biophysboy

They use a number to make the “prediction” seem more credible and “rational”

He is much more full of shit.

Let’s go back to the good ol days when half of movies were about cowboys.

If only we had all had such principled foresight. Thank you for this great lesson.

It is not an ad account - the other reply is stupid. I am saying ad accounts deal with this crap from FB all the time. Add this instance to the pile. It is the price of surrendering comms and marketing to a gigantic behemoth.

Autobans happen to ad accounts on FB all the time.

I am 99% sure it was an automated system false positive, but it is annoying that we all have to rely on these capricious systems. That is in itself a problem with massive scale platforms. It confirms my silly hot take that all online communities should not be bigger than 10k people. That keeps the number of bad eggs legible to an actual human moderator, and keeps the discussion less mob like

Yes, and I think over time there will be a blatant inequality in who is commoditized and who isn’t.

If AI engineers didn’t create the data to produce the models, and they increasingly don’t write the code to produce the models either, then why are they treated as the agents of change rather than the subjects of change, like mathematicians are?

what does SI stand for, subordinate imbecile?

I have been kind of disappointed by AI science-wise. I was anticipating more advancement. Experiments do take time, so I suppose they could be pending.

What attracted you to the field? (I’ll see myself out)

Same experience. The question "what is your most memorable experience" also kind of annoyed me, because I see AI mostly as a tool. It is like asking me what my most memorable experience with my iphone is. I'm sure I could cook up something, but it felt like they were mining users for happy PR.

People start with this prior in the examples you gave when they are completely uncertain and have no better mental model. It is not a good starting point for an event that has never happened and can only happen once.

Uninformative priors are a part of Bayesian models, which are useless if they cannot be updated with real or simulated data?

> But research has shown that people translate vague phrases like "a decent chance" into probabilities in inconsistent ways.

That is not a bad thing. It captures uncertainty, unlike the fake number.

> For an ML researcher who is already used to dealing in next-token probabilities that aren't rigorously determined, just stating a probability estimate directly is very natural.

Exactly, it is a rhetorical device to persuade a technically-inclined audience. It works because it implies that a quantitative model exists. I want a clear, incisive set of mathematical arguments. Otherwise, I’m ignoring predictions as the ramblings of arrogant idiot rich kids.

I think the reason for this is that the group who has the authority to do the former is much larger than the group that can do the latter. The group who could actually build safeguards seems to have no free time and is constantly being whipped to go faster and win the race.

There’s nothing wrong with bold predictions, but they should be paired with good methods. The doom predictions are not paired with good explanation.

Where does 50% come from? It is meaningless if the probability model is not explained.

My org has a leaderboard for AI spending each month, and I have found it interesting how fast the distribution decays, just within the top 10 users. I often think “what did these people do with all those tokens?” It’s interesting to think the answer to that question is “maybe not a lot?”

Why should I believe this 2.8B matters relative to the trillions put into the AI buildout? All of the "coordinated actions" from this camp - public resignations, hacking scandals, joint calls to "pause" - don't seem to have done anything. So far, it has been a lot of ineffectual hyperventilating.

In any case, I agree the p(doom) sci-fi is annoying secular milleniarianism. SV hyperfixates on imaginary futures. If they actually cared about safety, they would be using all this money to strengthen global cybersecurity, instead of writing LessWrong posts that gives kids in their 20s ulcers.

There are two things that make me sad about AI:

1) the mass proliferation of low-effort content. A thing can be implemented quickly and correctly but still be badly designed. When people say "oh this will go away as the models improve", I think they are making a category error. The AI cannot read your mind.

2) we really have no idea what to do about how this transforms society. As a result, the debate has been captured by quasireligious doomsday cultists and self-interested accelerationists. Obviously we should improve our cybersecurity and synthetic biology regulations (although I'm a skeptic on the bioterrorism risks). But we had no idea how much time people would pour into social media feeds when they were invented. I am sure we will be careless with this technology as well in this 2nd gilded age.

Can you give an example? I'm not dismissing - I just can't think of one myself

I enjoyed the book too, but as I read it, I felt more and more like he could have just as easily found examples where making a system simpler and more legible led to massive improvements.

I’m reading William Vollmann’s Imperial (a book about the eponymous county in CA) , Richard White’s “The Republic for Which it Stands” (a history of reconstruction and the gilded age), and Paul Starr’s “Social Transformation of American Medicine”.

“First principles thinking” is a Silicon Valley cliche. It is a tell for me

> As my regular readers will be aware, I have been engaged, on behalf of my clients, in legal combat with Internet censors around the world, agencies who think that they have both the standing, the competence, and the right to tell American companies what software they can write and run, for the better part of 18 months.

Is this supposed to make me trust their opinions on tech regulation?

David Sacks is one of the most predictable VCs in Silicon Valley. He could have written "I still have the same thoughts; here's a link to my first and only opinion" and the post would have communicated the same amount of info.

I understand that AI is not just cut and paste, but some documents will have more influence than others w/ power law scaling. I would be very surprised if this distribution were not extremely steep for arcane math

Well the argument is that B is no longer sustainable because of the scenario you describe in A. If you do a bunch of work on Navier Stokes but then OpenAI gets all the press, then what was the point?

I know this happens, but my impression during my phd was that many labs use popular methods to test popular questions, leading to a lot of simultaneous work. You see this in history as well.