Engineer at CoinList.co. Donor to LW 2.0.
ESRogs
Karma: 4,100
Oh, I had actually seen that paper. Forgot that they did that though. Thanks!
This makes me wonder, how would Monte Carlo tree search do for GPT? And could you do AlphaGo-style IDA?
You’d need an analogue of the value network (or value head). (Where current GPT seems analogous to the policy network.) And then ideally you’d also want some analogue of winning / losing to ground out the evaluation.
Maybe you could set it up like this --start with a task description like, “write a poem in the style of e.e. cummings about the romance between cryptographers Alice and Bob”
feed the task description (with some boilerplate) into GPT, and have it start generating continuations
do MCTS on the continuations; use your value network (head) to evaluate the continuations vs the task description; update the policy network based on the evaluations
include an “is done” head and evaluate it to decide when to stop
send completed works to humans to provide feedback; the feedback should include separate scores for “good so far” for the value head, and “is a completed work” for the “is done” head.
I’d be curious whether this would enable GPT to significantly improve. Specifically, would you be able to generate longer works with less intervention?
G: A 50⁄50 gamble between A and B
This bit in the first drawing should say ”… between A and C”, right?
It’s totally possible that there’s no way to get to reliably correct answers and you instead want decisions that are good regardless of what the answer is.
Good point!
It seems like teaching that fact, and instilling moral uncertainty in general into children
I would guess that teaching that fact is not enough to instill moral uncertainty. And that instilling moral uncertainty would be very hard.
Sounds like you probably disagree with the (exaggeratedly stated) point made here then, yeah?
(My own take is the cop-out-like, “it depends”. I think how much you ought to defer to experts varies a lot based on what the topic is, what the specific question is, details of your own personal characteristics, how much thought you’ve put into it, etc.)
Options 1 & 2 sound to me a lot like inside view and outside view. Fair?
Also asked on Twitter here.
If GPT-3 is what you get when you do a massive amount of unsupervised learning on internet text, what do you get when you do a massive amount of unsupervised learning on video data from cars?
(In other words, can we expect anything interesting to come from Tesla’s Dojo project, besides just better autopilot?)
the FDIC does not directly buy the spread; they allow the CD to exist, which the buyers should have access to
I’m not sure how important FDIC insurance is to the story here, but worth noting that it has a 250k per account limit. So I don’t think financial institutions would have access to an unlimited amount of it.
But if selling the box spreads doesn’t change the account value much—if the brokerage marks you as having $100k cash but also a ~$100k liability—then after withdrawing, the value is lower than before you sold the spreads.
This is correct. The brokerage shows cash + a liability.
Not sure I’ve understood your question, but if you withdraw from your account, the value goes down, whether you’ve sold a box spread or just have a margin loan.
Worth noting that you may also want to use this trick if you are using portfolio margin to lever up your portfolio more than 1x. Instead of withdrawing to a CD, you leave the cash in your account, and end up with a lower margin rate than you’d otherwise get.
Hmm, good question. I actually don’t have a good sense of how much of that is assets that remain on the balance sheet (e.g. manufacturing equipment) vs stuff like paying their employees to figure out how to make the batteries better, or how to set up the factory more efficiently.
And paying employees to figure stuff out would show up as just costs on the balance sheet, rather than assets, unless you actually patented something, right? (I don’t actually know accounting super well.)
EDIT: It’s intuitive to me though that when you’re growing revenue at 50% annually (as Tesla has since 2013), you’re just not going to be able to spend money as efficiently as when you’re operating at, or close to, the same scale from one year to the next. (That is, efficiently in terms of short-term profit and loss. From a long-term perspective it might be very efficient, if the spending is enabling future growth.)
I’m not sure exactly where that’s most likely to show up on an accounting statement. But I do think it’s what you’d expect by default. And it’s how startups operate. You spend to grow, and you don’t expect to be profitable right away. Tesla should perhaps be thought of as a rare public company that still operates like a high-growth startup.
I run the discussion forums where we’ve been continuously open to debate and questions since before LW existed.
Which discussion forums are you referring to?
The blackmailers need to end up sometimes disclosing in order for their threats to be credible, no?
This is a series of scenarios describing a human taking actions in the world, designed to test physical and common-sense reasoning.
Nitpick: why is this part bolded? Surely this was written by you and not GPT-3, right? (It’s becoming a new pet peeve of mine when people are not super clear and consistent with their formatting of GPT-3 conversations. I find it often takes me a significant amount of effort to figure out who said what if a convention is not followed religiously within a transcript.)
I’m sure I could go on for quite a while. Legalizing blackmail means that people are de-facto incentivized to exploit information when it would harm people, because their payout stops being derived from the public interest
What if blackmail was only made legal in cases where there was a significant public benefit? (See here.)
I believe that would still be a departure from laws today (you’re not allowed to pay off whistleblowers, are you? or to accept such payment?), but seems like it might not have the downsides that you and others have highlighted of incentivizing finding ways to hurt others by sharing information in general.
Suppose it were easy to split potential blackmail scenarios into whistleblower scenarios (where the value of the information to society is quite positive) and embarrassing-but-useless scenarios (where it is not).
Would you support legalizing blackmail in both classes, or just the first class?EDIT: I ask because, I think (at least part of) your argument is that if we legalize paying off whistleblowers, then that’s okay, because would-be-whistleblowers still have an incentive to find wrongdoing, and the perpetrators still have an incentive to avoid that wrongdoing (or at least hide it, but hiding has costs, so on the margin it should mean doing less). (This reminds me a bit of white hat hackers claiming bug bounties.)
Meanwhile, the anti-blackmail people argue that you don’t want people to be incentivized to find ways to harm each other.So, if you could cleanly separate out the public benefit from the harm, on a case-by-case basis (rather than having to go with simple heuristics like “gossip is usually net beneficially”), it seems like you might be able to get to a synthesis of the two views.
Taking the Bay’s weather to be a negative? Melbourne must really be idyllic...