johnthewise

If you do so many things, some are bound to be stupid.

Waymos disengage and get tele operated too?

It's more of a bet on the optimus

https://x.com/ThierryBreton/status/1823033048109367549 Are we supposed to think these prior public threats are unrelated and X is really fined for changing a design on their website?

How do you know? Maybe he doesn't name these Elon?

referring to legality is self referential, since they enact the laws, everything they do can be legal. US could sentence commissioners to prison by enacting certain laws or declare war, we wouldn't certainly say 'they have a right to do that' in that situation.

Of course no one cares about random companies in Czechia or France getting pressured; it's not meant to sway public opinion in Sweden, otherwise it would have been a waste of influence (money). I think alephnerd operates on a higher level of abstraction in his commentary, and you mistake this as him making specific validity claims about the policies. I think your grievances stem from this gap in abstraction.

For example, he might personally support DSA/GDPR, but he says that the US generally views these as “non-tariff barriers” to US service companies[0] and doesn’t bother evaluating the policies themselves. essentially saying for the purposes of predicting how the US will react, it's sufficient to analyze how the US views them and the actual policy details lose relevance in that context. He also shared a detail[0] about how the US placed their lobbyists as commissioners on GDPR, which is an interesting operational detail that argues against the broad support argument you’re making. Another question is whether there would still be broad support for some policy after it has been enacted and its adverse effects have been felt.

[0] https://news.ycombinator.com/item?id=46170027#46174642

I found your commentary very tasteful, do you write on X or elsewhere?

It's easier to justify stock markets banning insider information because there are ignorant participants through their investment funds who we would like to protect. Why would we protect willing participants betting on arbitrary events? Even if we ban on this one too, should we in general be able to create a market that explicitly allows insider information for some arbitrary thing, insideinformationverywelcomemarket, where everyone is aware it's the main point of the market or shall we just protect these people from themselves?

why? The purpose of the prediction market is not to be fair to market participants, it's to aggregate information regarding an event. There is a public benefit to allowing participants to bet on insider information. It could be even argued there is nothing unfair about it if everyone is free to do it/can anticipate others might have insider info. If we actually limited the market to participants who have insider information(not feasible because we can't verify), that'd be a great public utility. this is the next best thing for all those who don't participate in it. These are specialized markets and we shouldn't rush to 'protect' people who bet on very technical events happening.

would you kill murderers before they murdered anyone?

Napoleon didn't fight much at all in battles, too.

Many people do. I can't imagine ever going back to opening ten tabs for a question. It's not like information on the internet was bullshit-free pre-LLMs. For me, it's easier to verify information than to find it.

You wouldn't even need to wait to act. Just pay/bribe people.

But they are very far of meaningfully replacing a human

Do you think its decades away far or few more years than what people extrapolate?

AI that drives humans out of workforce would cause a massive disinflation.

Dollar is agreement between humans to exchange services and goods. You wouldn't use USD to trade with aliens, unless they agreed to it. Aliens agreeing to USD would mean we have something to offer to them.

In the event of mass unemployment level AI, cash stops being the agreement between humans. At first, cash value of services&goods converge to zero, only things that hold some value are what AI/AI companies care about. People would surely sell their land for 1M$ if a humanoid servant costs 100 dollars. Or pass a legislation to let OpenAI build 400GW data center in exchange for 100$ monthly UBI on top of your 50$ you got from a previous 20GW data center permit.

A perfect next token predictor is equivalent to god

well, if you can arbitrarily nullify degrees, you can also decide the new degree is not valid.

on terrorism charges(to give more context on the audacity of the attempt)

2016 wasn't secular coup. the last of few secular ones were in 2000s, https://en.wikipedia.org/wiki/E-memorandum https://en.wikipedia.org/wiki/2008_Justice_and_Development_P...

The October Revolution wasn’t a popular uprising but a surgical strike by a minority faction. just after the take over, SRs were more popular but bolsheviks crushed them fast without much resistance. so it wasn't about what people wanted but what a determined minority could pull off.

AI subscriptions are already very sticky . I can't imagine at least not paying for one, so I doubt they care about retention like the rest of us plebs do.

It was frustrating before, then scary, now I think a lot of us are thinking about more material actions we can take to stop this non-sense.

don't listen to it or come up with something people prefer over it?

this doesn't argue whether existence of profits necessarily implies exploitation of workers but asserts it and then proceeds to argue against philanthropy funded by profits. This line of reasoning only makes sense if one already accepts the initial assumption, whereas the original poster questions that very assumption, so it's a bit irrelevant quote.

He seems to understand more than you suggest.

https://www.astralcodexten.com/i/136923606/is-musk-smart-doe...

He could be one of the greatest learners of all time as he is likely the greatest entrepreneur of all time.

Emergent properties are nice. They show CoT now, but who knows if there is a better planning strategy? Second thing is it kind of implies every base model can be increased in capability just with some RL tuning, cheaply. So in theory you can plug in every observable and quantifiable outcome beyond math and coding(stock returns, scientific experiment results?) and let it learn how to plan it to solve it better. Train on Observed effects of various drugs on people, it then creates a customized treatment plan for you? Sft version would be limited by doctors opinion on why certain drugs affected the outcome, whereas RL version can discover unknown relationship.

Do you think its not emergent because you think this behavior was explicitly coded in or you dont think its emergent because you dont like the implications of thinking it?

Yes, that’s kind of a given. The model has to have all the knowledge components to solve a task, so a capable base model is needed and only thing thats being learned here is how to stitch base knowledge to plan an attack. No amount of RL with a dumb base model would have worked for example.

I feel it’s not related to data or the architecture but the process of reasoning in general. For these models, every token predicted condition or drives the output in certain direction. Semantic meaning of these tokens have a magnitude in solution space. Lets say ‘answer is 5’ is very large step and ‘and’ token is very small. If you are looking for a very specific answer, these smaller nudges of each token generation provide corrections to direction. Imagine trying to click on narrow button with high sensitivity mouse settings, obviously you need to do many smaller moves whereas with a big button maybe you can one shot it. The harder or specific a task is where a solution space is very narrow that it cant be possibly one shotted, you need to learn to take smaller steps and possibly revert if you feel overall direction is bad. This is what RL is teaching the model here, response length increases(model learns to take smaller steps, reverts etc) along with performance. You reward the model if solution is correct, model discovers being cautious and evaluating many steps is the better approach. Personally I feel this is how we reason, or reasoning is in general taking smaller steps and being able to evaluate if you are in a wrong position so you cna backtrack. Einstein didn’t one shot relativity after all and had to backtrack from who knows how many things.