https://sifted.eu/articles/deepmind-talent
https://www.businessinsider.com/ai-researcher-quit-google-op...
https://www.bloomberg.com/opinion/features/2023-07-13/ex-goo...
https://mdwdotla.medium.com/why-im-leaving-google-for-a-star...
4,236 karma · joined September 25, 2016
https://sifted.eu/articles/deepmind-talent
https://www.businessinsider.com/ai-researcher-quit-google-op...
https://www.bloomberg.com/opinion/features/2023-07-13/ex-goo...
https://mdwdotla.medium.com/why-im-leaving-google-for-a-star...
Additionally, it's complete and total oversimplification. If you look at Google's earnings it's pretty damn clear that at least until 2020 they were not just going for maximum total spend, but for a steady, gradual raise in total spend. Not too slow, not too fast. They were NOT taking every opportunity they had, in fact they're famous for systematically refusing many opportunities (see the original founders' letter, but even after that). They were farming the ad market, the ad spend, growing it, nurturing it. Then COVID blew up the farm.
Maybe you're right now, but I do hope they're recovering their old tactics. Because if they maximize it you'd see nothing but scams ... wait a second.
B) Google was built by providing a vision, and getting out of the way of ground-up engineer efforts. "Scaring workers into compliance" IS killing the golden goose.
You can see this in AI. Every story from an AI engineer that ran away from Google is the same. They didn't run away for the money, they ran away because they were getting scared into compliance.
Now AI may make it, or not. I don't know. But this is happening EVERYWHERE in Google. Every effort. Every good idea, and every bad idea runs away, usually inside the mind of "a worker". Not to make them personally maximum money, but it's natural selection: if the idea doesn't run away, the engineer it's in is "scared into compliance", into killing the idea.
Whatever the next big thing turns out to be, it simply cannot come out of Google. And it will hit suddenly, just like it did for Yahoo.
I mean, sorry, but the black flags and open hostility towards Jews shows what at least a large percentage of the movement is really about. Especially in Europe.
The rest of the movement is the same as leftist movements, imho. It's also not about supporting Palestinians it's about fighting the power, ie. attempting to have political impact by "campaigning" for something SO immoral, unacceptable and unrealistic that there is bound to be a fight. It's about the fight, NOT about a solution. And by campaigning I don't mean campaigning in the sense of political campaigning, or even the vitriol spouting semi-threatening Trump is doing, but being so in the way, sabotaging people's lives, that normal people pretty much have to react with violence (because that's the point of blocking, for example, the Golden Gate bridge: to threaten people's livelihoods, and get a strong reaction that way).
It's not about saving tax dollars ... Yes, there's 1% fringe rightists in there. But seriously? It's not about that.
This seems to apply to many conflicts, e.g. Sudan. The US could intervene. Now you might say "US tried something similar and that was not exactly a good experience", but the US certainly can intervene. Or, in Lebanon the US, and all countries in the UN security council promised to intervene (and disarm Hezbollah), but just don't do it.
The question is which is the least bad mistake? Everything from attacking Putin directly, to doing nothing, to weapons for Ukraine, to NATO soldiers fighting in Ukraine.
It's management responding to the customers. Yes they "have the power" (the higher up in management I go, the less power I seem to have. I can affect an invididual a LOT, but if you think going up in management will get you the power to, for example, make a new product ... best of luck with that)
In other words: we get what we collectively pay for.
Same with agression, animals, unacceptable collections (either hoarding, or something like a collection of knives), ...?
Are we going to "outlaw" homeless people with a dog?
2) Almost, Chinchilla concerns itself with minimizing cost(training) + cost(inference). The expectation most people have is that there's an insane amount of inference compute, and training is maybe 1% of that. The point of the Chinchilla paper is that that's not true, training uses such insane amounts of compute that despite all model inference by half the internet for a year or two is a huge amount of compute, training is still a very decent percentage of that. I believe in one of the examples they pointed out that even the whole internet inferring with a model for years was still only 20% of the cost of training that model.
People expect it works like compilers, that making a compiler produce 1% faster code is worth 50 highly-paid SWEs because while an individual program run isn't exactly expensive, the time and resources spent running programs is astronomically larger than the time and resources spent developing compilers.
The thing is most of the optimizations we know don't work during training. You can't quantize, you can't MoE (well, you can, obviously, but it doesn't save any training computation. In fact it increases training cost)
At 50-50, having a 10% cheaper-to-train model justifies 10% more expensive inference.
3) Combining both arguments ... at this point people should probably realize that Facebook's LLama is really an attack on Google (which is at least partially working, elon musk is tweeting about it)
If Facebook really doesn't care about AI (or ... about as much as, say, netflix does. Not zero, but as long as they beat reddit's efforts they feel very comfortable), but Zuck does care about destroying Google, the calculus changes. Zuckerberg may not want the best possible AI, he may want as many scammers as possible trying to Game the Google search quality team, to present them with challenges faster than they can adapt. Then the training cost becomes a moot point.
Hmmm, I should send my CV to meta ...
Likewise, these professionals hide that almost all experiences kids have with social services are negative for the kids. Now I suppose you could say the above is an example of that, but really, it goes further. Kids seek help with homework, and only get berated by someone that couldn't do the homework themselves ...
Studies keep pointing out that social services is exactly the wrong approach. What makes teachers, and social professionals good is excellent subject knowledge, combined with basic psychology. NOT the other way around. And in practice every mental help professional I've ever seen thinks they know what to do, and when pushed fail to produce even basic psychological facts, or outright deny them. I like to think you can explain this that when push comes to shove our minds are trying to solve problems in the real world.
The majority of mental problems are someone failing to solve real world problems, and repeatedly failing to influence the outcome. A little bit of psychology is needed to get them to try again ... and a LOT of knowledge of the real world is need to make sure the outcome is different.
Obviously generalizing over pictures of humans and pictures of porn will enable CSAM generation ...
Plus a LOT of things have been declared illegal pornographic material in various places: all porn (China), drawn CSAM, any kind of drawn porn with unclear ages, ...
Because totalitarian ideologies ... they can't fix reality, but they can certainly fix the appearance of reality and make the people judging THEM only observe a fixed reality.
Plus this is something that "infects" everything. If there is ONE western thing the CCP wants to hide from Chinese, they'll have to cut access to a LOT of western products just to hide a few small things they really don't want to get out (better ballpoint pens in the west perhaps?). It's funny how this starts too. Like in the USSR, in China when it started it was more-or-less positive. They outlawed better products and celebrated chinese industry catching up. Needless to say, they're a government. The focus has shifted to just outlawing better products, and now they're really more often destroying people who help chinese industry catch up, because ... that does the absolute #1 forbidden thing: it gives the wrong politicians a leg up over others. It gives, to be exact, random politicians a leg up over others ... which of course quickly gets a wrong one.
Hopefully it'll only mean Chinese people never see western products, unless specific products the party wants them to use. Hopefully we'll retain access to Chinese goods.
Though of course, it'll rapidly become INCREDIBLY rewarding for anyone with a great idea to run to the west. In the west, you'll get a great job at the very least. Idea great enough? You get to be rich. You stay in China with a great idea? You get killed ...
People call it a religion now ... because there isn't a state anymore (called a caliphate). Which illustrates yet again how central the state is to islam. All states, including the Taliban and the Mullahs of Iran, radically de-emphasize the fact that the islamic religion demands a global central totalitarian state, and emphasizes nationalism and, for example, the difference between Iranians and Pakistani. Every 6 year old Iranian knows how "their islam", first, centers around the Iranian state, and second how pakistani islam is totally wrong. Plus, of course, how trying to work out those differences, as the "one islam" ideal in the religion itself would seem to reward, is treason.
You could for example DPO based on another LLM's evalution of how a conversation is going, or you could use whether the conversation keeps going.
Well, you'll quickly find that bodies require constant control to survive. Whilst an animal's body can survive for months with an animal's mind operating at 1% capacity, if the output of the mind truly stops, the body dies. The heart, respiration, the circulatory system (deciding where the blood goes), ... all need to be controlled with some amount of intelligence for them to function at all. Even just to remain alive, they need intelligence. They don't work on reflexes alone.
For mammals, total loss of control leads to fatal injury ... in seconds. Without control, the heart will either stop or (wrong order of muscle firing) be totally ineffective at building pressure in the circulatory system. This will kill oxygen transfer through the blood-brain barrier as soon as the pressure drop reaches the brain, which will be much less than a second, this leads to further loss of control in 2-3 seconds. Irreperable damage sets in before 20 seconds pass, death (permanent and total loss of control) occurs in humans after about 1 minute 30 seconds. The plus side of this seems to be that, when that control is there, mammal bodies are far more efficient than a "passive" robot can be (any robot you can turn off -safely- at any time is a "passive" robot, the big counterexample is a flying drone, which will incur damage and may injure others if turned off or it loses control)
So "predict the next token" is really THE most important function of the brain. Second, it is incredibly important to KEEP predicting the next token, to keep going, and keep going, and keep going, no matter how stupid the output, because literally any output will give a result superior to no output at all. No matter the damage, no matter how terribly wrong things are going, it must keep going further.
It, for example, ruled that Belgium can't keep imprisoning Freddy Horion, a murderer (who murdered when he was 24, he's now 75). He is still in prison, 10 years after the conviction of the Belgian state, who seem to be keeping him in prison mostly because he refuses to confess (and testify against a suspected accomplice), apparently for over 21 years now that is the reason he's in prison (and he hasn't received one cent of the money the Belgian state was convicted to pay him, since it was a penalty per day doubtless it is millions of euros now). Plus he's 75 ... Additionally he's been denied medical care at least twice.
There's another Belgian prisoner, Farid Bamouhammad, who has been tortured in prison (hung up with cuffs by his throat to the heating system and left like that for more than 24 hours). The story has a horrific ending: he was denied medical care for 3 years. He ... has since died from the cancer (treatable at the beginning of those 3 years he was denied medical care by the justice system) he turned out to have. He was not paid damages either. There's more horror in this case.
AND there's a general conviction of the Belgian justice system, on 2 separate aspects. Nothing has changed. Damages have not been paid. For the general convictions the damages, a small per-day per-prisoner amount, ignored by the state for a decade, are now absolutely ridiculous amounts that will never be paid.
To give another example, ECHR ruled that the Danish MUST NOT allow child services to act in situations where parents cannot provide appropriate housing for children. Instead, it must use other institutions to provide housing to those families. Inappropriate housing is the number one reason children are placed out of their home, before ... and after the ruling. Obviously, once again, a few specific parents and children were to receive damages, they haven't received a cent.
(this is especially ironic, since it'scourts going EXPLICITLY against a superior court, the ECHR, just because the executive requests them too. This is against the separation of powers and violates the very concept of a judicial system ... and the violators ... are judges)
A Dutch general was convicted of war crimes during the Yugoslav war, and a bunch of people were promised payment for his crimes. He has not been punished, not even fined, and has since received the medal of Honor from the Dutch Royals. The money the Dutch state was convicted to pay has not been paid. Not a cent.
Germany was convicted to be in the wrong to arrest 2 people who had crossed into Germany at the request of the Dutch state (they were -alleged- shoplifters, and they are innocent, in the sense that they were never convicted), and hand them over to the Dutch police. This failed to respect a long list of legal rights EU citizens supposedly have.
Switzerland has been convicted of wrongfully convicting ...
France has been convicted of holding prisoners in inhumane conditions (and we don't mean "a few", but 60% of all judicial prisoners) ...
Now don't take this the wrong way. When ECHR convicts INDIVIDUALS (or companies) there is more enforcement. But against states ... states just ignore, nothing happens. Yes, there are cases of states actually paying, but looking up which ones they are (e.g. Turkey. One might remark ... Turkey ISN'T in the EU) makes this even worse.
At least, everybody just sort of shuts up about these, as opposed to Hamas complaining Israel "is convicted of genocide by ICC" (which is a double lie: Israel isn't convicted, yet they, Palestine, ARE convicted, and ARE persisting in the crimes they are convicted of, no this isn't about the hostages, it's about earlier stuff). And they're screaming this everywhere. Now, I understand why they do it, but it remains utterly distasteful to lie like that.
EU states, including the EU itself, have blanket exceptions to the GPDR, for example. For example, they get to pass around private medical data, without telling anyone, supposedly in cases where they deem it to be important for investigations (not just criminal), anything to do with protection of minors, ... And because nobody cooperates well with this "for some reason", you find more than a few examples of them passing around (and using in court) WRONG private medical data. There has never been any conviction for this (DESPITE EU countries getting convicted for basing judgements on wrong medical data in their own EU courts).
I imagine something similar is going on here.
Of course, the only thing that's really in danger is the careers (and egos) of those managers. Internal competition would be an extremely good source of information on the performance of the business those managers manage.
So anything remotely competitive gets eliminated.
Which is what the original thesis was, I believe, that someone needs to lose.
But it gets a lot worse. The problem is that growth AND decline are both exponential processes. Not just growth. Decline too. Which means whatever percentage you say has to go to taking care of old people ... give it 20 years and it'll be more. A factor more. In fact, it will quickly rise to over 100% if population doesn't stabilize. In nature, for many species the exponential nature of shrinking population causes, by itself, without other causes, extinction. Luckily, usually local extinction. Cats are a great example of this. Areas almost always have a lot of cats ... or none.
If you take current examples (that are getting worse) you see that what happens is that every generation about halves. BUT 4-5 generations are alive simultaneously. So in order to keep the same level of elderly care the expenses need to double 5 times. There's a tiny problem ... even theoretically 20% expenses on the elderly can only double twice. The third time you double it, it goes over 100%. But when maintaining elderly care takes 100% of global human labor, we're only about halfway to a real declining population.
Or to put it more realistically, after all humans don't just undergo natural processes (thankfully), we adapt: unless population is growing at a minimum rate, the elderly will be abandoned. As soon as it becomes clear that this cannot be avoided, if needed we'll breed humans like in the Matrix to prevent this from continuing. Why? The people making that decision ... are the ones who are about to meet that fate.
This is be why a declining economy is so incredibly, incredibly destructive. Exponential growth, with a growth rate smaller than one, is a killer.