

END_QUOTE

Other researchers are taking hints from neuroscience to explore the inner workings of LLMs. To examine how chatbots deceive, Andy Zou -- a computer scientist at Carnegie Mellon University in Pittsburgh, Pennsylvania -- and his collaborators interrogated LLMs and looked at the activation of their electronic neurons. He says: "What we do here is similar to performing a neuroimaging scan for humans." It's also a bit like designing a lie detector.

The researchers told their LLM several times to lie or to tell the truth and observed the differences in patterns of neuronal activity, creating a mathematical representation of "truthfulness". Then, whenever they asked the model a new question, they could look at its activity and estimate whether it was being truthful, with more than 90% accuracy in a simple lie-detection task. Zou says that such a system could be used to detect LLMs' dishonesty in real time, but the accuracy needs to be improved first.

The researchers went further and intervened in the model's behavior, adding these truthfulness patterns to its activations when asking it a question, enhancing its honesty. They followed these steps for several other concepts besides "truthfulness" as well: they could make the model more or less power-seeking, happy, harmless, gender-biased and so on. [TO BE CONTINUED]

[TUE 25 JUN 24] COVID RESISTANT

* COVID RESISTANT: As discussed in an article from SCIENCE.org ("One In Five People Who Contract The COVID-19 Virus Don't Get Sick" by Mitch Leslie, 19 July 2023), mysteries linger about the SARS-CoV-2 virus that causes COVID-19 -- one being that, while it can kill those who contract the disease, about a fifth simply shrug it off. A study now shows that people who don't get sick possess a version of a particular immune system gene. When these people are infected with one of the common coronaviruses that cause colds, they obtain strong immunity against infection by SARS-CoV-2.

The gene in question is one of the "human leukocyte antigen (HLA)" genes, which encode proteins that help the body deal with pathogens. If a virus invades one of our cells, HLA proteins on the cell's surface display fragments of the invader's proteins, alerting the immune system. T cells -- which recognize and "remember" pathogens -- then kill the infected cell, destroying the virus along with it.

There are thousands of varieties of HLA genes, which can influence how energetically people's immune systems combat particular viruses. For example, HIV-infected people with a certain "HLA-B" version have very low levels of the virus and don't show symptoms, while patients with another HLA-B type rapidly progress to AIDS.

At the outset of the COVID pandemic, nobody knew if there were HLA variants that provided resistance to SARS-CoV-2, but it seemed like a good possibility. Immunogeneticist Jill Hollenbach of the University of California, San Francisco, and colleagues decided to investigate. Given the lockdowns in progress at the time, getting subjects into lab for tests was impossible -- so the researchers devised a contact-free procedure, enlisting a group of nearly 30,000 people who had signed up to donate bone marrow, and so had already undergone tests to determine their HLA complement.

The participants downloaded a cellphone app and used it to report any symptoms, even mild ones like a runny nose or scratchy throat, that persisted for at least 3 days.

Of the more than 1400 subjects who tested positive for SARS-CoV-2, 136 reported that they did not feel sick. 20% of the symptom-free group carried the same version of the HLA-B gene, known as HLA-B*15:01. In contrast, only 9% of the subjects who reported symptoms on the app carried this HLA variant. Participants with two copies of HLA-B*15:01, one provided by each parent, got even more protection, being eight times less likely to suffer symptoms than were people with other HLA-B varieties.

What was the HLA-B variant doing? To figure that out, the researchers inspected blood samples collected from other people before the pandemic. In 75% of blood donors with HLA-B*15:01, T cells recognized a snippet of SARS-CoV-2's spike protein, which the virus uses to penetrate cells. The T cells seemed primed to attack the virus, even though the subjects had never encountered it before.

Apparently the blood donors had been infected by seasonal coronaviruses, widespread relatives of SARS-CoV-2 that are one of the many viruses that cause the common cold. The researchers discovered that T cells from the subjects with HLA-B*15:01 also reacted aggressively to a fragment of the spike protein from two of the seasonal coronaviruses -- this fragment being almost identical to the snippet of SARS-CoV-2's spike protein that the researchers had tested.

The conclusion was that, thanks to HLA-B*15:01, these subjects obtained a degree of immunity from SARS-CoV-2 from infection by common-cold coronaviruses. There had been suspicions that was possible from early on in the pandemic, but research had given mixed results -- as it turns out, because of a variation in host genetics. The gene doesn't actually stop a SARS-CoV-2 infection, but the immune response is fast enough to defeat the virus before the subject feels sick.

Further research is needed to determine exactly how the variant charges up the response to SARS-CoV-2. The research may help the search for new vaccines or treatments to deal with COVID-19. [ED: Early on in the pandemic, there was some thought that exposure to earlier coronaviruses conferred resistance, but some studies came up zeroes. It turned out that some people pick up resistance, but only those with the right mutation.]

[MON 24 JUN 24] THE WEEK THAT WAS 25

DAYLOG MON 17 MAY 24: In reaction to the Supreme Court's rejection of Congress's ban on bump stocks for semi-automatic rifles, Senate Majority Leader Chuck Schumer is talking about moving swiftly on passing the ban again.

Now, I realize Schumer has forgotten far more about legislating than I'll ever know, and I'm certain he knows what he's doing. The question is: What IS he doing? Can he really restore the ban against SCOTUS opposition?

That doesn't seem too hard to believe, because it appears the SCOTUS rejection of the ban on bump stocks involved ridiculous tapdancing on the definition of "machine gun". Could a new ban with a more explicit definition make SCOTUS think twice about voting it down?

On the other hand, Schumer is talking about a unanimous vote -- some Senate procedure I know nothing about -- and he may just be trying to make the Republicans look bad (or worse than they already do) ahead of elections. Eh, maybe it's both.

* NPR reports that trolls are now busily spreading disinformation online that sunscreen is dangerous to use. Of course, that's nonsense. I cannot figure out what motivates people to say such ridiculous things. The ultraviolet is strong here in mile-high Colorado, so sunscreen is essential for people who spend much time in the Sun. I did have a skin tumor some years back; it wasn't serious, the doctor just sliced it out, but it was a wake-up call. I get a skin checkup in September.

Along parallel lines, I went in to Kaiser Permanente for my yearly checkup, and unexpectedly got another COVID booster. I got to wondering how many vaccinations I've had in my life -- maybe a hundred? Half that many? Twice that many? Hard to say. What's not hard to say is that, other than feeling rocky for the day after sometimes, I've never had any problems with vaccinations. Antivaxers talk like I'm taking my life in my hands getting a vaccination. They lie; again, I don't really understand why.

DAYLOG TUE 18 MAY 24: The MAGA Right has been continuing their propaganda offensive against Joe Biden, these days notably with "cheapfakes" -- disinformation videos fashioned with selective editing, to show JB "lost and confused".

The White House has been hitting back, notably with Press Secretary Karine Jean-Pierre calling out the "cheapfakes" and the outlets, such as the NY POST, that peddle them. Of course the POST doubled down on the fakes, but the idea was to get the real story into the news flow. There's been much concern over the threat of AI "deepfakes", but it's hard to see it's such a worry -- because we're already saturated with fakery, and AI doesn't bring much new to the party.

* Regarding Chuck Schumer's effort to re-ban bump stocks, discussed yesterday: The GOP has already shut it down, so I guess Schumer really was trying to make them look bad. GOP == I2-C4-F2: ignorant, incompetent -- conflicted, confused, compromised, complicit -- fraudulent & fascistic.

* In Ukraine War news, Armed Forces Ukraine has scored major successes against the Russian Black Sea fleet with their sea drones AKA "sea babies". The latest effort along such lines looked like a small fishing boat with an outboard motor.

It had a covered deck, and a frame above the rear mounting a sensor turret. Apparently it is intended mostly for patrol / scouting and running supplies, not for use as a kamikaze drone. One wonders if it has a laser target designator in the turret to direct drone and missile attacks. As far as using it for supply goes, it hints at support of AFU Special Operations Forces in Crimea. There may well be hundreds of SOF there, presumably to provide intelligence.

On another front, the AFU is now obtaining thousands of volunteer troops from prisons. Since bringing in hardened criminals is not a good idea, convicts can only be recruited if their crimes are on a restricted list -- negligent homicide, for instance. Violent criminals need not apply.

DAYLOG WED 19 MAY 24: Donald Trump is, as usual, being a nuisance, headlines today saying that, if he's elected, corporate CEOs who didn't back him would pay for it. I replied on Spout: "Cry more, Trump."

Later on in the day, Trump announced plans to cancel all student relief programs. I thought: That's the most ridiculous campaign pledge ever. -- and posted on Spout: "Wotta loser."

Polls keep saying Trump is a big election challenge to Joe Biden, but I just don't see it. I keep saying: He lost handily in 2020, he'll be in better shape in 2024? Campaigning from criminal court? When he barely knows what planet he's from? With good reason, Trump is denounced as a villain -- but he's not a serious person, not even as a villain. He's like a villain in Saturday-morning cartoons. He doesn't even rate being a villain in Marvel Comix.

Trump and JB are scheduled to debate on the 27th. No audience, mic is off when it's not turn to talk, no prepared notes. Trump will be "lost and confused" through the debate, and it will be a disaster for him. Will he be a no-show? If he doesn't show, he will be conceding defeat. Trump is doomed either way.

* There's talk that sanctions on Russia have been ineffective, but all inspection suggests they are highly effective. China has been working to fill the gap left by the exit of Western firms from Russia, but transport infrastructure is a bottleneck.

One particular bottleneck turns out to be shipping containers. They are loaded with cargo sent to Russia, but Russia doesn't have cargo to put in them to send back. The result is that empty containers are piling up in Russia, scarce in China, and shipping is getting more expensive. Certainly, a profound shortage of containers in China makes it likely empty containers will be shipped back from Russia, but that increases expense as well.

DAYLOG THU 20 MAY 24: It appears F-16s are about to arrive in Ukraine, with the USA authorizing their use in striking targets in Russia where offensive actions against Ukraine are taking place. Some observers say it would not make sense to use them against Russia. The thought is that flying the F-16s too close to the battle lines risks their destruction. It is difficult to believe the Ukrainians don't know that, and so they certainly have specific and thought-out ideas of what they want to do with them.

My own thought is that they want the F-16s to engage Russian air power that has been carrying glide bombs and other munitions to attack Ukraine. The Russians have been dropping such munitions from out of range of AFU defenses but with long-range AMRAAM missiles, F-16s may be able to pick them off. Hints are that the F-16s headed for Ukraine are being upgraded to the latest "F-16V" configuration, most notably with AN/APG-83 Scalable Agile Beam Radar (SABR). The hints are that Ukraine asked for SABR, and the older AN/APG-68 doesn't have adequate range to support extreme-range intercepts. A fully refurbished F-16V is effectively a new aircraft; although refurbishment means delay, it would not make sense to do less.

Not incidentally, the USA is now prioritizing deliveries of Patriot and NASAM surface-to-air missile loads to Ukraine. This means postponing deliveries to countries already in the queue; I assume they consented. NASAMS is a ground launcher for AMRAAM air-to-air missiles, in this context being known as "Surface-Launched AMRAAM (SLAMRAAM)". Interesting question: are the Ukrainians receiving the heftier and more formidable extended-range SLAMRAAM-ER? Nobody's talking.

Other Ukraine War news says the Orcs are building up forces to take the town of Borova, southeast of Kharkiv. That is clearly a threat, but I suspect the current AFU munitions buildup will crush them. We'll see by the 1st of September.

DAYLOG FRI 21 MAY 24: I have an account on the DeviantArt DreamUp image-generation system, and occasionally run across some interesting accounts by others. Recently I came across an account that features images of imaginative celebrity couples.

How about Marilyn Monroe and Nat King Cole? Hedy Lamarr and Marlon Brando? Hedy Lamarr and Christopher Walken? That one threw me. Ava Gardner and Jamie Foxx? Hedy Lamarr and Chris Evans? Evans of course as Captain America, works for the time period. Liz Taylor and Mel Gibson? Liz Taylor and Prince? There was other stuff in that account, celebrities as zombies for example; much of it is "Not Safe For Viewing", so I won't give the account name. That could lead to problems. I found it a clever and inspirational idea.

AND SO ON: Nigel Farage, one of Britain's foremost fascistic trolls, stuck his foot in it by declaring that the Ukraine War was the West's fault: "We provoked this war."

Tory Prime Minister Rishi Sunak said the comment was "completely wrong and only plays into Putin's hands," accusing Farage of "appeasement" that was "dangerous for Britain's security". Labour leader Sir Keir Starmer said Farage's comments were "disgraceful", while Lib Dem leader Sir Ed Davey called Farage "an apologist for Putin".

In response, Farage of course gaslighted his comments, condemning Vladimir Putin and the invasion, but still insisting that the expansion of NATO was the real cause. Anyone familiar with the origins of the war knows that's nonsense; it was driven by Putin's determination to restore the Russian Empire of the Tsars, along with his tyrant's fear and loathing of Ukraine's democratic revolution.

Sunak has called a general election for 4 July. There was some puzzlement as to why he did so, since the general belief is that the Tories will be slaughtered at the polls. Investigation suggests that Sunak realized that the Tories would suffer worse if they stalled the election until it was mandated. Sunak seems to be sensible, if not sensible enough to give up on the Tories. Farage's Reform Party -- once the Brexit Party -- is expected to pick up a few seats in Parliament, which will be doing good by their standards. Rightist trolls seem to be going out of fashion, however, so that's likely to be about as far as they will get.

[FRI 21 JUN 24] JOE BIDEN (31)

* JOE BIDEN (31): Back home, Joe was characteristically determined to recuperate as fast as possible, while Jill did all she could to get him to take it easy. Events demonstrated that Joe did indeed need to take it easy; he suffered from major dizzy spells on occasion. Dr. George assured him that was nothing unusual after major brain surgery: "It will go away, but healing is a process."

Much worse, one morning Joe woke up immobilized by agonizing pain. Jill called a local doctor, the same one who had diagnosed Joe's warning pains as a pinched nerve, who suggested it was just intestinal gas. That wasn't convincing; a trip to the hospital in Wilmington showed he had a blood clot in a lung, which was not unusual for patients who had been bedridden for an extended period of time. The clot meant ten more days at Walter Reed, where he had a blood filter to catch clots surgically installed in his circulatory system, was given anti-clotting drugs, and subjected to repeated tests -- some painful.

During that second stay at Walter Reed, Joe decided that for once in his life, he might not be in such a hurry. Nonetheless, he was glad to be released. The reporters were waiting for him; he joked with them, saying: "I've asked you all to come today, because I'm announcing I'm re-entering the race for president." They laughed.

Back in Wilmington, Jill kept Joe in strict isolation: no work, no phone calls, minimal contact with the outside world. Ronald Reagan called twice, but he couldn't get through; Jill just told Reagan how things were going, said she would pass his best wishes on to Joe, and sincerely thanked the president for his concerns. The Democratic primary race was rolling on in the meantime, with Dukakis becoming the leader, but Joe knew little about it. He managed daily walks, but spent most of his time sleeping.

The prospect of surgery on the second aneurysm worried him. He hadn't really dreaded the surgery on the first aneurysm, because he'd been so overwhelmed at the time and things had moved so fast. With time on his hands, he had to think about the finite chance the operation would kill him, and the bigger chance it would leave him impaired.

Joe went under the knife again at Walter Reed in early May. The operation was a complete success, but Dr. George said Joe needed to stay in the hospital to look out for blood complications. Joe remained bedridden at Walter Reed through the rest of the month, which proved extremely boring. The only thing he had to do was watch the construction of a radio tower from the window in his room.

The main thing, however, was that he had gone through the worst and could contemplate the long plod to get his life back on track. There was also a subtler lesson for him to consider over the longer run: he had survived in large part because he had obtained the best medical care available. What, then, would happen to Americans who barely had access to medical care?

Jill, who was teaching full-time, commuted to Walter Reed at nights and weekends. Neighbors helped out to keep the Biden household going; the Biden's favorite sub-sandwich shop made subs for Joe several times a week, with Jill ferrying them to Walter Reed. He gave the subs he didn't eat to the nurses there, who appreciated them.

When he was finally released to go back to the Station, he was feeling much stronger, but he was very thin, his attention span was short, and he tired easily. Most disturbing was that the surgery had affected nerves and left the right side of his face partly paralyzed. The doctors told him it might -- or might not -- clear up after a time; it would do so about six weeks later. [TO BE CONTINUED]

[THU 20 JUN 24] SPACE NEWS

[ED: It has become very laborious to keep up with space launches, all the more so because they're increasingly not so well documented. As a result, I'm not going to provide a comprehensive list any longer, instead focusing on highlights.]

* Space launches for May included:

• Ten batches of "Starlink" comsats were launched on SpaceX Falcon 9 boosters from Vandenberg SFB and Cape Canaveral, flying over 200 satellites.
  
  
• There were 3 other Falcon 9 launches, payloads including a pair of Maxar "Worldview Legion 1 & 2"; a classified National Reconnaissance Office (NRO) payload, designated "NROL 146" and referred to as "Starshield"; and the ESA-JAXA "EarthCARE" environmental monitoring satellite -- AKA "Earth Explorer 6" of the Living Planet Program.
  
  
• China performed nine launches, most on Long March boosters, but also one launch each of the "Kuaizhou 11", "Ceres 1", and sea-launched "Ceres 1S" light boosters.
  
  
• Russia performed two launches, payloads including a surveillance satellite, designated "Cosmos 2576", and a Progress freighter-tanker for the International Space Station, designated "MS 27 / ISS 88P".
  
  
• North Korea attempted to launch the "Malligyong-1 1" reconnaissance satellite, but the undescribed booster did not make orbit.

Noteworthy launches included:

[03 MAY 24] CN WC / LONG MARCH 5 / CHANG'E 6 -- A Long March 5 booster was launched from Wenchang at 0927 UTC (local time - 8) to send the "Change'e 6" lunar sample return probe to the lunar farside.

Chang'e 6 had a launch mass of 8,200 kilograms (18,075 pounds). It was a backup lander to the earlier Chang'e 5 mission. The probe was expected to return up to two kilograms (4.4 pounds) of lunar soil samples to Earth. Earlier missions had returned samples from the lunar nearside. The mission system consists of four major modules: an orbiter, lander, ascent vehicle, and sample-return capsule. The lander had a mass of about 3,200 kilograms (7,055 pounds), not counting the 700-kilogram (1,545-pound) ascent vehicle, while the orbit and its sample-return capsule had a mass of 4,300 kilograms (9,480 pounds).

The lander acquired samples and put them in the ascent vehicle, which boosted back into lunar orbit and docked with the orbiter. The orbiter then returned to Earth and dropped the sample-return capsule. The lander included a set of sophisticated instruments from foreign partners:

• France provided the "Detection of Outgassing Radon (DORN)", which conducted studies of the transport of lunar dust and other interactions between lunar regolith, the lunar water, and more. DORN was developed by the French Research Institute of Astrophysics and Planetology (IRAP), in a tight relationship with CNES. DORN was to assess the outgassing from the lunar crust, which includes elements such as helium-4, argon-40, and radon.
  
  
• Italy provided the "INstrument for landing-Roving laser Retroreflector Investigations (INRRI)", consisting of two passive laser retro-reflectors to be used in the future for laser range-finding of the lander.
  
  
• Sweden provided the "Negative Ions on Lunar Surface (NILS)" ion-sampling instrument.

Pakistani students provided a CubeSat, named "ICUBE-Q", to accompany the flight to the Moon. It took pictures of the Earth and Moon, and also measured the Moon's magnetic field.

[25 MAY 24] NZ / ELECTRON / PREFIRE 1 -- A Rocket Labs Electron light booster was launched from New Zealand's Mahia Peninsula at 0741 UTC (local time - 13) to put the "Polar Radiant Energy in the Far-InfraRed Experiment (PREFIRE) 1" satellite into orbit. It was a 6-unit CubeSat, intended to perform infrared surveys of the polar regions.

[WED 19 JUN 24] GEN-AI ARRIVES (19)

* GEN-AI ARRIVES (19): Harvard University computer scientist Kenneth Li, working with David Bau, Martin Wattenberg and others, trained an LLM to play the board game Othello, in which opponents place black and white discs on a grid. The researchers fed their model, called "Othello-GPT", sequences of moves in text form from past games, until it learned to predict the likely next moves. The team also trained a smaller model to interpret the internal activations of the AI, and discovered that it had constructed an internal map of the discs based on the text descriptions of the gameplay. Wattenberg says: "The key insight here is that often it's easier to have a model of the world than not to have a model of the world."

Since chatbots are of course chatty, some researchers are investigating their workings simply by asking them to explain themselves, in much the same way as humans do to a psychologist. Thilo Hagendorff, a computer scientist at the University of Stuttgart in Germany, says:

QUOTE:

END_QUOTE

Hagendorff has written papers on "machine psychology", in which he argued that conversation with an LLM can shine light on the sophisticated behaviors that emerge from the underlying calculations.

A 2022 study by a team at Google introduced the term "chain-of-thought prompting" to describe a two-step method for getting LLMs to explain their "thinking". The user provides a sample question and describes, step by step, how it can be answered -- and then asks the real question. This prompts the model to follow a similar process, also explaining its chain of thought -- and some studies show it is more likely to get a correct answer.

For example, here is a simple math word problem, of the sort that has long plagued schoolkids, using standard prompting:

QUOTE:

END_QUOTE

Note that the second query is a little ambiguous, in that it should have specified "have left" instead of just "have". Humans can figure out the cafeteria doesn't have the 20 used for lunch any more, but it can make trouble for machines. After being prepped with explained sample questions, chain-of-thought prompting gets clearly different and better results:

QUOTE:

END_QUOTE

[TO BE CONTINUED]

[TUE 18 JUN 24] PSYCHE PROBE

* PSYCHE PROBE: A SpaceX Falcon 9 booster was launched from Cape Canaveral on 13 October 2023 to send the "Psyche" asteroid probe into deep space. Psyche's target is the metallic asteroid 16 Psyche; it will be the first "M-type" (metallic) asteroid to be given a close inspection. It is suspected that it was once a planetary core.

The Psyche spacecraft was intended to perform a 21-month primary science mission at Psyche, following its arrival at the asteroid after a six-year journey, which will include a gravitational-assist flyby of Mars. Psyche was based on the Maxar SSL-1300 platform, with the NASA Jet Propulsion Laboratory handling the development of all communications, command, and data handling subsystems, as well as flight software. Psyche had a launch mass of 2,610 kilograms (5,570 pounds); it used four SPT-140 Hall-effect ion thrusters, driven by the probe's twin solar panels, and had four science experiments:

• A high-resolution multispectral imager. It used a set of filters to identify metallic and silicate compounds on the asteroid's surface.
  
  
• A gamma-ray / neutron spectrometer, intended to map the elemental composition of the asteroid.
  
  
• A magnetometer, to see if the asteroid has any appreciable magnetic field that would suggest it was once a planetary core.
  
  
• A "gravity science experiment", with gravitational variations around the asteroid from shifts in the signal from the probe's X-band communications system.

It also carried a "Deep Space Optical Communications (DSOC)" laser communications link to test high-bandwidth communications for deep space missions. The mission's science objectives included:

• Determine whether Psyche is a core, or if it is unmelted material.
  
  
• Determine the relative ages of regions of Psyche's surface.
  
  
• Determine whether small metal bodies incorporate the same light elements as are expected in the Earth's high-pressure core.
  
  
• Determine whether Psyche was formed under conditions more oxidizing or more reducing than Earth's core.
  
  
• Characterize Psyche's morphology.

Following launch, Psyche will to cruise through space for five years and 10 months, with a gravity assist from Mars in 2026, arriving at 16 Psyche in August 2029. Once at 16 Psyche, the spacecraft will orbit the asteroid in four different orbits, labeled named "A:D", during its 21-month mission:

• The first orbit, Orbit A, will be 700 kilometers (435 miles) in altitude at a 90-degree inclination. Psyche will be in Orbit A for 56 days and will perform magnetic field characterization.
  
  
• Psyche will then move to Orbit B1, which is 303 kilometers (188 miles) in altitude at a 90-degree inclination. The spacecraft will stay in Orbit B1 for 92 days for continued magnetic field and topography characterization.
  
  
• Next, Psyche will lower to Orbit D, which is 75 kilometers (47 miles) in altitude at 160 degrees in inclination. Psyche will stay in Orbit D for 100 days and will investigate the chemical composition of the asteroid's surface.
  
  
• Psyche will then move up to Orbit C, a 190-kilometer (118-mile) high orbit at a 90-degree inclination. The spacecraft will be in Orbit C for another 100 days and will perform investigations into 16 Psyche's gravitational and magnetic fields.
  
  
• The final orbit Psyche will move to during its primary mission is Orbit B2, with the same altitude and inclination as Orbit B1. However, this time the spacecraft will stay in Orbit B2 for 100 days, performing more topography and magnetic field characterization.

16 Psyche is the largest M-type asteroid to ever be discovered, and its large size is one of the main reasons why scientists selected the asteroid for the Psyche mission. Its mean diameter is around 220 kilometers (137 miles). It is so large that it contains approximately one percent of the entire asteroid belt.

Psyche was the 14th mission in NASA's Discovery Program, which supports solar system exploration missions that are lower in cost than NASA's New Frontier and Flagship programs. It was selected in 2017, along with the companion Lucy asteroid mission.

[MON 17 JUN 24] THE WEEK THAT WAS 24

DAYLOG MON 10 JUN 24: I have a neighbor across the street who's a Trump fan. Today I noticed he was flying Old Glory upside-down, which I vaguely knew was a Trump sorta thing. Later I got the specifics: flying a flag upside-down is a distress signal, It's been done in many protest campaigns, the Trump variant being over the "sToLeN eLekShUN". I didn't need to know the specifics to be annoyed, and I wanted to do something about it -- but what?

I fly Old Glory on national holidays, when I remember. I think: *Why not now?* So I put it up -- then went over to my neighbors, Kent & Freya, and rang the doorbell. I asked Kent if he had a flag he could fly, and when he asked why, I pointed across the street: "Did you see that?"

"Oh yeah, geez, that!" I replied: "I was REALLY annoyed!" "He must be in distress." "Heh! Not as much as he's gonna be over the longer run." I suggested to Kent he fly his flag too, and he did later. I said to Kent: "I was thinking of putting a Confederate flag on his lawn -- but naw, this is plenty good enough." I'm not big on indignation, but I'm good with mockery.

* Spout's Chris Bouzy ran a poll on Spout asking if Joe Biden was responsible for the Gaza crisis. The poll came back with 2% UNDECIDED, 5% YES, and 93% NO. Social media polls are usually just for laughs, but this one rang true.

Spout is Left of Center, trolls not being welcome there, and the poll clearly reflected a legit sample of liberals. Only the Redlined Left would say YES, and 5% sounded like a valid number. If the poll had been on Xitter, JB would have done much worse, because the trolls would have voted him down.

DAYLOG TUE 11 JUN 24: The verdict came in on Hunter Biden today, and it was GUILTY. That was disappointing; I was hoping for a hung jury. The sentencing won't be for at least four months. I'm hoping he gets probation, but the judge does not seem lenient. On the plus side, although the matter is heartbreaking for JB, the verdict did him a lot of good: nobody can credibly use Hunter against him any longer. When JB was asked if he would pardon Hunter, the answer was a solid NO.

* Today, Ukraine publicly announced the introduction of its "Unmanned Systems Forces" military branch. It's not clear how the Robot Forces will work, but it does point directly to the dominance of robot weapons in the war.

To that end, much has been made of AFU drones hitting two advanced Su-57 fighters in Russia, hundreds of kilometers from the border of Ukraine. The interesting thing about that these aircraft were almost certainly at well-defended airfields, and they got hit anyway.

Reading between the lines suggests that the attack drones were accompanied by jammer drones and decoy drones. As Ukrainian production of drones ramps up, we'll be seeing more tricks. The Ukrainians have been crushing the Russian Black Sea Fleet with their sea drones, along with air drones. The Russians seem to be helpless to stop the drones -- suggesting that every attack introduces something new.

DAYLOG WED 12 JUN 24: The US has now announced a new set of sanctions on Russia, focused on Russian companies involved in the war, as well as foreign companies that help the Kremlin evade sanctions. It's hard to stop the flow of "dual-use" components -- the Russians just set up more front companies in intermediary countries -- but it is possible to make such components more expensive and their supply more uncertain.

* Regarding Hunter Biden's conviction on gun charges: Joe Biden has said Hunter will not be pardoned, but now appears to be leaving the door open for commutation. The judge in the case, Maryellen Noreika, seems determined to be harsh on HB but the vague hints of commutation may be to encourage her to sentence HB to home arrest / probation. Putting HB in jail is problematic, because he requires Secret Service protection -- he needs it -- and commuting the sentence to home arrest could be seen as merely prudent.

* I've mentioned repeatedly in the past that a Center-Right party is unlikely to emerge, because America's Center has shifted Left. Joe Biden is now the Centrist candidate. That leaves GOP "Moderates" with no place to go.

On further thought, it's even worse for them than that. All the Republican "Moderates" can do is support more or less the same sort of policies as MAGA, just not in as extreme a fashion. Voters do not care about weak-tea MAGA. The Republican Party is doomed.

DAYLOG THU 13 JUN 24: Today, the US Supreme Court judged against a group of doctors who were out to ban the "abortion drug" mifepristone -- the court saying that the doctors had no standing to press the suit, not having suffered any damages themselves. The doctors claimed mifepristone represented a health hazard, which is nonsense -- substantial adverse reactions only occur 0.3% of the time. However, more than a few states still ban mifepristone; the war is not over yet. I'm wishing it were over, but that won't be for a while.

* Reports by Western intelligence from Ukraine say the Orcs are getting hammered: 1,000 killed a day in May -- *killed*. That means Ukraine is getting new and effective weapons in serious quantity.

The puzzling question is: *What kind of new weapons?* Ukrainian operational security is very tight and nothing leaks out. The clues will come out of photos taken by the Orcs of fallen dud Ukrainian weapons -- there's always a few.

In the meantime, the USA having passed new sanctions, Russians have reacted with a wild rush on banks. It is impossible to say how much longer Vladimir Putin can hold out, but sooner or later he will abruptly cave in.

DAYLOG FRI 14 JUN 24: House Speaker Mike Johnson spoke of having constructive conversations with Donald Trump. This was leaving out a reported phone call in which Trump, in between a carpet of F-bombs, demanded that Johnson overturn Trump's conviction.

Of course, Johnson has no way of doing that. 80 US CEOs also listened to Trump speak at their Business Roundtable quarterly meeting, but it wasn't a "Trump meeting" by any means; Jeff Zients, Joe Biden's chief of staff, was also there to give a pitch -- and the big takeaway reported for the CEOs was that they thought Trump was senile, rambling like a "drunken uncle" at a family gathering.

* I was hoping that today SCOTUS would announce the decision on Trump's expansive claims of presidential immunity from prosecution. Instead, what we got was ... overturning the ban on bump stocks, which turn semi-automatic rifles into jury-rigged full automatic weapons.

All the GOP justices voted to overturn the ban. Words fail me. The only thing I can say is that, given the dismal state of America's gun-safety laws, this only makes the situation somewhat worse. Now I'm kinda worried about what other mad things they will do.

AND SO ON: I've been using Google Assistant -- running on a smartphone on a stand on my work table -- for over two years now, and I've found it extremely useful. Most of it's just trivial stuff, turning on or off the two lights I've got set up on wi-fi, and asking for the weather. However, I can also ask it to perform unit conversions -- kilometers to miles, for example -- and it's much quicker than fumbling with a calculator. It's also great for getting simple factoids, like the distance between two cities. It can generally answer simple questions, such as when someone was born. If the question is too complicated, I get linked to a web page for investigation. Just this morning, I asked Google Assistant when the Biden-Trump debate was going to happen, and she -- I use a British female voice -- replied (more or less): "Thursday, June 27, in Atlanta, with Jake Tapper and Dana Bash moderating."

These days, I'm also making much more good use of the Microsoft Copilot generative-AI chatbot. It is, despite limitations, extremely useful as well. In the old days, if needed some particular information, I'd try to Google it to get links to web pages that would allow me to piece together the information. Now I just ask Copilot for the information, and it gives me the answer, along with links if I want to use them. Of course, it only works if I pose an articulate question.

I'm finding out more I can do. For example, some Japanese-language tutorial web pages give me Japanese text, but not the "romanji" -- roman text -- equivalents. I can dope out the Japanese "kana", the phonetic set, but the Chinese-style "kanji" can have different readings, and it's a chore to figure them out. Some web pages use "furigana" -- painfully tiny kana characters above a kanji character to give its proper reading -- but some don't bother to even do that.

I finally decided to tell Copilot: "romanji: <pasted kanji>" -- and to my surprise, I got the romanji. I'm also surprised that I can ask fairly difficult questions on Japanese syntax and usage, and get useful answers. I even asked Copilot if it had anything to do with the OpenAI chatbot, OpenAI being backed by Microsoft -- but Copilot said it was derived from Microsoft in-house work, and warned me about some of its limitations.

[FRI 14 JUN 24] JOE BIDEN (30)

* JOE BIDEN (30): At Walter Reed, Joe was entrusted to the care of a Dr. Eugene George. Joe had been given an angiogram -- a scan to trace out the blood vessels in his head -- at Wilmington, but Walter Reed had better equipment, so Dr. George had it done again. The new angiogram located the troublesome aneurysm on the left side of Joe's head, and also found a smaller one on the right. Dr. George wanted to schedule the operation to deal with the left aneurysm immediately; the smaller one on the right could wait.

Dealing with it involved getting to the aneurysm and cutting it out, blocking off the blood vessel connections to it permanently. Blocking the connections wasn't a problem, since other blood vessels could take up the slack; the problem was that the operation was very delicate, with a fair chance of the patient dying in surgery, and a worse chance of a brain impairment, particularly a loss of speech, if the patient didn't die. Dr. George reassured Joe that he had done many similar surgeries before with good results, and that a highly skilled surgeon would assist. Nonetheless, Dr. George told Joe to speak with his family before the operation; he might not be able to do so later.

Jill went and got the two boys, who came to Joe's bedside, trying to be upbeat. Joe told them: "You're great young men, and I'm so proud of you. So I know you'll live up to your obligations. I know you'll take great care of your mother and your sister."

Beau replied: "Don't say that, Dad. You're not going anywhere."

Joe went on: "Look, I'm probably going to be fine, guys, but in the event something happens, you know what I expect of you. You take care of each other. You take care of your sister. You take care of your mom. I know you'll do it."

Joe was wheeled on a gurney to the operating room. He felt no fear, in fact he felt entirely calm; there was no action he could take to change matters, his fate was entirely in the hands of the doctors. They put him under general anesthetic and, after some confused thoughts, he was out like a light.

The next thing Joe knew, he was waking up to see the time on a clock, reading "4:00" -- in the AM or PM? He didn't know. Then he felt a kiss on his forehead, and a feminine voice asking: "Joe?"

"Jilly?"

"It's me. I'm here."

"Am I alive?" Jill didn't laugh; with his shaved, swollen, and surgically mutilated head, Joe looked like death warmed over.

She replied: "You're alive." He could see, he could talk, he wasn't paralyzed, he'd come through. On later discussing the emergency, the family concluded that giving up the presidential campaign saved Joe's life. It was touch-&-go as it was, and being on the road would have greatly complicated dealing with the crisis -- all the more so because Joe would have tried to keep on pushing. "You wouldn't be alive," Jill told him. Maybe so, maybe not, but the crisis would have still completely derailed his presidential campaign.

Joe was alive, but not out of the woods. As Jill wrote later:

QUOTE:

END_QUOTE

Jean was correct: legally speaking, a spouse had priority. In any case, Joe spent ten days recuperating at Walter Reed, with nothing to do but watch the monitors when he was awake, wondering if he was really going to pull through, wondering if he wanted to bother. When he was finally released, he was brought out on a wheelchair, to be picked up by Jimmy for the drive back to Wilmington. Joe was wearing a baseball cap to cover up the staples in his skull, making him more presentable to the gaggle of reporters waiting for him outside the hospital.

Joe was supposed to come back in a month or two, once he'd recuperated enough, to get the other aneurysm dealt with; the reporters asked him about the second surgery, and he replied that he expected it to be a "piece of cake". Dr. George had actually emphasized that the surgery was not at all trivial, with risks similar to those of the first, but Joe didn't want to dramatize his situation. Dr. George was then called by colleagues who browbeat him for misleading a high-profile patient. [TO BE CONTINUED]

[THU 13 JUN 24] GIMMICKS & GADGETS

* GIMMICKS & GADGETS: As discussed in an article from SCIENCENEWS.org ("Want A Shrinky Dinks Approach To Nano-Sized Devices?" by James R. Riordon, 18 January 2023), it is not easy to fabricate devices in the nanoscale. A new approach for that job is to shrink them down after building them, instead of making them small to begin with.

The key is spongelike hydrogel materials that expand or contract in response to immersion in chemicals. By carving patterns in hydrogels with a laser and then shrinking the gels down to about a 13th of their original size, the researchers created patterns with details as small as 25 billionths of a meter across.

Bioscientist Zhao Yongxin of Carnegie Mellon University in Pittsburgh and colleagues deposited a variety of materials in the patterns to create nanoscopic images of Chinese zodiac animals. By shrinking the hydrogels after laser etching, several of the images ended up roughly the size of a red blood cell. They included a monkey made of silver, a gold-silver alloy pig, a titanium dioxide snake, an iron oxide dog, and a rabbit made of luminescent nanoparticles.

Zhao says that the group also fabricated holographic patterns, then shrunk them down and rendered them unreadable. To read them, they had to be re-expanded to the same size. What excites Zhao the most about the work, however, is the wide range of materials that can be used: "We will be able to combine different types of materials together and make truly functional nanodevices."

* As reported in a press release from the Massachusetts Institute of Technology ("MIT's New Desalination System Produces Freshwater That Is 'Cheaper Than Tap Water'" by Jennifer Chu, 03 October 2023), there's nothing new about solar-power desalinizers: the sun heats the desalinizer to evaporate the salt water, fresh water accumulates on the black sloped roof, then trickles down to be accumulated.

Nothing could be simpler, but such traditional desalinizers are not very efficient, and they unsurprisingly tend to accumulate salt and clog up. MIT engineers, working with collaborators in China, have devised an improved solar desalinizer. The trick is that water circulates in and then out of the desalinizer, forming eddies as it goes -- which slows the flow and encourages evaporation. The new desalinizer actually has two vertical chambers, with the water vapor accumulating in the lower chamber, and multiple stages. It is more efficient and doesn't accumulate salt.

The device, if scaled up to the size of a small suitcase, could produce about 4 to 6 liters of fresh water an hour at low cost. The device should be reliable -- presumably, the only moving parts are solar-powered pumps.

* As discussed in an article from GIZMODO.com ("Apple Updates iMessage With a New Post-Quantum Encryption Protocol" by Lucas Ropek, 21 February 2024), Apple has now introduced an update to its "iMessage" encryption, which the company grandly calls its "most significant cryptographic security upgrade" ever.

The new scheme, labeled "Post-Quantum Level 3 (PQ3)" addresses a weakness in the current RSA method of "asymmetric encryption" -- that is, an encryption scheme in which a freely-distributed "public key" can be used to encrypt a message, but only a secret "private key" can then decrypt that message. The problem with RSA is that it is computationally dependent on multiplications of large prime numbers -- and the new emerging "quantum computers" are capable of rapidly factoring such multiplications back out again.

PQ3 uses a new, more secure asymmetric encryption scheme that can't be cracked by fast factoring It is based on a "post quantum secure" algorithm known as "Kyber", developed by researchers at the US National Institute for Technology and Standards (NIST). Apple modestly announced: "To our knowledge, PQ3 has the strongest security properties of any at-scale messaging protocol in the world."

The only difficulty with PQ3 is that quantum computers haven't really arrived yet. However, Apple is sending a message to its users that it is serious about security. When quantum computers do become common, they won't have an advantage in cracking PQ3. In the meantime, PQ3 provides security just as good, likely better, than RSA.

[WED 12 JUN 24] GEN-AI ARRIVES (18)

* GEN-AI ARRIVES (18): As discussed in an article from NATURE.com ("How does ChatGPT Think?" by Matthew Hutson, 14 May 2024), software has long proven hard to debug, the problem getting tougher as systems become more complicated. David Bau -- a computer scientist at Northeastern University in Boston, Massachusetts -- says: "I spent 20 years as a software engineer, working on really complex systems. And there's always this problem."

The arrival of artificial intelligence (AI) systems has raised the difficulty to a new level. Bau says that with traditional software, usually someone who was intimately familiar with it could dope out the problems. That's not so true any more, Bau saying: "Here's what really terrifies me" about the current flood of AI systems: "there is no such understanding", even among the people building it. Researchers often talk about AI systems as "black boxes", the inner workings of which are a mystery.

In the face of this difficulty, researchers have turned to the field of "explainable AI (XAI)", building up a bag of tricks and tools to help reverse-engineer AI systems. Methods include, for example, highlighting the parts of an image that led an algorithm to label it as a cat, or getting software to build a simple "decision tree" that approximates an AI's behavior. This helps to show why, for instance, the AI recommended that a prisoner be paroled or came up with a particular medical diagnosis. XAI is getting results, but it's still very much a work in progress.

Large language models (LLM), the AI engines that power chatbots like ChatGPT, are particularly troublesome to decipher, in part because of their size. LLMs can have hundreds of billions of "parameters", the variables that the AI uses to make decisions. The opacity of LLMs is very troublesome. People are using them to seek medical advice, write computer code, summarize the news, draft academic papers, and much more -- but everyone who uses them finds out that they can "hallucinate", producing misinformation, reflecting prejudices, and leaking private information.

XAI tools are being devised to take on LLMs. Researchers want to devise more reliable and trustworthy AI; users want explanations and clarifications to make sure they're getting valid results from a chatbot; and regulators want to know if an AI will go out of bounds, and in fact want to know where the bounds should be set in the first place. Martin Wattenberg -- a computer scientist at Harvard University in Cambridge, Massachusetts -- says that understanding the behavior of LLMs could even help us better understand human cognition.

Researchers have called LLMs "stochastic parrots" -- meaning that the models write by probabilistically combining patterns of text they've encountered before, without really understanding what they're saying. However, that leads to the question: *Don't humans work in much the same way?* It makes no sense to say that machines can't think like humans, if we don't really know how humans think, at a fundamental level, in the first place. Of course, as noted LLMs sometimes go badly off the rails -- but humans do, too.

A team at the AI company Anthropic -- based in San Francisco, California -- highlighted the reasoning powers of AI in a 2023 study that tried to backtrack why a chatbot says what it says. Anthropic's researchers scaled up a common XAI approach to probe an LLM that had 52 billion parameters to reveal which bits of the training data it used to answer questions. They then ran in through tests, a significant one being:

QUOTE:

END_QUOTE

The researchers traced the response to Lana del Rey's song "Lust for Life", the HAL 9000 computer in 2001: A SPACE ODYSSEY, and several other sources. The researchers described the model's behavior as "role-playing" -- more than parroting, but less than planning actions. The AI was, in effect, asked to tell a story, and it made up a story. If it had been simply told it was going to be shut down, it would not have protested. It doesn't have the capability of independent behavior given to an ant. [TO BE CONTINUED]

[TUE 11 JUN 24] SOLAR LIGHTSHOW

* SOLAR LIGHTSHOW: As discussed in an article from NATURE.com ("Dazzling Auroras Are Just A Warm-Up As More Solar Storms Are Likely, Scientists Say" by Alexandra Witze, 13 May 2024), our Sun has an 11-year activity cycle, and right now it's at a peak. In early May, one big solar flare caused auroras at unusually low latitudes, with images flooding social media.

The specific cause is a cluster of sunspots, known as "active region 3664 (AR3664)", that appeared below the Sun's equator on the side currently facing Earth. The cluster is about 17 times as wide as Earth. According to Shawn Dahl, -- a space-weather forecaster at the US National Oceanic and Atmospheric Administration's Space Weather Prediction Center in Boulder, Colorado -- It is one of the biggest and most elaborate sunspot region observed during the current solar cycle, which began in 2019.

Around 8 May 2024, AR3664 generated at least seven "coronal mass ejections (CME)", blasts of energetic plasma, in the direction of Earth at speeds of up to 1,800 kilometers (1,120 miles) per second. The CMEs swamped space-weather detectors, being rated as a "superstorm", and set off auroras when they collided with the Earth's magnetic field and atmosphere -- with gases such as oxygen and nitrogen becoming ionized and glowing in various colors. Auroras are typically only seen at high latitudes, but on 10 May they were seen as far south as Mexico.

The solar storm disrupted radio and GPS communications around the world. The SpaceX Starlink orbital internet constellation suffered some degradation of service, while electrical-grid operators took protective measures against currents induced in power lines by the superstorm. New Zealand's electrical-transmission service temporarily turned off some circuits around the country to prevent equipment damage.

NASA announced on 10 May that there was no threat to the four US and three Russian astronauts aboard the International Space Station. Some satellites did stop making scientific observations. For instance, NASA's Chandra X-ray Observatory was put into a safe mode before the solar storm. During the storm NASA's ice-measuring ICESat-2 satellite automatically stopped doing science when it experienced unexpected rotation -- apparently caused by drag due to the extension of the upper atmosphere in the face of the solar bombardment.

The European Space Agency's Solar Orbiter probe is nearly behind the Sun with respect to Earth, giving it a different view of the storm. AR3664 is now rotating off the side of the Sun seen from Earth and into the field of view of Solar Orbiter.

Researchers expect that the current solar cycle will persist for some time, the indicator being the number of sunspots they are observing. The biggest storms typically happen months, even years, after this official peak. Furthermore, as Dahl points out, as the solar cycle progresses, sunspots tend to appear closer to the Sun's equator, increasing the chances of CMEs that will hit for Earth and not fly out into empty space.

* Related articles discussed the colors of auroral displays. The color is predominantly green, in part because the human eye is less sensitive to other colors of auroras, and in part because of the altitude of the show. Different gases predominate at different altitudes and in varying concentration, with each gas fluorescing in a particular color in response to solar particle stimulation.

Most solar particles typically collide with our atmosphere at an altitude of around 95 to 240 kilometers (60 to 150 miles) where there are high concentrations of oxygen. When the oxygen is "excited" at these altitudes, it generates green light. At higher altitudes, above 240 kilometers, oxygen produces reds instead. This is because, at lower altitudes, when an oxygen molecule is stimulated by a solar particle, the molecule quickly collides with another gas molecule and then decays. At higher altitude the atmosphere is very thin; that gives an excited oxygen molecule time to decay to a lower energy state before releasing a lower-energy red photon.

An aurora may have a purplish lower border which comes from emissions from molecular nitrogen, at altitudes between 120 and 200 kilometers (80 to 125 miles). Nitrogen can either promptly decay after being excited, giving off blue light, or give off red light after a delay. Blue and red give purple. Very rarely, under intense solar activity, yellow and pink may show up -- as mixtures of red with green or blue.

[MON 10 JUN 24] THE WEEK THAT WAS 23

DAYLOG MON 03 JUN 24: Regarding the Ukraine War peace summit to be held in Switzerland in mid-month: of course Russia hasn't been invited and is discouraging all its other allies from attending. China is making it clear it is not interested in the event.

That means the peace summit will essentially be a propaganda exercise, intended to diplomatically isolate Russia and its allies. At the very least, Russia and China will not be able to credibly push their own bogus peace plans any more. US President Joe Biden will not attend, sending Vice President Kamala Harris and National Security Advisor Jake Sullivan instead. There have been some Ukrainian complaints about Biden's no-show, but it can be seen as an expression of confidence in the vice-president.

* Along parallel lines, the Biden Administration has promoted a Gaza peace plan and claimed it has Israeli backing -- but Prime Minister Benjamin Netanyahu has disavowed the plan, rejecting it as unsatisfactory.

It seems unlikely that the Biden Administration would claim Israeli ownership of the plan if no Israelis backed it, so it may well be the case that the USA is coming down on the side of Netanyahu's political opponents, becoming directly involved in Israeli government infighting.

Netanyahu is obviously under severe internal and public pressure. It is clear is that Joe Biden has lost patience with Netanyahu, and is leaning on him very hard. Where this goes is anyone's guess.

DAYLOG TUE 04 JUN 24: Hunter Biden's trial on gun charges in Wilmington DE just started, with jury selection. Hunter was there with his wife, Melissa Cohen-Biden, the two walking into the court hand-in-hind; First Lady Jill Biden was there as well.

Also in attendance was Garrett Ziegler, previously an aide in the Trump White House, and a prime mover in the harassment of Hunter Biden -- who is now suing Ziegler for his activities. Not much background on Melissa Cohen-Biden is available, but all indications are that she is formidable. She confronted Ziegler outside the courtroom, pointed a finger at him, and said in a loud voice: "You have no right to be here, you Nazi piece of s***!"

She says that Ziegler had, in the course of his attacks on Hunter, smeared her in anti-semitic terms as well. Ziegler denied everything -- but then he would, wouldn't he? Trump and his people are not noted for either civility or truthfulness. Besides, she was right, Ziegler was not an interested party in the trial; he was only there to troll Hunter Biden.

As discussed previously, it is unlikely Hunter Biden will do time on the gun charge. Given that the Biden name is solid gold in Delaware, the jury is likely to be sympathetic.

* Much speculation on Spout about whether Donald Trump will go to jail when Judge Merchan sentences him on 11 July. We'll see what happens -- but I really wish the judge ends the sentencing with: "YOU'RE FIRED!"

DAYLOG WED 05 JUN 24: Dr. Anthony Fauci testified to Congress this week, with Congressional MAGA assailing him for "mishandling" the COVID-19 pandemic. It's the same old trash, attempting to weaponize the pandemic against the Democrats.

These days, it seems MAGA Republicans are trying to shift the blame for their mishandling of the pandemic, which unnecessarily cost hundreds of thousands of Americans their lives. In reality, for the public the pandemic is old news, and the fuss is pointless.

DAYLOG THU 06 JUN 24: I got my Colorado mail-in ballot yesterday -- it's automatically sent out to all registered voters -- for the primary election. I filled it out today, and I'll drive over to the drop-box at the county office campus tomorrow to deposit it.

This election was different for me in that I used Microsoft Copilot to check out the candidates. I'd type in: "Background: <candidate_name> running for Colorado <public_office>." -- and get back a nice summary. It was somewhat superfluous, since I'm going to vote for all Democrats and no Republicans. Incidentally, Colorado has open primaries, I got two ballots, one for each party -- but I can only use one of them. Still, I felt reassured on finding background for the people I voted for.

The ballot also included a special election, to replace Ken Buck in the House in Colorado's 4th District. I voted for the Democrat, of course. The winner will only serve to the end of this Congressional session in January. More interesting is the battle between Lauren Boebert and Ike McCorkle for the same seat in the November general election. Boebert is currently sinking badly, but we'll see what happens. What is particularly interesting that McCorkle is a rock-solid US Marine veteran -- whose campaign page pledges, among other things, to back a revived Equal Rights Amendment, focused on women's and LGBT rights.

That from an ex-Marine gunny sergeant -- he ended as a captain, however -- who served four combat tours. (I got that out of Copilot, too.) I see in him a glimpse of the future: traditionally, a tough ex-Marine would stereotypically be a Rightist, favored by rural folks, but what we're going to see more of is the same sort of guys (and gals) who are definitely Blue in color, and rural folks will still vote for them. In this vision of the future, the GOP amounts to nothing.

On consideration, it's not so surprising that stalwart ex-military are turning Blue. There is no more integrated component of American society than the US military; these days it's even accepted transgenders. Anyone in the ranks who has a problem with DEI will be told to get with the program.

DAYLOG FRI 07 JUN 24: I dropped off my ballot yesterday at the country office campus this AM -- noticing a sign next to the ballot box that I hadn't noticed before, saying in effect NO ELECTIONEERING OR GUNS NEAR HERE. Colorado authorities, it seems, are tired of voter interference.

* It appears the MAGA Right is saying: "U lIbz wIlL B rEaLly sOrRy iF jUdGe MeRChaN tHrowZ TrUmP N jAiL!!!" I think in reply: *I don't believe we will.* I suspect they will be disappointed at how little protest follows. There will, however, be noisy celebrations.

* I'm in my 70s now, and my short-term memory is getting dodgier. I was vaguely thinking for a long time of getting a reminder app for a smartphone, and see if that helped. I finally took the time to figure it out.

I originally thought of a voice recorder app for taking memos, but on checking such things out, it became apparent they wouldn't work very well. Checking the notes would mean having to listen through them, which would be time-consuming. It was much better to get an app that converted voice to text notes. After downloading and trying several apps, I finally found "Voice Notes".

It works easy: (1) I bring up the app and immediately see a list of short notes I've already made, and can check out at a glance. (2) I touch a button to make a note, I get a pop-up to tell me to talk, and tell the app my comment. Nice feature: I get a list of possible readings of the voice input to select from. (3) Having selected the right reading, I touch the "back" button to go back to the notes list. (4) I can select an existing note to add to it, edit it, or delete it. Simple as a brick. I've only been using it for a little while, but it seems really effective.

So now, I've got my smartphone in my pocket or in arm's reach all my waking hours. Not only do I remember things better, but obtaining Voice Notes led to a general tightening-up of procedures all down the line. I'll see how well things work over the longer run.

AND SO ON: Back in 2016, a British intelligence wonk named Christopher Steele built up a file describing the Trump presidential campaign's contacts with the Russians. The "Steele Dossier", as it became known, heightened FBI interest in the Trump campaign -- and led to a backlash from MAGA trolls, who denounced it as a vicious, fraudulent attack on Trump.

Steele himself said the file was no more or less than a collection of unverified tips that needed validation, and he had no intention of making it public -- but it was leaked. Trump eventually sued Steele for defamation in British court; the suit was dismissed as "without merit", with Trump then slapped with a fine of 300,000 GBP for the defendant's legal fees. He appealed, and the appeal has now been rejected. He says he'll appeal again, but it's hard to believe British courts will accept the appeal.

Such bringdowns of Trump are often played up on The Palmer Report (PR), a blog run by one Bill Palmer and friends. This last week the PR reported how the Trump gang is playing up a Facebook posting by one "Michael Anderson" -- stating that his cousin was a juror in the Trump trial, and she had informed him during the trial that Trump was going to be convicted. Of course, Trump is screaming: "MISTRIAL!" -- but it's foolish. Anybody can say anything on Facebook, and can be automatically assumed to be bogus until proven different.

The PR shows Trump no mercy, as of late focusing on his mental decline. I tend to think they've got it turned to 11, but what can I say? Turn it down to 9? Not all the bad things said about Trump are true, just most of them.

One Robert Harrington, posting on PR, stepped back to look at the bigger picture, describing how Flat Earthers and Moon Hoaxers can only maintain their goofy ideas by assuming massive conspiracies that have perfect security. As Harrington points out, the current political situation shows that isn't how the real world works:

QUOTE:

END_QUOTE

There's still some caution about assuming the win in November, I think partly driven by the feeling that the Republicans can't really be in as sorry and completely ruined condition as they look. On second thought, however, I'm reassured: *But they are.* The Republican Party today looks like a sap from a Looney Toons cartoon who has wandered off the edge of a cliff -- but will only fall when he looks down.

[FRI 07 JUN 24] JOE BIDEN (29)

* JOE BIDEN (29): Joe went back on the road in February 1988, the focus being to talk about the current arms-control efforts. The INF Treaty wasn't controversial, in fact it was Reagan at his best; for all his reputation as a Rightist troglodyte, he could be pragmatic and conscientious.

Along with the INF Treaty, however, Reagan was also pushing a "Strategic Defense Initiative (SDI)", popularly known as "Star Wars", to develop a defensive "shield" for America to defeat a nuclear attack. It demonstrated Reagan's conscientious nature again, reflecting his distaste for the established doctrine of "Mutual Assured Destruction (MAD)" -- or in simple terms, if the Soviets attempted to destroy America with their nuclear weapons, America would destroy the USSR in turn.

Reagan correctly saw MAD as completely unethical, a monstrous proposition. The problem was that building a shield against nuclear attack was not only extremely difficult and promised to be very expensive, it was impossible to make it leakproof. A kill ratio of 90% would be very good -- but if the Soviets launched a hundred missiles at the USA, that still meant the loss of ten American cities, which would be unacceptable damage. Since the missiles could in principle carry multiple warheads, as well as decoys, they could likely do a lot more damage than that. The Soviets were very frightened of Star Wars, seeing it as work towards systems that would allow the USA to destroy the Soviet Union with impunity; they were greatly over-reacting, but the bottom line was that Star Wars gave them an incentive to build up and refine their nuclear strike force.

Joe spoke at the University of Scranton, in his old hometown, on 9 February, and then spoke at the University of Rochester in New York the next day. He got an enthusiastic reception, being given many questions about arms control; he got some questions about his presidential campaign, but it wasn't really an issue. The presidential campaign was moving on, and he was old news on that score.

Joe got back to the hotel late; he was so happy and pumped up that he couldn't sleep, so he thought of calling out for a pizza. The next thing he knew, he woke up on the floor. He didn't know where he was for a moment, then recollected he'd been hit with a staggering lightning bolt of pain that put him down. He looked at the clock; he'd been out for about five hours.

Joe had been traveling with Bob Cunningham, a friend from Wilmington, who managed to get him on the airplane and back to Wilmington. One of Joe's aides, Tommy Lewis, picked the two up at the airport, to drive them to Joe's house. Joe looked terrible, but he insisted: "I'll be fine. Just get me to bed." Obviously, he was not thinking clearly. They did get him home, but he collapsed again. He was dimly aware of Jill coming home, and then being taken to Saint Francis Hospital in Wilmington as snow was falling.

The doctors told Jill they were going to give Joe a spinal tap, which would take a while. Jill left to get a babysitter for Ashley and call Hunter. When she got back to the hospital and went to Joe's room, a nurse outside the door told her: "Oh hi, Mrs. Biden -- don't go in right now, they're giving him last rites." Jill went in and interrupted the proceedings, shouting at the priest: "Get out of this room! Get OUT!" After things settled down, she was told that the results of the spinal tap were dire: there was blood in Joe's spinal fluid, meaning something in his brain was bleeding into it.

Further tests showed that Joe had suffered an aneurysm, meaning a blood vessel that was swollen up, leaking blood, and ready to burst. The problem could be surgically corrected, but it was tricky. The Biden clan -- Joe SR and Jean on down -- had congregated at the hospital, with Jimmy calling around to find the best place to have the surgery down. That turned out to be Walter Reed Army Medical Center, outside Washington DC.

Joe was packed into an ambulance to be taken to Walter Reed, while the snow was coming down hard. The ambulance had a Delaware State Police escort, Beau riding with the troopers. At the state line, the Delaware State Police passed the escort over to Maryland State Police, with the hand-off resulting in confusion. Jill ended up shouting at the mob: "Dammit! Move this ambulance!" Jill rarely used harsh language; whether that had any particular effect is unknown, but things got sorted out one way or another, and the ambulance was on its way. [TO BE CONTINUED]

[THU 06 JUN 24] SCIENCE NOTES

* SCIENCE NOTES: As discussed in an article from CBS.com ("Super-Puff Exoplanet Is As Big As Jupiter But 10 Times Lighter" by Sophie Lewis, 19 January 2021), astronomers have discovered a big but lightweight "super-puff" or "cotton candy" planet about 212 light years from Earth.

This super-puff exoplanet, designated "WASP-107b", is about the same diameter as Jupiter, but only about one-tenth the mass -- or about 30 times more massive than Earth. The discovery, made by doctoral student Caroline Piaulet under the supervision of professor Bjoern Benneke at the University of Montreal, suggests that gas giants form much more easily than previously believed.

WASP-107b was first spotted in 2017. It is found in the Virgo constellation; the planet is very close to its star, over 16 times closer than Earth is to the Sun, completing one orbit every 5.7 days. WASP-107b is one of the least dense exoplanets scientists have ever found. They determined the mass by observing spectral shifts in the light from the parent star from the gravitational hookup between the star and the planet. They were startled to find out the solid core of WASP-107b has a mass that is no more than four times that of the Earth, meaning more than 85% of its mass stems from the thick gaseous layer surrounding the core.

This is much different than with Neptune, which has a similar mass but holds just 5% to 15% of it within its gas layer. Based on their knowledge of Jupiter and Saturn, scientists previously believed that a solid core at least 10 times the mass of Earth would be needed to acquire enough gas for a gas giant planet to form. Professor Eve Lee -- an astronomer in the Department of Physics and the McGill Space Institute at McGill University -- says:

QUOTE:

END_QUOTE

A second, more massive planet was also discovered in an eccentric orbit beyond that of WASP-107b. The planet, designated "WASP-107c" a mass about a third of that of Jupiter, and orbits the star once every 3 years. Pialet says:

QUOTE:

END_QUOTE

The research team is now using the Webb Space Telescope to uncover more details.

* As discussed in a press release from Virginia Technical Institute / VTI ("Submerging Skylines"), research by VTI and the US Geological Survey / USGS has shown that major US Atlantic coast cities are sinking at rates faster than global sea level rise, with areas like New York City, Long Island, Baltimore, Virginia Beach, and Norfolk experiencing significant land subsidence, up to about five millimeters per year.

Population centers such as New York City and Long Island, Baltimore, and Virginia Beach and Norfolk, are seeing areas of particularly rapid subsidence, alongside more slowly sinking or relatively stable ground. The rapid subsidence poses a risk to roadways, runways, building foundations, rail lines, and pipelines. Research lead Leonard Ohenhen -- a graduate student working with Associate Professor Manoochehr Shirzaei at Virginia Tech's Earth Observation and Innovation Lab -- says:

QUOTE:

END_QUOTE

Shirzaei and his research team obtained a mountain of data points obtained by space-based radar satellites, using the measurements to build digital terrain maps that show where sinking landscapes present risks to the health of vital infrastructure. These maps show that a large area of the East Coast is sinking at least 2 millimeters per year -- with several areas along the mid-Atlantic coast of up to 3,700 square kilometers (1,400 square miles), sinking more than 5 millimeters per year. For contrast, the current global rate of sea level rise is 4 millimeters per year. Ohenen says:

QUOTE:

END_QUOTE

[ED: As discussed here in the past, the extraordinary subsidence in some northerly East Coast regions is due to the slow sinking of land that was raised by glaciation -- which by pressing some land down, raised land beyond the edges of the glaciers -- in the last Ice Age, while that of Norfolk and other locales around Chesapeake Bay is to the subsidence of locales around the edge of the asteroid impact crater that established the bay.]

[WED 05 JUN 24] GEN-AI ARRIVES (17)

* GEN-AI ARRIVES (17): Collin Lynch of NC State emphasizes that it's crucial that any chatbot used in education is carefully checked for its tone, as well as accuracy; it should not insult or belittle students, or make them feel lost. Lynch says: "Emotion is key to learning. You can legitimately destroy somebody's interest in learning by helping them in a bad way."

What complicates matters, as Khan's Kristen DiCerbo points out, is that Khanmigo responds differently to each student in each situation. That makes it harder to nail down, but she believes it also makes the bot more engaging than previous tutoring systems. Khan Academy expects to share its research on Khanmigo's efficacy in late 2024 or early 2025.

Other tutoring companies are offering LLMs as assistants for students or are experimenting with them. The education technology firm Chegg in Santa Clara, California, for example, has introduced an assistant based on GPT-4. TAL Education Group, a Chinese tutoring company based in Beijing, has created an LLM named "MathGPT" that the company says is more accurate than GPT-4 at answering maths-specific questions. MathGPT also aims to help students by explaining how to solve problems.

Another approach to creating an AI learning partner integrates the LLM with outside sources of knowledge, such as a textbook or a set of scientific papers, that have been thoroughly verified. The goal of this "retrieval-augmented generation (RAG)" method is to get around the extreme difficulty of verifying the billions of sources of text that give an LLM its conversational power. A RAG LLM becomes instead a traceable animated version of its source texts.

AI company Merlyn Mind in New York City is using RAG in its open-source "Corpus-qa" LLM, which is targeted at education. Like ChatGPT, Merlyn Mind's LLM is initially trained on a big body of text not related to education specifically, giving it conversational ability. However, according to Satya Nitta -- the CEO of the country -- when Corpus-qa answers a query, it prioritizes knowledge from its prime reference sources, which minimizes hallucinations and other errors. The LLM will also apologize for having a low-quality response, though that is increasingly normal for GAI chatbots.

Claire Zau -- vice-president of GSV Ventures, an investor in educational-technology companies in New York City -- says that Arizona State is also using RAG, describing ASU as one of the most progressive universities for LLM adoption. ASU has developed a toolbox that allows its faculty members to experiment with LLMs in education through a web interface. The toolbox provides access to six LLMs, including GPT-3.5, GPT-4 and Google's Bard, as well as RAG capabilities. Researchers are using the toolbox to build chatbots for their students to use.

Elizabeth Reilley -- ASU's executive director of AI acceleration -- says the bots are already having a positive impact on education at the university. For example, she says, a bot created for use in ASU's introductory chemistry course uses RAG to combine GPT-3.5 with PDF and PowerPoint course materials. She gives an example of a test that imagined a baseball-loving student asking the LLM for an explanation of dipole-dipole interactions in molecules based on that sport, with the chatbot providing an answer in a way that the student can more easily grasp.

Danielle McNamara -- executive director of ASU's learning engineering institute in Tempe -- says that using a general LLM combined with RAG differs from earlier machine-learning approaches, which attempted to train an AI system to simulate a science expert. The earlier ML tools lacked generalized capabilities, such as the capacity to incorporate baseball into chemical concepts, that could help students.

Other institutions are also embracing LLMs. Vanderbilt University in Nashville, Tennessee, which has given students on certain courses access to a paid version of ChatGPT, including access to specialized plug-in tools. Researchers at East China Normal University in Shanghai have created a dedicated open-source educational LLM -- not based on an available LLM like ChatGPT or Bard -- named "EduChat" that combines essay assessment, dialogue-based tutoring and emotional support in one chatbot.

There is, of course, the issue of who pays for educational chatbots. DiCerbo says Khan Academy is now looking for philanthropists and grants to help to pay for computing power and to provide access for under-resourced schools, having prioritized such schools in the pilot phase. She says: "We are working to make sure that digital divide doesn't happen."

Another challenge is to ensure that the information LLMs provide is not biased, and that the models consider knowledge and viewpoints from under-represented groups. Given that source materials inevitably reflect majority views, this is a troublesome issue. Sean Dudley -- ASU's associate vice-president for research technology -- says that RAG allows ASU's LLM platform to provide users with the sources of its answers. That doesn't eliminate the bias problem, but it at least provides transparency. Referencing is also becoming a norm for other chatbots.

How well LLMs work out in education remains to be seen. Lynch accepts that they are powerful tools, but also seeks to keep their shortcomings in focus. He says: "It isn't like overnight we've learnt to fly." [TO BE CONTINUED]

[TUE 04 JUN 24] DIGITAL ESTONIA

* DIGITAL ESTONIA: As discussed in an article from CNBC.com ("How A Tiny Country Bordering Russia Became One Of The Most Tech-Savvy Societies In The World" by Elizabeth Schulze, 8 February 2019), the country of Estonia, with a population of only about a million, is one of the most digitally wired countries in the world: taxes can be fully handled online, almost all public services are available on the web, and citizens can vote on the internet.

When Estonia became independent from the Soviet Union in 1991, the country embarked on a series of reforms to modernize the economy, with an emphasis on digital technology. Kersti Kaljulaid -- president of Estonia from 2016 to 2021 -- says:

QUOTE:

END_QUOTE

One of the first efforts was in education, with the government putting computers in every classroom; by 2000, all of Estonia's schools were online. The government also offered training to bring adults up to speed on using computing, with the result that the percentage of Estonians using the internet went from 29% in 2000 to 91% in 2016.

In 2002, Estonia launched a national ID system. Physical ID cards are paired with digital signatures, to allow Estonians to securely pay taxes, vote, do online banking, and access their health care records online.

Another major Estonian innovation is "e-Residency", a scheme that allows individuals to start businesses in the country without living there. The program serves as a launching pad for companies wanting to do business in the European Union (EU) and benefit from the EU's single market. Tens of thousands of people from around the world have taken advantage of e-Residency since it launched in 2014. It was complemented by a "digital nomad visa" system, which gives digital workers the right to live outside their home country but still virtually work there.

Estonia is well-known for its startup tech sector. The most famous example is Skype, the video chatting service that was bought by Microsoft, which was launched in Estonia in 2003. It has a very large number of tech "unicorn" companies per capita.

Of course, Estonia's digital environment is affected by the menacing presence of Russia directly to the east. Russian dictator Vladimir Putin has made it clear that he wants to bring the Baltic States back under Russian control, with the Russians performing cyber attacks on them. In 2007, Estonia suffered a massive cyber attack that brought down most of its digital infrastructure.

Since then, the Baltics have focused on cyber defense. Estonia became home to the NATO Cyber Defense Center of Excellence, which conducts large-scale cyber defense drills. The government also set up a "data embassy" in Luxembourg where it stores copies of all of its data. Along with a highly professional data security organization, Estonia works on educating the public in cyber defense, from kindergarten up.

* Along related lines, I was playing a videogame I downloaded off the Steam service, to find to my surprise that it was made in Poland. As it turns out, Poland is home to a substantial community of videogame developers that do a world-class business -- the best-known player being CD Project, maker of CYBERPUNK 2077. Apparently, the videogame community there got started in the 1980s, trading in bootleg foreign computer games, and after the fall of the Iron Curtain "went legit" to start producing their own titles.

[MON 03 JUN 24] THE WEEK THAT WAS 22

DAYLOG MON 27 MAY 24: On Saturday, Donald Trump pitched himself at the Libertarian National Convention at the Washington DC Hilton. It did not go at all well for him; he did get loud cheers, but it was hard to sort them out from the loud boos.

It was not exactly a case of Libertarians embracing sanity, with signs being waved around saying END THE FED -- Libertarians hate the Federal Reserve and tend to like Bitcoin -- and signs saying FREE ROSS -- meaning Ross Ulbricht, creator of the notorious "Dark Web" site Silk Road. Ross is now serving life in prison. Senator Mike Lee, a hardcore Trump supporter, also spoke to the crowd, and was booed at least as much. Joe Biden was invited to speak at the convention along with Trump, but wisely decided not to waste his time.

Libertarians, as I have said repeatedly on Spout, have always had mixed feelings about Trump. He's like them in many ways, except that he has no grand intellectual pretensions. They do not like seeing that the Emperor really has no clothes.

DAYLOG TUE 28 MAY 24: Reports are coming in from Ukraine that the Orc offensive towards Kharkiv is bogging down in its own blood. Reports also indicate that it jumped off with about half the troops needed for success.

Putin, it appears, realized that once US aid started flowing to Ukraine again, he wouldn't have a better chance in the future. Aid is not merely being resumed, it is being enhanced. Allied weapons production is ramping up slowly, with more effective weapons being supplied.

In the meantime, Switzerland is hosting a peace summit for two days in mid-June. Since Putin does not want peace except on his own imperialistic terms, the question arises: *Why bother with peace negotiations?*

China has been invited, and it may be the case that China is the target of the exercise. China wants to be seen as a world leader, and bringing about a peace settlement would greatly enhance Chinese prestige. China is holding all the cards in their dealings with Putin.

Many European leaders will attend the summit; Joe Biden will not. The Ukrainians have complained, but how sincerely? Might it all be a show, allowing China to "upstage" its rival, the USA? The USA could accomplish nothing at the summit. We'll see what happens.

DAYLOG WED 29 MAY 24: There has been considerable frustration in Ukraine and among Ukraine's Allies over Vladimir Putin's ability to fund his war of aggression, with economic sanctions often being called a failure. As an article from NEWSWEEK explains, not really.

Sanctions cannot be airtight; they can, however, be stifling. Consider the difficulties of Russia's gas supply company, Gazprom. At the outset of the war, Putin believed he could use Russian gas as a weapon against European countries aiding Ukraine. Unfortunately for Putin, most of the European countries managed to obtain substitute sources; Austria and Hungary still remain dependent on Russian gas, but they are not enthusiastic supporters of Ukraine. Russia's European market for gas has dried up, it seems permanently.

Gazprom lost billions of dollars in 2023. China remains a market, but infrastructure to do it right -- a pipeline -- would cost about $100 billion USD. China is not giving Russia a deal on gas purchases, and in fact the Russians may be selling at a loss.

Putin is keeping his war going by desperate economic measures, in effect burning all the furniture in the house to keep it warm. He survived 2023 by such means and is able to continue the war in 2024 -- but will he run out of money in 2025? We'll see.

* Trump's "hush-money" court case has now gone to the jury, which will deliver a verdict in a few days. The buzz on Spout was anticipation of a GUILTY verdict, but commented that I was simply "waiting patiently" -- adding: "I've learned that's the only appropriate mindset for observing the workings of the law." -- adding after that: "I'm not worried. Trump ... he's worried."

That got good feedback, with me throwing in a parting shot: "Trump lives in a self-glorifying fantasy world. Reality is intruding on it."

DAYLOG THU 30 MAY 24: This afternoon, reality intruded on Donald Trump's fantasy world in a big way, when a jury found him guilty of 34 counts of falsifying business records. I was trying to play it cool, saying I was content to be patient, but I did get anxious as today went on. Nobody was saying Trump was going to be acquitted, but there was the possibility of a hung jury. Likely not, Noo Yawkers do not like Trump much, and Judge Merchan gave the jury highly detailed and explicit instructions that, the legals said, made a GUILTY verdict almost inescapable. Nonetheless, Trump has got out of tight spots in the past, though these days not nearly as often as he used to.

With the announcement of GUILTY on Spout, something of a party followed. As one of the hundreds of posters, I put up a GIF of Obama doing his famous "mic drop", adding a caption: *BOOM CHAKALAKA!* I got LIKES out of that.

A Brazilian asked if Trump was likely to go to jail, but I said the murmur among the legals was that Trump would probably get a big fine and probation. How campaigning for president would work under probation is an interesting question. I added that sentencing was up to Judge Merchan, and Trump had greatly flamed off the judge -- so Trump might get some time in jail just to teach him a lesson. Sentencing won't be until well into July -- about the same time as the GOP national convention.

Incidentally, Brazilians got Spout's Chris Bouzy to come out with a Portuguese version of the Spout app, so Brazilians are now common on Spout. BR.GOV hasn't signed on yet, however.

DAYLOG FRI 31 MAY 24: To no surprise, the GUILTY verdict slapped on Donald Trump yesterday set off the MAGA faithful in a big way -- some trying to claim it was really a "big win" for Trump in disguise. *Yeah sure.*

Congressional annoyance Marjorie Taylor Greene defiantly stood up for Trump, saying she would unfollow anyone who suggested replacing him as the Republican nominee. That got over 20,000 replies saying, in one way or another, REPLACE TRUMP.

I'm not understanding why people are still afraid of Trump. Bill Maher was commenting that Trump was more dangerous than ever, and will try to steal the election in 2024. I doubt that's any threat, because the people who tried it in 2020 have largely been busted.

There was another comment on Spout about the "undecided voters" going into the election, who I had to compare to the "Bigfoot voters", saying: "Anyone who could say it was hard to choose between Trump and Joe Biden is not a serious or trustworthy person."

In any case, now we wait on the sentencing. I'm hoping Trump gets at least a few days of jail time. I want to see him in an orange prison suit for real, and it will take him down a peg, or more. Backlash? That's the fun part: there won't be a backlash to speak of.

The IRISH STAR of Dublin released an article in response to the judgement titled: "Donald Trump Supporters Threaten Civil War As US Braces For Wave Of Violence". From their side of the Pond, they didn't hear the sound of all the partying quite right.

AND SO ON: The Israelis, clearly under pressure from the US and Europe, have released a peace proposal to end the fighting in Gaza. The proposal envisions three stages:

• Stage 1: A six-week ceasefire in which the Israel Defense Forces (IDF) would withdraw from populated areas of Gaza. There would also be a surge of humanitarian aid, as well as an exchange of some hostages for Palestinian prisoners.
  
  
• Stage 2: All remaining living hostages would be returned, including male soldiers. The ceasefire would then become permanent.
  
  
• Stage 3: The remains of all deceased hostages would be returned. A reconstruction plan would go into effect, based on US and international assistance to rebuild homes, schools and hospitals.

Israeli Prime Minister Benjamin Netanyahu says the war will end when all objectives are achieved, including the return of all hostages and the elimination of Hamas' military and governing capabilities. He says the plan is consistent with these principles. Hamas is responding positively to the proposal, but nobody thinks it's going to be easy.

For myself, I've ceased to take sides in the fighting, instead just wanting the shooting to stop. Once it stops, cleanup begins. That will be hard. In any case, I want the problem to go away. We've got other things to worry about.