Continues Learning is indeed a big deal.
Continues learning does not mean you need a continuously learning LLM—this can be a property of a system.
But:
The bitter lesson says a data‑scaled meta‑learner that learns to learn will beat our hand‑engineered scaffolds.
Speculation on the available info: They must have questioned him on that. Discovering that he was not entirely candid with them would be a good explanation of this announcement. And shadowbanning would be the most discoverable here.
Thanks for the clarification. I was more thinking about the attitude towards China.
A honest question from an outsider (to American politics): isn’t it now convenient for the current political establishment to go with the the lab leak hypothesis and so the change should not be taken into account when updating priors?
“If alien aircraft were on Earth, they would need to be carefully calibrated to give us grainy distant glimpses (in every possible way) but never more. If alien aircraft are here, they’re screwing with us.”
I don’t know—this inference seems rather weak. I try to be on time to appointments, most of my failures are pretty small—observing me being late a minute or two a couple of times does not mean that I calibrate being late to a minute or two. Failures would naturally cluster near the border line.
Plus they might be actively erasing evidence when it is too obvious.
My heuristic about this is that the public is currently oversensitive on zoonotic viruses—so I feel free to not following this one until there is some more serious info.
I am surprised that nobody linked to Scott Alexander on ivermectin: https://astralcodexten.substack.com/p/ivermectin-much-more-than-you-wanted (plus: https://astralcodexten.substack.com/p/higlights-from-the-comments-on-ivermectin?s=r). For me it more or less settles the subject—the most probable hypothesis is ivermectin de-warms people and it has a huge positive outcome on peoples health when people have warms (independently from covid—or maybe even dependently—because the warms decrease immune response fighting the virus) and unfortunately in many places on Earth worms are still prevalent enough that it shows in covid related trials.
But on the other hand I am pretty much sure that much of vaccines related info is now suppressed—not through a centralized censorship but via an emergent mechanism.
The clustering is pretty telling—anyone who claims that vaccines are dangerous also claims that ivermectin or HCQ treats covid, everyone who rejects ivermectin will also reject any claims about adverse effects of vaccines.
Yet another clustered subject is vaccine efficacy—it is pretty clear that injected vaccines don’t prevent getting ill and spreading the virus. Vaccination can limit the time of virus shedding—but actually the dangerous case is when we shed the virus when hot having other symptoms and for all we know vaccines might even increase the durability of that infection phase. This is something that is admitted by the ‘main stream’ - but the implication of that is that “vaccine passports” are silly, because vaccinated persons can still spread the virus and that will never be accepted by the ‘pro-vaccines’ side (which is now main stream).
It is also pretty clear now that the current vaccines are less effective against the new variants—the problem is that the ‘pro vaccine side’ does not want to admit that, and ‘the other side’ does not want to admit that the right conclusion is that we need new vaccines (maybe the nasal onese—that would provide sterilizing immunity): https://erictopol.substack.com/p/the-covid-capitulation?s=w
I must admit I am not an expert in this—but I would assume that the low hanging fruit is patrolling bots with AI for spotting the enemy. The advantage of using AI is two things—one is relieving people from paying attention to the video feeds, another one is that it would compress the communication needs—the bot would only need to communicate after it spots something interesting and do everything else autonomously.
I don’t know if the loitering munitions have such capabilities—wikipedia only says: “Switchblade has sensors to help spot enemy fighters ”—it might be classified.
There is over a million people moved from Ukraine to Russia now (https://wyborcza.pl/7,75399,28394968,ponad-milion-ukraincow-wywiezionych-do-rosji-czesc-z-nich-miala.html) and there is very little information about their whereabouts. Some of them might be voluntary refugees—but a huge part was deported against their will, sometimes into faraway parts of Russia. I’ve read they get papers stating that they must stay in those faraway regions for two years. I am curious what is the legal status of that. I suspect it is all very shaky even in Russian law standards and they can be helped with some pressure from the international community and maybe with just getting the information about their rights to them.
https://en.wikipedia.org/wiki/2022_Ukrainian_refugee_crisis#Neighbouring_countries “According to the Ukrainian authorities, Russian troops in the territories of Ukraine occupied by Russia are engaged in the forcible deportation of people from Ukraine to Russia, passing them off as refugees. According to the Russian government figures, 656,381 refugees had gone to Russia by 28 April.[1] In March, the Ukrainian foreign ministry alleged that 2,389 Ukrainian children had been abducted from Russian-occupied territory in Donetsk and Luhansk, and transferred to Russia, as well as “several thousand” residents of Mariupol”
One of the big surprised in this war for me is no news about usage of small drones with AI. Maybe there are some but still classified—but I would imagine that sophisticated opponents would now use massive amounts of them. They could be quite cheap and detecting people and weapons seems like pretty standard neural networks application. They could give the fighters full awareness of enemy moves with minimal attention requirements. I read about commercial drones being used, sometimes ones that require full immersion from the operator via goggles etc—drones with simple AI would be a game changer in this task.
To understand Bucha (and other places) one first needs to look at: https://www.youtube.com/watch?v=Zrf0yLxViRo, https://www.youtube.com/watch?v=5KFcfVWbyzU, https://www.youtube.com/watch?v=ev0x9pqYqvs and similar videos (https://www.google.com/search?q=ukrainian+civilians+confront+russian+soldiers+youtube). What you see there is people who are not afraid. But fear is an indispensable ingredient of authoritarian regimes. Russians know that they cannot govern people like those in these videos and decided to do something about that.
This is why it wasn’t concealed—they wanted the Ukraininans to know and start to be afraid.
I’ve read on twitter how the wheat situation might be improved by India surpluses. But when I googled for other sources I found: https://www.reuters.com/world/india/exclusive-india-acts-seize-gap-wheat-export-market-left-by-ukraine-war-2022-03-15/ That gives only about 5 million tonnes more than last year—less than one fourth of what Ukraine exports.
What do you mean when you write that it is about ‘limbic system’? For me it suggests that you imagine a big enthusiastic nationalistic crowd—but read https://twitter.com/kamilkazani/status/1504910499418234882 - most of the people are forced to be ‘supporters’. Many of them will learn to be more ‘enthusiastic’ with time—because the Russian society has been trained for that for a long time. For sure there are also true believers—but the whole game is about making everyone look like a believer—it is all about https://www.lesswrong.com/tag/common-knowledge
I have been living in communistic Poland and I attended the 1st of May parades—I remember the coercion, but I also remember how absurd it all felt, I did not understand why it happens when everyone seem to be against it. Maybe I just lived at the very decline of that system in Poland. Only after reading Solzhenitsyn I started to build a theory of how it works. Some people can really just force themselves to believe, this is easier when you don’t have a strong drive to gather information or are incapable processing the contradictions, but on the other hand the more intelligent are also better at resolving the contradictions to the ‘right’ conclusions—delusions are a hell of a drug (in Gulag Archipelago Solzhenitsyn writes about the true communists in the prison camps and the ways they justify their own suffering). For others it is about developing a split personality, something like a https://www.lesswrong.com/posts/imxPxMjXE2JT9rLaF/tulpa-references-discussion that you can switch on when needed. You also would expect the psychopaths to be very successful in that society.
As others pointed out it is very difficult to have a formal theory in these circumstances—but I still believe that we can have heuristics about evaluating sources. Something like https://astralcodexten.substack.com/p/bounded-distrust?s=r
An important case for me personally would be evaluating Bellingcat—I can understand how cynical people believe they are just CIA psyops, or that they could easily be misled by media purposefully planted in the places they look for information—etc. It is a new kind of media/investigation institution and it has many surprising strengths, but most probably also many unexpected weaknesses.
I am looking for a python coder. The project is in Object Detection (finding Lego Blocks) - but I am looking rather for someone junior. I still do it just by myself—no company established—so I cannot help with visa sponsorship or something—but it can be fully remote and I could help with the other aspects of relocation to Poland (if for some reason Warsaw was attractive for any Russians fleeing the empire).
It just a flashmob. A flashmob in Macy’s was so precovid—now we have flashmobs on capital markets: https://zby.medium.com/gamestop-capitol-hill-riots-flashmobs-everywhere-3c1df6333328
Also GME should just raise capital and sell enough shares to get the price more reasonable. SEC should not just allow that—but encourage them.
See also Ben Thompson on social networking 2.0: https://stratechery.com/2020/social-networking-2-0/
I would presume that many people here had bought stuff at Silk Way—it is puzzling that the discourse about antigen tests is still about the official approval and not about how you can import reliable tests from abroad. For my personal use I found online pharmacy in Germany that sends to Poland and does not check if I am really a doctor—but they probably don’t send to US, so you need to find your own ways.
Regarding Belgium—have a look at Czech Republic too.
https://x.com/revishvilig/status/2001001161508446451
“from Deputy commander of Ukraine’s HATRED reconnaissance battalion:
The war evolves roughly every six months. At the moment, it is entering a new era — the widespread introduction of artificial intelligence.”