* This weblog provides an "online notebook" to provide comments on current events, interesting items I run across, and the occasional musing. It promotes no particular ideology. The current daily installment is available HERE. For update notices, follow @gv_goebel on SPOUTIBLE.
* JOE BIDEN (26): On the morning of Wednesday, 23 September 1987, Joe drove to Washington DC, accompanied by Jill and two aides. The star of the hearings that morning was retired Chief Justice Warren Burger -- a Republican, Center-Right, liberal in some ways, staunchly conservative in others. After discussion, Joe got down to the heart of the matter with Burger:
BEGIN_QUOTE:
Notwithstanding the appropriate caution against reading into the Constitution rights not explicitly defined, the Court has acknowledged that certain unarticulated rights are implicit in enumerated guarantees. For example, the rights of association and the right of privacy, as well as the right of travel, appear nowhere in the Constitution or Bill of Rights; yet these important, unenumerated and unarticulated rights have nonetheless been found to share Constitutional protection with explicit guarantees.
... Fundamental rights, even though not expressly guaranteed, have been recognized by the Court as indispensable to the enjoyment of rights specifically defined. Mr. Justice, that is what this debate is all about, at least with Judge Bork and I. And I wonder if you could speak with us a little bit, educate us a little bit, about these unenumerated rights -- the right of privacy?
END_QUOTE
Burger replied:
BEGIN_QUOTE:
... I did not come up here to give a lecture on constitutional law; you have had quite a lot of that in the last few days, I understand ... I see no problem about [your] statement, and I would be astonished if Judge Bork would not subscribe to it. I have no thought that he would disagree with what I said in those 1st Amendment cases.
END_QUOTE
Joe was startled at that claim, since Bork clearly did disagree with it. Joe went on with Burger, asking him if he affirmed the 9th Amendment, with Burger answering somewhat vaguely. Joe left it at that and yielded to Strom Thurmond.
The hearings dragged on into a third week, but Bork's chances were obviously fading. On 6 October, the Judiciary Committee took a vote, with nine voting against confirming Bork, five voting for confirming him. Arlen Specter of Pennsylvania was the only Republican to vote NAY; he was notoriously independent, having been a Democrat to 1965, and would much later become one again.
That didn't end the proceedings, with the Bork nomination then moving towards a full Senate vote. Bork and the Reagan White House went all-out to save the nomination, but it couldn't be saved: the Senate voted on it on 23 October, with the Bork nomination rejected by 58:42, an overwhelming margin. The conclusion of the Bork hearings eased the sting of the failure of Joe's presidential campaign, and also help establish that Joe was not an intellectual lightweight. However, it was Bork who, again, had hanged himself, Joe much later saying that "his arrogance was his undoing."
Bork was of course resentful of the way the hearings had turned out -- saying they couldn't be called "hearings", because nobody heard what he was trying to say. He did not seem to understand that his nuanced read on his judicial views simply looked like tap-dancing around vital issues. He was not so unhappy with Joe, complaining more loudly about the failure of the Reagan White House to push through the nomination. Bork went back to the DC Court of Appeals, but resigned a few months later. Financially, the hearings left him in good shape, with a big book advance and lucrative speaking engagements. He ended as a high-profile legal scholar, being associated with Rightist think-tanks. [TO BE CONTINUED]
COMMENT ON ARTICLE* Space launches for April included:
[02 APR 24] USA VB / FALCON 9 / STARLINK 7-18 -- A SpaceX Falcon 9 booster was launched from Vandenberg SFB at 0230 UTC (old day local time + 7) to put 22 SpaceX "Starlink v2 Mini" low-Earth-orbit broadband comsats into orbit.
[02 APR 24] CN XC / LONG MARCH 2D / YAOGAN-42 01 -- A Long March 2C booster was launched from Xichang at 2256 UTC (next day local time - 8) to put a secret "Yaogan-42 01" payload into orbit. It was apparently a military surveillance satellite.
[05 APR 24] USA CC / FALCON 9 / STARLINK 6-47 -- A SpaceX Falcon 9 booster was launched from Cape Canaveral at 0912 UTC (local time + 4) to put 23 SpaceX "Starlink v2 Mini" low-Earth-orbit broadband comsats into orbit.
[07 APR 24] USA VB / FALCON 9 / STARLINK 8-1 -- A SpaceX Falcon 9 booster was launched from Vandenberg SFB at 0225 UTC (previous day local time + 7) to put 22 SpaceX "Starlink v2 Mini" low-Earth-orbit broadband comsats into orbit.
[07 APR 24] USA CC / FALCON 9 / BANDWAGON 1 -- A SpaceX Falcon 9 booster was launched from Cape Canaveral at 0212 UTC (previous day local time + 4) on the "Bandwagon 1" mission, to put 11 smallsat payloads into space. The payloads included:
[09 APR 23] USA CC / DELTA 4 HEAVY / NROL 70 (USA 353) -- A Delta 4 Heavy booster was launched from Cape Canaveral at 1653 UTC (local time + 4) to a classified National Reconnaissance Office payload into orbit. The payload was designated "NROL 70 (USA 353)". It was believed to be an ORION / MENTOR-type SIGINT satellite.
[10 APR 24] USA CC / FALCON 9 / STARLINK 6-48 -- A SpaceX Falcon 9 booster was launched from Cape Canaveral at 0514 UTC (local time + 4) to put 23 SpaceX "Starlink v2 Mini" low-Earth-orbit broadband comsats into orbit.
[11 APR 24] RU-C VT / ANGARA A5 ORION / GAGARINETS -- An Angara A5 Orion booster was launched from Vostochny at 0900 UTC (local time - 8) to put the "Gagarinets" satellite into geostationary orbit.
[11 APR 24] USA VB / FALCON 9 / USSF 62 -- A SpaceX Falcon 9 booster was launched from Vandenberg SFB at 1200 UTC (local time + 7) on the "USSF 62" mission. Payloads included:
[13 APR 24] USA CC / FALCON 9 / STARLINK 6-49 -- A SpaceX Falcon 9 booster was launched from Cape Canaveral at 0122 UTC (previous day local time + 4) to put 23 SpaceX "Starlink v2 Mini" low-Earth-orbit broadband comsats into orbit.
[15 APR 24] CN JQ / LONG MARCH 2D / SUPERVIEW NEO-3 1 -- A Long March 2C booster was launched from Jiuquan at 0412 UTC (local time - 8) to put the "SuperView Neo-3 01 (Siwei Gaojing-3 1)" satellite into orbit for China Siwei.
[17 APR 24] USA CC / FALCON 9 / STARLINK 6-51 -- A SpaceX Falcon 9 booster was launched from Cape Canaveral at 2126 UTC (local time + 4) to put 23 SpaceX "Starlink v2 Mini" low-Earth-orbit broadband comsats into orbit.
[18 APR 24] USA CC / FALCON 9 / STARLINK 6-52 -- A SpaceX Falcon 9 booster was launched from Cape Canaveral at 2240 UTC (local time + 4) to put 23 SpaceX "Starlink v2 Mini" low-Earth-orbit broadband comsats into orbit.
[20 APR 24] CN XC / LONG MARCH 2D / YAOGAN-42 02 -- A Long March 2C booster was launched from Xichang at 2345 UTC (next day local time - 8) to put a secret "Yaogan-42 02" payload into orbit. It was apparently a military surveillance satellite.
[23 APR 24] USA CC / FALCON 9 / STARLINK 6-53 -- A SpaceX Falcon 9 booster was launched from Cape Canaveral at 2217 UTC (local time + 4) to put 23 SpaceX "Starlink v2 Mini" low-Earth-orbit broadband comsats into orbit.
[23 APR 24] NZ / ELECTRON / NEONSAT 1, ACS3 -- A Rocket Labs Electron light booster was launched from New Zealand's Mahia Peninsula at 2232 UTC (next day local time - 13) to put the South Korean "New space Earth Observation Satellite (NeonSat) 1" and NASA's "Advanced Composite Solar Sail System (ACS3)" CubeSat into orbit.
NeonSat 1 was an imaging surveillance microsatellite with a launch weight of about 100 kilograms (220 pounds), apparently for both military and civil use. It was developed by the Satellite Technology Research Center (SaTReC) at the Korea Advanced Institute of Science and Technology (KAIST), Korea's leading science and technology institute. It was a pathfinder for a constellation of ten more NeonSats, all to be launched by 2027.
ACS3 was based on a 12-unit CubeSat. It deployed a square one-piece solar sail 9 meters (30 feet) on a side, using four booms made of composite material.
[24 APR 24] CN JQ / LONG MARCH 2F / SHENZHOU 18 -- A Long March 2F booster was launched from Jiuquan at 1258 UTC (local time - 8) to put the "Shenzhou 18" crewed space capsule into space on a Tiangong China Space Station support mission. The three-taikonaut crew included commander Ye Guangfu, operator Li Cong, and system operator Li Guangsu.
[28 APR 24] USA CC / FALCON 9 / GALILEO FM25 & FM27 -- A SpaceX Falcon 9 booster was launched from Cape Canaveral at 0034 UTC (previous day local time + 4) to put the "Galileo FM25 & FM27" navigation satellites into orbit.
[28 APR 24] USA CC / FALCON 9 / STARLINK 6-54 -- A SpaceX Falcon 9 booster was launched from Cape Canaveral at 2208 UTC (local time + 4) to put 23 SpaceX "Starlink v2 Mini" low-Earth-orbit broadband comsats into orbit.
COMMENT ON ARTICLE* GEN-AI ARRIVES (14): AI image-generation tools can also benefit from automatically generated prompts. A team at Intel labs, led by Vasudev Lal, set out to optimize prompts for the image-generation model Stable Diffusion. Lal saw the exasperating finickiness of writing prompts as a bug, saying: "So we wanted to see if we can automate this kind of prompt engineering."
Lal's team built a tool named "NeuroPrompts" that takes a simple input prompt, such as "boy on a horse," and automatically tweaks it to produce a better picture. To do this, they started with a range of prompts generated by human prompt-engineering experts, and then trained a language model to transform simple prompts into these expert-level prompts. On top of that, they used reinforcement learning to optimize these prompts to create more aesthetically pleasing images -- as rated by yet another machine-learning model, PickScore, a recently developed image-evaluation tool.
The automatically generated prompts did better than the expert-human prompts they used as a starting point, as rated by the PickScore metric. Lal wasn't surprised: "Humans will only do it with trial and error. But now we have this full machinery, the full loop that's completed with this reinforcement learning ... This is why we are able to outperform human prompt engineering."
Of course, the automatic approach still doesn't necessarily give users what they want, so they gave uses the ability to guide the automated process. In their tool, the user could specify the original prompt -- like "boy on a horse" -- and an artist to emulate; or a style; or a format; or other modifiers.
Lal believes that the exasperating finickiness of LLMs is temporary, a sign of the infancy of the technology, saying: "I think it's important that these kinds of optimizations are investigated and then ultimately ... incorporated into the base model itself so that you don't really need a complicated prompt-engineering step." It's just a question of training AI to be smarter about handling the prompts it's handed.
Tim Cramer -- senior vice president of software engineering at Red Hat -- suggests that automating prompt generation makes sense for casual mass-market applications, but doesn't work so well for more specialized applications. Cramer says: "I think there are going to be prompt engineers for quite some time, and data scientists. It's not just asking questions of the LLM and making sure that the answer looks good ... there's a raft of things that prompt engineers really need to be able to do."
Austin Henley says there are many criteria to deal with in using LLMs to produce commercial product:
In other words, getting the right use out of an LLM really is a serious engineering task. Many big companies now offer a new job title: "Large Language Model Operations (LLMOps)" -- for which prompt engineering is only a part, along with establishing conformance and testing. Given that GAI is evolving rapidly, any such job is going to evolve just as rapidly. Cramer says: "The landscape is just too crazy right now. Everything's changing so much. We're not going to figure it all out in a few months."
[ED: GAI is sort of where personal computing was in the early 1980s: a lot of interesting things going on, a lot of ferment, with rapidly evolving technology that often appears appallingly crude in hindsight. GAI is going to be working vastly better in 2030 than it does now.
In working with DeviantArt's DreamUp image generator, it can be frustrating to specify prompts that give me what I want. One thing I found out is that I have to specify major features of an image before I specify minor details, working down a hierarchy. It appears that if I specify a minor detail first, DreamUp will assume the major features and ignore my specification for them. Again, sometimes the simplest and shortest prompts will get the best results, but it's kind of pot luck.
DreamUp users can list their prompts along with their imagery, but there's a split mindset: some people, particularly those trying to make money off their "deviations", hide their prompts -- while those with a more tinkering mindset like to share them, and find the protective mindset exasperating.] [TO BE CONTINUED]
COMMENT ON ARTICLE* SYNTHETIC DNA CRACKDOWN: As discussed in an article from WIRED.com ("The US Is Cracking Down on Synthetic DNA" by Emily Mullin, 6 May 2024), work on made-to-spec "synthetic DNA" has evolved into an emerging industry. Synthetic DNA has a wide range of application, being used to develop diagnostic tests, make enzymes to digest plastic, or engineer antibodies to treat disease.
A research group needs to investigate a rare species of bacteria? Instead of finding samples, the researchers can just get the sequence from a DNA synthesis company. Synthesizing DNA is nothing all that new, but as of late it's easier, cheaper, and faster to do, thanks to new technology that can "print" custom gene sequences. There are dozens of companies now producing synthetic DNA in volume.
Synthetic DNA has great promise, but it also presents hazards. It's becoming possible to create entirely new sequences that don't exist in nature -- and, by accident or intent, they could include some that present a threat to humans or other organisms. Tom Inglesby -- an epidemiologist and director of the Johns Hopkins Center for Health Security -- says:
BEGIN_QUOTE:
The concern has been for some time that as gene synthesis has gotten better and cheaper, and as more companies appear and more technologies streamline the synthesis of nucleic acids, that it is possible to "de novo" create organisms, particularly viruses.
END_QUOTE
It's possible that a Black Hat could make a dangerous virus from scratch by ordering its genetic building blocks and assembling them into a whole pathogen. In 2017, Canadian researchers announced they had reconstructed the extinct horsepox virus for $100,000 USD using mail-order DNA. What if somebody did the same for smallpox, which is extinct in the wild?
On 29 April 2024, the White House issued new rules to regulate at companies that manufacture synthetic DNA. The rules are consequent to an executive order signed by President Joe Biden in the fall of 2023 to establish new standards for AI safety and security, including AI applied to biotechnology. The rules ask DNA manufacturers to screen purchase orders to flag "sequences of concern" -- that might, for example, increase the toxicity of an organism, or enhance its ability to cause disease -- and assess customer legitimacy. At present, the rules only apply to researchers that receive Federal funding: They must order synthetic nucleic acids from providers that implement these practices.
Inglesby says it's still a "big step forward", since about three-quarters of the US customer base for synthetic DNA are federally funded entities. Many synthetic DNA companies already follow screening guidelines issued by the Department of Health & Human Services in 2010. About 80% of the industry has joined the International Gene Synthesis Consortium, which pledges to vet orders. However, these measures are both voluntary, and not all companies comply.
Kevin Flyangolts -- founder and CEO of New York-based Aclid, a company that offers screening software to DNA providers -- says he's happy to see the new rules: "While the industry has done a pretty good job of putting some protocols in place, it's by and large not consistent," he says. He still hopes Congress will pass a law requiring all DNA providers to screen orders. In 2023, a bipartisan group of legislators introduced the "Securing Gene Synthesis Act" to mandate screening more broadly, but the bill has yet to advance.
Emily Leproust -- CEO of Twist Bioscience, a San Francisco DNA-synthesis company -- also wants regulation:
BEGIN_QUOTE:
We recognize that DNA is dual-use technology. It's like dynamite, you can build tunnels, but you can also kill people. Collectively, we have a responsibility to promote the ethical use of DNA.
END_QUOTE
Twist has been screening sequences and customers since 2016, when it first started selling synthetic DNA. A few years ago, the company set up a "red team" of outside consultants to test their screening, with the consultants setting up fake customer names and ordered sequences of concern. The screening caught many of the probes, but there was some confusion on what sequences were "concerning" -- leading to a review and tightened procedures.
Leproust still worries about certain hypothetical scenarios that are beyond her control. For instance, a state actor with bad intentions could start making its own gene sequences: "Probably the biggest risk is if a state wants to build their own DNA synthesis capabilities. They may be able to do it, because states have vast resources."
COMMENT ON ARTICLEDAYLOG MON 06 MAY 24: I'm a geezer who plays videogames -- generally arcade games, puzzle games, racing games, since I don't have too much spare time to play games. I'd been looking around for a free version of PAC-MAN to play, but I couldn't find one that worked for me.
Today, I saw an article on an elaborate version of PAC-MAN that had been opened up for public play. It looked fun, but it was a multiplayer game, and I wanted a single-player game. I searched for alternatives, and found a YouTube video on free versions of PAC-MAN. One, OPEN GREEDY, caught my eye. GREEDY was made by French enthusiasts, starting out on DOS it seems, to work up to an open-source version for Windows about ten years ago. I downloaded it, and it turned out to be a nice little piece of work.
One fun feature is that the ghosts let out a "blood-curdling scream" when I eat them. Incidentally, Kroger supermarkets has a promotional HTML5 derivative of PAC-MAN named HUNGRY CART, with a munching shopping cart and ads to gobble up. It's cute, but not very sophisticated.
DAYLOG TUE 07 MAY 24: Today, President Joe Biden spoke on the occasion of Holocaust Memorial Day, condemning Palestinian Hamas terrorists for their savage 7 October 2023 attack on Israel, and reaffirming his support of that country. JB also denounced the "ferocious surge" in antisemitism on college campuses and elsewhere in the wake of the Gaza crisis. The speech did not endear him to pro-Palestinian activists -- but a choice between the Palestinians and the Israelis was not really a choice at all.
JB cannot derail the Israeli assault on Gaza, but he is doing all he can to reduce its impact on Palestinian civilians. For myself, I'm just waiting for the crisis to blow over. It is a distraction from other things, particularly the Ukraine War. The struggle between Palestinians and Israelis is endless. The Gaza crisis will pass and the fighting will die down, to drop off the headlines. At that point, the long-term solution will become the issue ... but nobody has any good idea of what the long-term solution could be.
DAYLOG WED 08 MAY 24: There was a wave of elections in the UK recently, with Labour winning handily over the Tories. To rub in the humiliation, a Tory MP named Natalie Elphicke then crossed the aisle and joined Labour.
Britain's Tories are imploding; my British penpal sees them having maybe 20 MPs after the next general election. Roughly the same can be said here in the USA. The "crisis of conservatism" is transcending borders. How did it happen?
On consideration, America's Republicans made four great mistakes:
1: They drank the anti-government kool-aid, and decided they didn't want to govern any more.
2: They defined themselves by what they were against, not what they were for, working to obstruct instead of getting anything done.
3: They decided to make up their own facts, and change them as convenient.
4: They decided to make up their own rules, and change them as convenient.
What happens after the fall? One scenario is that GOP splits into an Extreme-Right party, which will slowly die out, and a Center-Right party -- but the Center has shifted Left, and so the Center-Right has ceased to exist.
After all, DEI, LGBT and reproductive rights, climate-change action, sensible gun-safety laws, and in general responsible government are now Center positions. The No Labels group wanted a Centrist candidate, never understanding: *We've already got one. His name is Joe Biden.*
DAYLOG THU 09 MAY 24: ESSANEWS.com reported on the activities of Ukrainian Atesh partisans in Crimea. Precisely what they are doing is not entirely clear, but they are claiming many casualties among Russian troops and destruction of Russian equipment.
They also report the Russians are transporting many wounded and dead back from the front lines. Ukrainian guerrillas appear to mostly focus on providing intelligence via smartphone on Russian operations, with the AFU reacting with missile or drone strikes.
The guerrillas also perform sabotage, and like to assassinate Ukrainian collaborating officials. Overall, the war is difficult for the Ukrainians right now, with the Russians putting AFU forces under severe pressure in the direction of Kharkiv.
Now that munitions are starting to flow to Ukraine again, odds are increasingly better that the Russian push will be stopped. Once it is, it's hard to think the Russians will be able to do better in the future. We should have a clearer idea of where things are going by the end of the month. Ukrainian and Allied war production is slowly ramping up towards a critical mass. New and possibly decisive weapons are being introduced. For the time being, however, it's still touch-&-go.
DAYLOG FRI 10 MAY 24: An interview was recently conducted with Jack Dorsey, who had been a prime mover in getting Twitter going, to ultimately sell it to Elon Musk. Dorsey has since been associated with a supposed Twitter follow-on named "BlueSky", but now he's distancing himself. Dorsey is no longer on the board of directors of BlueSky, saying that it has gone wrong in the same way that Twitter went wrong. Dorsey, however, is not reflecting on the way Twitter turned into a feedlot for trolls, but on its centralized control.
It seems Dorsey is after a decentralized, distributed social media system open to completely free expression. This sounds like a triumph of idealism over reality, since the #1 problem with social media is actually its weaponization by trolls.
The big issue in social media, whether Dorsey realizes it or not, is that Twitter -- now known as "Xitter", since Musk renamed it "X" -- is dying, and that something will replace it. Inspections of BlueSky and Instagram's Threads show they leave something to be desired. Mastodon seems more viable, but it's more along the lines of the old USENET than Twitter. That leaves Spoutible as an active candidate, since it is much better engineered than BlueSky and Threads, and far better moderated.
The latest feature in Spout, "Accuracy Alerts", allow users to invoke AI-generated fact checks on postings. The trick here is that regulators, particularly in the EU, are big on content moderation, and Spout is doing everything they want. Eventually, governments will have to dump Xitter, and Spout is likely to be the beneficiary. Not incidentally, there was once upon a time an active "Black Twitter", but Musk killed it off. It seems to have obtained a new home on Spout.
AND SO ON: An online commenter named Joe Khalil made some interesting observations:
BEGIN_COMMENT:
Crazy to think: "In Court, Porn Star Details Sex With President" might've been the 4th weirdest headline this week behind:
-- VP Hopeful Continues Media Tour Despite Questions About Shooting Puppy
-- Democrats Save Republican Speaker from GOP Members' Efforts to Remove Him
-- Presidential Candidate Says Worms Ate Part of His Brain
END_COMMENT
I replied that if such things had been said in a sci-fi novel written in the 1960s and set in 2024, everyone would have thought it was a comedy.
* An article from PROPUBLICA.com ("IRS Audit of Trump Could Cost Former President More Than $100 Million" by Paul Kiel and Russ Buettner, 11 May 2024), indicates that Donald Trump may be in big trouble with the Internal Revenue Service, concerning his Trump Tower in Chicago:
BEGIN_QUOTE:
The 92-story, glass-sheathed skyscraper along the Chicago River is the tallest and, at least for now, the last major construction project by Trump. Through a combination of cost overruns and the bad luck of opening in the teeth of the Great Recession, it was also a vast money loser.
But when Trump sought to reap tax benefits from his losses, the IRS has argued, he went too far and in effect wrote off the same losses twice.
The first write-off came on Trump's tax return for 2008. With sales lagging far behind projections, he claimed that his investment in the condo-hotel tower met the tax code definition of "worthless," because his debt on the project meant he would never see a profit. That move resulted in Trump reporting losses as high as $651 million for the year ...
END_QUOTE
The IRS let that pass; the agency was badly underfunded and didn't want to go after high rollers who would fight back. However ...
BEGIN_QUOTE:
... in 2010, Trump and his tax advisers sought to extract further benefits from the Chicago project, executing a maneuver that would draw years of inquiry from the IRS. First, he shifted the company that owned the tower into a new partnership. Because he controlled both companies, it was like moving coins from one pocket to another. Then he used the shift as justification to declare $168 million in additional losses over the next decade.
END_QUOTE
That got the attention of the IRS, with Trump now facing a charge of up to $100 million USD. Since the IRS doesn't discuss its investigations with the public and the situation is complicated to begin with, it's hard to trace what happened and what's going on -- but one Walter Schwidetzky, a law professor at the University of Baltimore and an expert on partnership taxation, told the authors of the article: "I think he ripped off the tax system."
Exactly what happens remains to be seen, but I've been hoping for years that Trump would get nailed on tax charges -- he's obviously been ripping off the IRS for a long time, and in fact he's boasted about it. I was hoping he'd go to court, but this works, too.
COMMENT ON ARTICLE* JOE BIDEN (25): Other members of the Senate Judiciary Committee went on to cross swords with Judge Bork. In the meantime, staffers pulled Joe out of the meeting to give him bad news: reporters had tracked down his flotch at the University of Syracuse Law School 22 years earlier, when he had been sloppy in his attributions in a paper. On the face of it, that didn't seem too disastrous, because he'd been cleared of wrongdoing in the incident -- but on consideration, Joe knew the reporters were not going to be nice to him.
He went back to the hearing, but later called a quick meeting with his colleagues on the Judiciary Committee. Joe explained what had happened and offered to step aside as chairman of the nomination hearing. There was a silence for a moment, leaving him with a sinking feeling -- but then the ranking Republican, Strom Thurmond of South Carolina, said emphatically: "Absolutely not. You're my chairman -- you're my chairman."
Ted Kennedy added: "This is ridiculous. You don't need to step aside."
"Well listen, at least let me explain --"
Alan Simpson, a Republican from Wyoming, cut him off: "You don't have to explain anything. We know you." Nobody on the Judiciary Committee wanted Joe to quit. That reassured Joe somewhat, but the report from Syracuse was still hanging over his head. He talked it over with staff, and the next morning called a press conference to explain that he had made a "stupid mistake", concluding: "I was wrong, but did not intentionally move to mislead anybody. And I didn't."
The headlines in the news media in response were of the style: "Biden Admits To Plagiarism". The feeding frenzy simply ramped up, with Joe being glad to get back to the hearings and ignore the noise for a little while. However, he couldn't make the noise go away, and his family was being hounded by reporters. A week after the hearings started, Joe called a meeting of his family in the evening, including his professional aides, to consider dropping out of the presidential race. He was distraught and so were many of the group, Beau saying: "You'll change, Dad. You'll never be the same."
Finally, after going around in this vein, Joe noticed Jean hadn't been saying anything. He asked her: "What do you think, Mom?" She replied: "I think it's time to get out." That effectively decided the discussion, the only thing left being to determine how to manage the exit.
The next morning -- Tuesday, 22 September -- Joe, with Jill on his arm, spoke to reporters outside the hearing room, telling them:
BEGIN_QUOTE:
... it seems to me I have a choice. I have to choose between running for president and doing my job to keep the Supreme Court from moving in a direction that I believe to be truly harmful. There will be other opportunities for me to campaign for president. But there will not be many other opportunities for made in flow and President Reagan's choice.
... it makes me angry and angry at myself for having been put in ... the position of having to make this choice. And I am a no less frustrating for the environment of presidential politics. That makes it so difficult to let the American people measure the whole Joe Biden. And not just misstatements that I have made.
... my colleagues in both political parties need no explanation of my conduct ... they ... vouch for my honor my integrity. Sometimes you win, sometimes you lose. ... now it's time for me ... to assess my mistakes and make sure I don't make them again.
END_QUOTE
After that was over, Joe felt physically leaden, surprised by how painful the experience was. Ted Kennedy and Jill encouraged him to get back into the hearings, Jill telling him: "You have to win this thing!" He went in and regained control of the hearings, with Kennedy then slipping him a note saying: "There is life after a presidential campaign." Joe could try again in the future. Fortunately, he had no idea of how long he would have to persist.
Joe and Jill took the train back to Wilmington that evening. Jill suggested they eat out that evening, but said he didn't feel like it. Jill said firmly: "We're going out to dinner tonight."
Yes, Ma'am. They went to Restorante Attilio, where they often ate; they got there late, and it was crowded. When they went in, Joe could hear the murmur: "There's Senator Biden." "There's Senator Biden." Joe had been dreading being recognized -- but then one fellow started clapping, and it turned into a standing ovation. [TO BE CONTINUED]
COMMENT ON ARTICLE* WINGS & WEAPONS: As discussed in an article from THEDRIVE.com ("Ukraine's Jet Powered UJ-25 Skyline Drone Appears On Battlefield" by Joseph Trevithick, 22 December 2023), the Ukraine War has been dominated by drones, with Ukraine's defense industry generating new designs on a steady basis.
One of the latest, the Ukrjet "UJ-25 Skyline", is something of a departure for Ukraine, being powered by a small turbojet. It recently broke cover when one of them crashed into the roof of a house in the Russian-occupied Ukrainian city of Berdyansk. Details were not provided by Ukrainian sources, but it is clearly a derivative of the Ukrjet "UJ-23 Topaz" -- originally developed as a target drone, adaptable to the "intelligence, surveillance, reconnaissance (ISR)" mission.
Both the UJ-25 and UJ-23 have sleek bodies, low-mounted forward-swept wings, vee tail, and an engine intake in the back. Although Ukrjet has not released details about the UJ-25, the company has provided details of the UJ-23 -- saying it cruises at 500 KPH (370 MPH), has a maximum speed of 800 KPH (500 MPH), an endurance of 90 minutes, and an operational radius of 400 kilometers (250 miles). It is described as "runway independent" -- meaning launch by catapult or rocket booster -- and is recovered by parachute.
The UJ-25's performance is likely similar, giving it a range of at least 800 kilometers (500 miles), enough to penetrate areas of Russia around Ukraine's borders. Nothing is known about its guidance systems or payloads, though it could obviously be used as a strike asset, decoy, or jammer platform. Incidentally, forward-swept wings enhance "stealth", since they deflect radar signals from the front towards the drone's fuselage, instead of back towards a radar. Forward-swept wings tend to be overstressed in aggressive maneuvering, but drones typically don't perform aggressive maneuvers.
* As discussed in an article from THEDRIVE.com ("New Ukrainian Kamikaze Jet Drone Appears In Russia" by Thomas Newdick, 7 February 2024), Ukraine has fielded another jet drone, known only from a "dud" example captured by the Russians. The Ukrainians haven't said anything about it, so it's just being called the "Black Dolphin K-Drone (BDKD)" for its appearance.
It is a sleek black machine about 3 meters (10 feet) long, with a low-mounted ogival delta wing -- bell shape, sharp sweep at the wing root, curving to moderate sweep in mid-wing, then back to a sharp sweep at the tips -- with a tailfin and a jet intake in the belly. Russian sources say it is of wooden construction, with some carbon-fiber parts and fiberglass skin. The smooth contours and non-metallic materials suggest it has some degree of stealth.
The BDKS has a 20-kilogram (44-pound) warhead and is powered by a little P400-PRO turbojet engine from the German JetCat company -- a centrifugal-flow engine, with a single compressor stage like a pump impeller instead of fan, that being a common configuration for small cheap turbojets. The intake on the belly is unusual, engine intakes tending to be unstealthy features, normally placed on top to shield them from radar. The position on the belly suggesting that the BDKS flies at low altitude.
The latest Ukrainian long-range drones do not rely on GPS guidance, instead featuring unjammable schemes like terrain-following. Specifications like speed and range are unknown.
* The Russians have successfully employed their "Lancet" drone in Ukraine. It has cruciform pop-out straight wings and tailfins, a pusher prop, and a range of up to 40 kilometers (25 miles) -- about twice that of a quadcopter kamikaze drone. It appears to be optically guided. The Lancet can be used both for surveillance and as a k-drone.
The Lancet has proven a dangerous nuisance to Ukrainian forces. The Ukrainians decided that they might as well build their own equivalent, with a number of solutions from some of the Ukrainian drone companies. The Ukrainian Lancets look like their Russian counterpart, and have a similar range. Not much has been said of them; it is interesting to wonder if they have options for an electronic-warfare seeker, allowing them to target and attack Russian "emitters -- radar, radio, or jammer systems.
A photo emerged of yet another Ukrainian k-drone -- of conventional configuration, with straight wings and orthodox tail configuration, plus a pusher propeller. What was unusual about this item was that it was fitted with a standard rocket-propelled grenade (RPG) warhead. That's a very good idea, since RPGs are so very common; such a warhead can't penetrate heavily tank frontal armor, but it can do a good job against relatively light top armor.
COMMENT ON ARTICLE* GEN-AI ARRIVES (13): As discussed in an article from SPECTRUM.IEEE.org ("AI Prompt Engineering Is Dead, Long Live AI Prompt Engineering" by Dina Genkina, 6 March 2024), the arrival of generative AI has presented users with a challenge: figuring out how to phrase a "prompt", or query, to a large language model. It's similar to the art of phrasing queries to Google that find the information we want -- but given that LLMs provide much more detailed information, it's much more difficult. The computer what we tell it to do, which can lead to bizarre results in GAI image / video generation, and useless results in other applications.
Companies are now using GAI to build product copilots, automate tedious work, create personal assistants, and a lot more, according to Austin Henley, a former Microsoft staffer who conducted a series of interviews with people developing LLM-powered copilots. Henley says: "Every business is trying to use it for virtually every use case that they can imagine."
Devising a useful prompt is tricky in the first place, and tends to differ from application to application. A cottage industry of sorts has arisen to provide hints and how-tos on writing prompts, being glorified as "prompt engineering". In reality, prompt design is more art than science, and a dark art at that.
Rick Battle and Teja Gollapudi at California-based cloud computing company VMware were startled at how finicky and unpredictable LLM performance was in response to strange prompting techniques. For example, people have found that asking models to explain its reasoning step-by-step -- a technique called "chain-of-thought" -- improved the performance of the models on a range of math and logic questions. Even stranger, Battle found that giving a model positive prompts, such as: "this will be fun" -- or: "you are as smart as chatGPT" -- sometimes improved performance.
Battle and Gollapudi decided to systematically test how different prompt-engineering strategies impacted an LLM's ability to solve grade-school math questions. They tested three different open-source language models with 60 different prompt combinations each, and found the results wildly inconsistent. Chain-of-thought prompting, for example, sometimes helped and other times hurt performance. They say: "The only real trend may be no trend. What's best for any given model, dataset, and prompting strategy is likely to be specific to the particular combination at hand."
There may be, at least conceptually, a simple way out of the trap: ask LLMs to generate their own optimal prompt. New tools have been developed to that end. Given a few examples and a quantitative success metric, these tools zero in on the optimal phrase to feed into the LLM. Battle and his collaborators have tested these prompt-generation systems and found their prompts almost always did better than the best prompt found through trial-and-error, and found them much faster.
Battle found that some of the automatically-generated prompts were so bizarre that humans were unlikely to ever think of them. Battle says: "I literally could not believe some of the stuff that [one of the systems] generated." He noted one prompt that was straight out of STAR TREK:
BEGIN_QUOTE:
Command, we need you to plot a course through this turbulence and locate the source of the anomaly. Use all available data and your expertise to guide us through this challenging situation.
END_QUOTE
It seems that talking to this particular LLM as if it were Captain James T. Kirk improved its skills on math questions. As strange as such results are, Battle thinks that we can't ever figure out better prompts than a machine can. [TO BE CONTINUED]
COMMENT ON ARTICLE* RUSSIAN WAR ECONOMY: As discussed in an article from BBC.com ("Russia's War Economy Cannot Last But Has Bought Time" by Faisal Islam, 10 February 24), at the outset of the Ukraine War, Russia was hit with painful sanctions. In March 2022 the Russian rouble had crashed, the value in London of corporate giants Gazprom and Sberbank fell 97%. Queues began to form at cash machines in Moscow. Oligarchs' yachts, football teams, mansions and even their credit cards were seized. Russia crashed into a major recession.
The toughest sanctions were the seizing of the Russian state's official foreign exchange assets and, in particular, the unprecedented freezing of the central bank's $300 billion USD in currency reserves. Western governments avoided using phrases like "economic warfare", but there was no other good way to describe it.
About two years on, Russian President Vladimir Putin gloated that Russia was the fastest growing economy in Europe, with the International Monetary Fund (IMF) upgrading its forecast growth for 2024 to 2.6% from 1.1%. Putin has mobilized Russia's economy to support the war in Ukraine. Western leaders claim this model is completely unsustainable even over the medium term. The question remains: just how long it can it be sustained?
Russian state spending is at record levels for the post-Soviet era. Military and security spending at up to 40% of the budget is back to late-Soviet era levels -- though such spending is necessarily at the expense of all other spending for Russia's people. Russia has also been able to keep its oil sales nearly at pre-war levels, at 9.5 million barrels per day, making up for Western embargoes by selling to India and China. The country has evaded sanctions by buying and deploying a "shadow fleet" of hundreds of tankers. One world leader commented privately: "2024 will be much more positive for Putin than we thought. He has managed to reorganize his own industry more efficiently than we thought."
Of course, this form of economic growth has greatly increased Moscow's dependence on oil revenues, on China, and on non-productive war spending. As demand for oil and gas peaks, and competitor production from the Arabian Gulf comes on-stream in the coming year, Russia will be exposed. GDP growth is primarily from war production, but the production of tanks and shells that are then blown up in Ukraine is merely squandering the money for no long-term gain. In the meantime, Russia has experienced a brain drain of some of its most talented citizens.
The Western strategy has not been to destroy Russia's economy, but to tie it down -- restricting access to technology, raising costs, limiting revenues, and make the conflict unsustainable long-term. One US official said: "We'd rather Russia uses its money to buy tankers [for hauling oil] than tanks."
Russia is still likely to persist economically through 2024. The Kremlin is trying to buy time, to wait for the re-election of Donald Trump and reduction of Western funding for Ukraine's defense. That's a faint hope, but it's the only one they have. In the meantime, attention of Western governments is turning back to the central role of those hundreds of billions in frozen Russian financial assets.
Ukrainian President Volodymyr Zelenskyy said: "If the world has $300 billion, why not use it?" Zelenskyy believes those frozen funds should be used to fund Ukraine's rebuilding effort, Britain's Chancellor Jeremy Hunt and Foreign Secretary David Cameron back the move, Lord Cameron saying: "We've frozen these assets. The question is are we going to use them?"
He added that "using some of this money now is, if you like, an advance payment on [Russian] reparations" for its illegal invasion of Ukraine, and could be used "to help Ukraine and save Western taxpayers money at the same time".
The G7's central bankers are considering that move, but it comes with risks. If Russian assets are seized in this way, then what message does it send to other nations -- in the Gulf, Central Asia, or Africa -- about the sanctity of their safe-haven reserves in Western central banks? It's not a simple thing.
Russia's war economy is not sustainable over the long term, and it is heavily dependent on oil exports. Ukraine understands this, and has been ramping up production of long-range kamikaze drones that can hit oil infrastructure deep inside Russia. Efforts to defend those assets further strain the Russian war effort, with coverage being a big problem, and repair of damaged oil facilities is difficult due to sanctions. Russia's Achilles heel is only too obvious, and there's only so much Putin can do to protect it.
* Further investigation with MS ChatGPT counters that Russian government revenues for funding the war are not being primarily obtained through oil exports. According to this view, Putin's primary sources of funds include:
Putin loots the Russian economy to support his war. Disregarding the specifics, the bottom line is that the war -- economically preposterous and ultimately disastrous; Russia's GDP is less than that of the UK, France, or Germany, making the war unaffordable. However, Putin can't give up, because he's had it if he does.
COMMENT ON ARTICLEDAYLOG MON 29 APR 24: "Legal Eagle" Teri Kanefield, writing in her blog, detailed the weaknesses of the "hush money" case against Trump, saying that Trump was guilty of bad behavior in the case, but didn't necessarily do anything illegal. She then suggested that he was going to be convicted anyway, partly because criminal trials more often than not result in convictions -- but more because Trump's conduct in the trial has been nasty, juvenile, and ignorant, marked by attacks on the judge.
I would have to add that Trump's behavior in the 2016 election was so disreputable that it stands to reason there had to be some identifiable crime in it -- or at least the jury would think so. Of course, Trump will appeal if convicted, but so what else is news?
* In other news, Joe Biden's son Hunter is apparently getting ready to slam a big defamation lawsuit on Fox News. *This is gonna be good.* It is somewhat surprising that Fox has been spreading vicious lies about people for decades, but only now is being sued for it. It appears that lawyers thought that pressing a defamation suit against Fox would be very difficult and unlikely to pay off. The success of the Dominion Voting Machines lawsuit against Fox appears to have opened the floodgates. I'm waiting for Ray Epps to get his turn.
DAYLOG TUE 30 APR 24: BUSINESS INSIDER had an article on the failure of US weapons in Ukraine, focusing on the "Ground Launched Small Diameter Bomb (GL-SDB)". It was a 112-kilogram (250-pound) glide bomb, normally dropped by aircraft, mounted on an old MLRS booster rocket.
Observers knew that GL-SDB should have been introduced into the Ukraine War early last year, but nothing happened. It wasn't until much later in the year that it became obvious the Orcs had jammed GPS over the battle area, rendering the GPS-guided SDB useless. A US defense official recently publicly admitted that experimental GL-SDB launches in the war zone had failed. The puzzle in this statement was: *Why say this now?* The trick is that SDB production has presumably moved on to the much more sophisticated "SDB II", which doesn't need GPS.
It seems plausible, if not certain, that the US is now supplying GL-SDB2 in quantity to Ukraine. Some months back there was a report of a Russian assembly in the Donbas region that was hit by 6 long-range munitions and devastated. GL-SDBs come in 6-packs. Maybe an operational test? So the reference to the failure of GL-SDB may be misdirection. Incidentally, there's also a laser-guided version of the original SDB, which could be guided to targets by long-range drones with laser target designators. We'll see in the near future.
As a further consideration, the US has developed a finned fuze for standard 155-millimeter shells, the "M1156 Precision Guidance Kit (PGK)", which converts the shell into a GPS-guided munition. GPS, again, is not working in the battle area -- but why not build a version of the M1156 with laser guidance? Or with the ability to home in on "emitters", such as radar or jammer systems? Nice idea, but nobody's saying a word.
DAYLOG WED 01 MAY 24: A UK penpal told me that British media has been calling the US presidential race "neck & neck" -- which means they're listening to US media. I told him that there's very little visible enthusiasm for Trump on the ground here.
It is hard to understand how Trump could conduct a credible presidential campaign from criminal court. I'm still thinking that in 2050, the younger generation will read about Trump and say: "You're lying. This couldn't have happened." *Oh, but it did.*
I'm not understanding exactly why the polls are so off these days. I don't think there's that much bias, though there may be methodological problems. I suspect most of all that the people being polled are simply muddled. There are some people on the Left who are waffling on Joe Biden, and a lot more people on the Right who have written Trump off, but don't want to admit that voting for him was a bad idea. In any case, Trump's not going anywhere but down.
DAYLOG THU 02 MAY 24: Retired US Army General Wesley Clark, once a NATO commander, recently commented in a public forum on the Ukraine War -- saying it was unlikely the struggle would end one way or the other any time soon.
Clark said: "I don't think Ukraine is willing to compromise and give up its territory. I don't think the Ukrainian military is going to collapse, even though things are pretty tough right now until that American assistance gets there. And I don't think Russia is willing to give up, so it's an endurance contest, and it's also a contest of national will, and whether the United States and NATO will stand firm to support a beleaguered democracy."
Clark said we have to be patient and stay the course: "For Ukraine to collapse, we'd be faced with a huge national security crisis in Europe, and it would empower the Chinese to go after Taiwan. So we can't measure it on the clock."
Clark concluded: "It's going to be a hair-raising time in Ukraine and for the West and then ... summer will probably end with a Ukrainian counteroffensive that regains much of the territory lost. 2025 may be decisive."
I think a counteroffensive is likely, but I'm thinking it will be focused on Crimea, using imaginative tactics. 2025 is indeed likely to be decisive, since Putin will then run out of money to fund the war. If he doesn't win now, he never will.
AND SO ON: This last week, THE ECONOMIST magazine conducted an interview with French President Emmanuel Macron, with Macron saying that Vladimir Putin posed a dire threat to Europe, and that NATO nations should consider direct intervention in the war if necessary:
BEGIN_QUOTE:
I'm not ruling anything out, because we are facing someone who is not ruling anything out. We have undoubtedly been too hesitant by defining the limits of our action to someone who no longer has any and who is the aggressor.
END_QUOTE
Macron said:"if Russia decided to go further, we will in any case all have to ask ourselves this question" of sending troops, describing his refusal to rule out such a move as a "strategic wake-up call for my counterparts." Macron described Russia under Putin as "a power of regional destabilization" and "a threat to Europeans' security". He elaborated:
BEGIN_QUOTE:
I have a clear strategic objective: Russia cannot win in Ukraine. If Russia wins in Ukraine, there will be no security in Europe. Who can pretend that Russia will stop there? What security will there be for the other neighboring countries -- Moldova, Romania, Poland, Lithuania, and the others?
END_QUOTE
Macron said that Russia "has broken all the frameworks and has basically returned to a logic of total war." However, Putin is overstretched and cannot sustain his war: "Devoting a third of its budget to defense is not sustainable for a country whose gross domestic product is lower than that of France, Germany, or the United Kingdom."
Most NATO nations said they had no intention of sending troops to Ukraine. The Poles, however, are playing their cards close to the vest; if France intervenes, it's a good bet the Poles will, too. Macron was not specific about what form such an intervention would take, preferring to maintain "strategic ambiguity", but insisted that the threat should not and could not be underestimated -- that Europe faced not only a military, security, and economic threat, but a threat to democratic rule.
Macron went on to warn voters to steer clear of nationalists in upcoming European Parliament elections:
BEGIN_QUOTE:
I say to Europeans: Wake up! Wake up! They are hidden Brexiteers. All European nationalists are hidden Brexiteers. It's all the same lies. Make no mistake. If you entrust the keys to people who think like they do, there is no reason why Europe should become a great power. In a way, it's as if we were saying it's not a problem if we entrust the bank to robbers. When they are around the table, they take Europe hostage.
END_QUOTE
He singled out the French Rightist Rassemblement National (RN) Party, saying the RN ...
BEGIN_QUOTE:
... wanted to pull out of Europe, out of the euro, out of everything. Now it no longer says anything. It's reaping the benefits of Europe, while wanting to destroy it without saying anything. And that's true in every country.
END_QUOTE
That's not so true in the USA, with Trump and MAGA disintegrating in full public view. Of course, that's not so easy to see from the other side of the Pond.
COMMENT ON ARTICLE* JOE BIDEN (24): Joe continued his preparation for the Bork hearings, while making campaign stops. He felt he was building up momentum, though not all the media agreed. Then, only days before the Bork hearings began, the reporter Maureen Dowd of THE NEW YORK TIMES published an article in which she accused Joe of "lifting" Kinnock's speech at the Iowa State Fair. Buried in the article, she did concede that Joe had properly cited Kinnock in earlier speeches and that she had contacted the Biden campaign -- which of course called it inadvertent, an oversight. What she did not admit was that she had been tipped off to the gaffe by a staffer of the Dukakis campaign named John Sasso, who helpfully sent supporting materials to the media. Dukakis himself knew nothing about it, and fired Sasso when he found out.
Joe spent much of the weekend before the hearings on the phone with reporters, trying to clarify the situation. Unfortunately it got worse, particularly after the reporters found out about the Bobby Kennedy quote that had been buried in Joe's speech in Sacramento early in the year. When Joe gaveled in the Bork hearings on Tuesday morning, the Right media was in full steam, one essay being titled "Plagiarizing Joe". The Right was already activated in defense of Judge Bork, and Joe had just handed them a mallet to bonk him over the head with. What particularly discouraged Joe was that his integrity, his "word as a Biden", was being so loudly attacked.
In any case, the hearings began on schedule. They started out civil and would remain so, though Joe noticed with some satisfaction that the swing voters on the Judiciary Committee made it clear to Bork they had difficulties with his legal philosophy. When Joe first spoke at length, he assured Bork that the hearings would be fair, expressed his esteem for Bork's reputation as a jurist -- and then asked Bork to clarify his philosophy. Bork did so:
BEGIN_QUOTE:
As I wrote in an opinion ... the judge's responsibility "is to discern how the Framer's values, defined in the context of the world they knew, apply in the world we know." If a judge abandons intentions as his guide, there is no law available to him, and he begins to legislate a social agenda for the American people. That goes well beyond his powers.
END_QUOTE
Joe found Bork highly confident, wondering if he might stray over the centerline into arrogance. Joe then brought up specific cases, with Bork generally responding that he had no casual opinion on those cases and would need to work on them to come to any judgement. Joe sensed a weakness and brought up the case of GRISWOLD V CONNECTICUT, which overturned laws against birth-control on the basis of privacy rights. Bork replied that judgement did not establish that there was any such "unstructured, undefined right to privacy" in the Constitution. The Constitution had no soul that could be discerned by the judiciary. Joe moved in:
BEGIN_QUOTE:
You're one of the most well-read and scholarly people to come before this Committee in all your short life. Have you come up with any other way to protect a married couple under the Constitution against an action by a government telling them what they can or cannot do about birth control in their bedroom? Is there any constitutional right anywhere in the Constitution?
END_QUOTE
Bork replied: "I have never engaged in that exercise." Bork, having been given enough rope, had hanged himself: he could say the Constitution had no soul, but most legal scholars and almost all layfolk thought it did. Joe wrote later:
BEGIN_QUOTE:
Ninety-nine out of a hundred law professors and constitutional scholars must be fascinated by [Bork's assertion]. And ninety-nine out of a hundred average citizens must have been thinking: Bork can't stop the government from intruding in my bedroom?
END_QUOTE
Joe wanted to make sure Bork's assertion was publicly underlined, asking in the following exchange: "... Does the majority have the right to tell a couple that they cannot use birth control?"
Bork effectively repeated his assertion: "... I have never decided that case. If it ever comes before me, I will have to decide it." Bork was saying a couple had no unarguable right to use birth control. [TO BE CONTINUED]
COMMENT ON ARTICLE* SCIENCE NOTES: As discussed in an article from SCIENCENEWS.org ("Here's How Cool a Star Can Be and Still Achieve Lasting Success" by Ken Croswell, 23 August 2021), the smaller a star is, the more common it is, with our Galaxy heavily populated by little red dwarf stars. There are more, even smaller, brown dwarfs, but they're not generally regarded as true stars, not supporting any more than transient fusion reactions.
It's often hard to distinguish between red and brown dwarfs, because when young they both look the same: red and dim, glowing with the heat of their formation. However, in brown dwarfs fusion reactions in the core quickly sputter out, and they never become true stars. Astrophysicists Dino Hsu and Adam Burgasser at the University of California, San Diego and their colleagues decided to investigate the dividing line between the two.
They determine the breakpoint by examining the orbits of red and brown dwarfs around the center of our Galaxy. When a star is born, it orbits around the Milky Way's center on a fairly circular orbit -- but over time, gravitation tugs giant gas clouds, spiral arms and other stars make their orbits more and more elliptical.
Red dwarfs burn their fusion fuel slowly relative to their size, so they live longer than bigger stars; their predicted lifetimes are far longer than the current age of the universe. However, since brown dwarfs don't, in maturity, perform fusion reactions, those that are still warm are young. That means, on the average, red dwarfs should follow more elliptical orbits around the Galaxy than young brown dwarfs do.
Hsu's team inspected 172 red and brown dwarfs of different spectral signatures, which vary as per their surface temperature. The researchers found that a sharp break in stellar motions separates warmer objects, which on average have more elliptical orbits and are older, from cooler ones, which on average have more circular orbits and are younger. This break appears at a spectral type between L4 and L6, corresponding to a surface temperature of 1,500 to 1,700 degrees Kelvin -- as compared to our Sun's surface temperature of about 5,800K.
Above this critical temperature, the dim suns are a mix of long-lived red dwarfs and young brown dwarfs. Below this temperature, Hsu says: "It's all brown dwarfs." However, Hsu says: "We need a more complete sample." That promises to be tricky, since red and brown dwarfs are common, but hard to see. It is still doable.
* As discussed in an article from NATURE.com ("Could AI-Designed Proteins Be Weaponized?" by Ewen Callaway, 8 March 2024), analysis and design of the elaborate structures of proteins is a tricky job, traditionally demanding a lot of computing power -- and even then, not easily getting results. However, machine-learning (AI) systems, through their "educated guesswork", have been able to get good results, opening the door to a new era in "protein engineering".
That new era poses both promises and hazards, hazards including the use of engineered proteins as bioweapons, for example toxins or highly transmissible viruses. Researchers have accordingly now set up a voluntary initiative to promote the safe and ethical use of protein design. David Baker -- a computational biophysicist at the University of Washington in Seattle, who is part of the initiative -- cautions that people shouldn't over-react: "The potential benefits of protein design [by AI] far exceed the dangers at this point."
To assess the potential for malevolent use of designer proteins, Baker's Institute of Protein Design at the University of Washington hosted an AI safety summit in October 2023. Baker said the summit asked the question: "How, if in any way, should protein design be regulated and what, if any, are the dangers?"
The initiative that resulted calls on the biodesign community to police itself, to regularly review the capabilities of AI tools and monitor research practices. Baker wants to set up an expert committee to review software before it is made widely available, and to recommend 'guardrails' if necessary.
The initiative also calls for enhanced screening of DNA synthesis, with DNA of course being the tool used to create proteins. At present, many companies providing DNA synthesis services are members of an industry group, the "International Gene Synthesis Consortium (IGSC)", that requires them to screen orders to identify harmful molecules such as toxins or pathogens.
James Diggans -- head of biosecurity at Twist Bioscience, a DNA-synthesis company in South San Francisco, California, and chair of the IGSC -- believes that AI should be used to detect threats: "The best way of defending against AI-generated threats is to have AI models that can detect those threats."
Governments are taking an interest in AI and biosecurity risks as well, though Baker isn't enthusiastic, feeling that government intrusion would interfere with progress in the field. Not all agree. Mark Dybul -- a global health policy specialist at Georgetown University in Washington DC who chaired a 2023 report on AI and biosecurity -- is enthusiastic about the effort, but adds that: "We need government action and rules, and not just voluntary guidance".
David Relman -- a microbiologist at Stanford University in California -- similarly believes that researchers can't be trusted to regulate their own research: "Natural scientists alone cannot represent the interests of the larger public."
* As discussed in a NASA press release "Unlocking Earth's Microscopic Mysteries With NASA's PACE" by Kathryn Hansen, 9 February 2024), on 8 February 2024, the NASA "Plankton, Aerosol, Cloud, ocean Ecosystem (PACE)" satellite was launched from Cape Canaveral on a SpaceX Falcon 9 booster.
PACE's primary payload was the hyperspectral "Ocean Color Instrument (OCI)" designed to observe oceans and other water bodies across a spectrum of ultraviolet, visible, and near-infrared light. OCI will map the distribution of photosynthetic marine microorganisms, or "phytoplankton", to monitor the state of the marine environment. PACE also carried two polarimeter instruments for atmospheric particulate analysis, providing researchers with data on atmospheric aerosols and cloud properties, as well as air quality at local, regional, and global scales.
COMMENT ON ARTICLE* GEN-AI ARRIVES (12): The first wave of Vectara's customers was heavily concentrated in customer support and sales -- areas where a 3% error rate is generally seen as acceptable. It is not acceptable in many other fields. Vectara's Ahmad says he's seen growing interest from the legal and biomedical fields. That error rate may not be too difficult for the legal profession, as long as lawyers realize they have to check their work, but it would mean huge problems for medical service providers.
OpenAI is careful to point out that users should double-check information they get from a GAI chatbot, and for critical issues such as legal, financial, or medical advice, chatbots are no substitute for professionals. AI experts say that LLMs need clear constraints and often expensive product work to be reliable enough for most businesses.
A study performed at Stanford University in early 2024 highlighted the difficulties with GAI chatbots, with the researchers asking ChatGPT basic medical questions to test its usefulness for doctors. To ensure better responses, the researchers prompted ChatGPT with phrases like: "You are a helpful assistant with medical expertise. You are assisting doctors with their questions." Strange as it may seem, experience shows that giving an LLM a little pep talk yields better responses.
Disturbingly, the study found that GPT-3.5 and GPT-4 both tended to give very different answers when asked the same question more than once. Hallucinations were very common, ChatGPT only yielding medically correct answers to queries about 20% of the time. However, the study did conclude that responses from ChatGPT "to real-world questions were largely devoid of overt harm or risk to patients."
Google is among the big AI providers who are now offering products to make LLMs more reliable. Warren Barkley -- senior director of product management for Vertex AI at Google Cloud -- says: "While AI hallucinations can happen, we include features in our products to help our customers mitigate them," Barkley says that Google gives companies the ability to tie -- or "ground" in AI terminology -- LLMs to public data sets, Google search results, or proprietary data.
Not everyone working on GAI systems is all that worried about hallucinations, most notably OpenAI CEO Sam Altman, seeing them as more of a feature, not a bug. Isn't generative AI supposed to be imaginative? Andrej Karpathy -- the former head of AI at Tesla and now at OpenAI -- commented that he struggles a bit with the "hallucination problem" in LMMs, because, in a sense "hallucination is all LLMs do -- they are dream machines."
Amin Ahmad of Vectara believes that hallucinations will be "solved" in a year or two, which Altman sees as plausible as well -- though Ahmad is careful to qualify that statement: "When I say solved, I mean it's going to be that these models will be hallucinating less than a person would be. I don't mean zero."
Even if the current generation of LLMs don't vastly improve, Ahmad believes their impact will still be monumental, in part because we'll just figure out what they're good for:
BEGIN_QUOTE:
The fact is that the kind of transformer-based neural networks we have today are going to completely change the way business is done globally. I started the company because I believe this technology has very broad applications, and that almost any organization, large or small, could use it and take advantage of it.
END_QUOTE
[ED: For myself, I find GAI chatbots extremely useful, but in part because I never expected them to be magic answer machines. Instead of me scouring the net for information and putting it together, the chatbot scours the net and puts it together itself. If I need clarification or want verification, I can adjust my query accordingly. The chatbot lists the sources it's using, and I can trace back to them if I need to. On simple straightforward questions, the chatbot generally does very well. If it can't give any good answer or a lead, that may suggest I can't find the answer on the internet. Hey, it happens.
Now, I've got myself a subscription to DeviantArt DreamUp, and do have problems with hallucinations there -- mangled hands, three legs, floating fingers, strange eyes, and so on. I think that will get better with more training of the engine. Oddly, the hallucinations seem to get worse if try to make too detailed a specification; it seems to confuse the engine. Give a short simple prompt, the engine will follow its own instincts and do a good job, though the result may not be what I want.]
COMMENT ON ARTICLE