LessWrong 2.0 Reader

Mm I think it's hard to get optimal credit allocation, but easy to get half-baked allocation, or just see that it's directionally way too low? Like sure, maybe it's unclear whether Hinge deserves 1% or 10% or ~100% of the credit but like, at a $100k valuation of a marriage, one should be excited to pay $1k to a dating app.

Like, I think matchmaking is very similarly shaped to the problem of recruiting employees, but there corporations are more locally rational about spending money than individuals, and can do things like pay $10k referral bonuses, or offer external recruiters 20% of their referee's first year salary.

Where are all the successful rationalists?  [LW · GW]

https://x.com/JDVance/status/1854925621425533043

Is it too soon to say a rationalist is running the White House?

The constant bound isn't not that relevant just because of the in principal unbounded size, it also doesn't constrain the induced probabilities in the second coding scheme much at all. It's an upper bound on the maximum length, so you can still have the weightings in codings scheme B differ differ in relative length by a ton, leading to wildly different priors

Your phrasing is vague here (and also your reply contains a number of typos one of which seems to invert your likely intended meaning), so I have difficulty telling if what you say is simply misleading, or false. Regardless:

If you have UTM1 and UTM2, there is a constant-length prefix P such that UTM1 with P prepended to some further bitstring as input will compute whatever UTM2 computes with only that bitstring as input; we can say of P that it "encodes" UTM2 relative to UTM1. This being the case, each function indexed by UTM1 differs from its counterpart for UTM2 by a maximum of len(P), because whenever it's the case that a given function would otherwise be encoded in UTM1 by a bitstring longer than len(P + [the length of the shortest bitstring encoding the function in UTM2]), the prefixed version of that function simply is the shortest bitstring encoding it in UTM1.

One of the consequences of this, however, is that this prefix-based encoding method is only optimal for functions whose prefix-free encodings (i.e. encodings that cannot be partitioned into substrings such that one of the substrings encodes another UTM) in UTM1 and UTM2 differ in length by more than len(P). And, since len(P) is a measure of UTM2's complexity relative to UTM1, it follows directly that a UTM2 whose "coding scheme" is such that a function whose prefix-free encoding in UTM2 differs in length from its prefix-free encoding in UTM1 by some large constant (say, ~2^10^80), P itself must be on the order of 2^10^80—in other words, UTM2 must have an astronomical complexity relative to UTM1.

I have no idea how you're getting to this, not sure if it's claiming a formal result or just like a hunch. But I disagree both that there is a neat correspondence between a system being physically realizable and its having a concise implementation as a TM. Even granting that point, I don't think that nearly all or even most of these physically realisable systems will behave identically or even similarly w.r.t. how they assign codes to "natural" optimization criteria

For any physically realizable universal computational system, that system can be analogized to UTM1 in the above analysis. If you have some behavioral policy that is e.g. deontological in nature, that behavioral policy can in principle be recast as an optimization criterion over universe histories; however, this criterion will in all likelihood have a prefix-free description in UTM1 of length ~2^10^80. And, crucially, there will be no UTM2 in whose encoding scheme the criterion in question has a prefix-free description of much less than ~2^10^80, without that UTM2 itself having a description complexity of ~2^10^80 relative to UTM1—meaning, there is no physically realizable system that can implement UTM2.

I'll try to make it but I might be moving that day so I'm not sure :\

I think this page would be more useful if it linked to the individual sequences it lists.

As far as I've seen, there is no page that links to all sequences in order, which would be useful for working through them systematically.

I had a twenty hour drive by myself recently, and binged Otherness in the Age of AGI. It was tremendous. Ambitious to take on the entire thing in one session! 

Cost of inference is determined by the shape of the model, things like the number of active parameters, which screens off compute used in training (the compute could be anything, cost of inference doesn't depend on it as long as the model shape doesn't change).

So compare specific prices with those of models with known size^[1]. GPT-4o costs $2.5 per million input tokens, while Llama-3-405B costs $3.5 per million input tokens. That is, it could be a 200-300B model (in active parameters). Original GPT-4 is rumored to be about 270B active parameters (at 1.8T total parameters). It's OpenAI, not an API provider for an open weights model, so in principle the price could be misleading (below cost), but what data we have points to it being about the same size, maybe 2x smaller if there's still margin in the price.

I think the post was a deliberate attempt to overcome that psychology, the issue is you can get stuck in these loops of "trying to try" and convincing yourself that you did enough, this is tricky because it's very easy to rationalise this part for feeling comfort.

When you set up for winning v/s try to set up for winning. 

The latter is much easier to do than the former, and former still implies chance of failure but you actually try to do your best rather than, try to try to do your best. 

I think this sounds convoluted, maybe there is a much easier cognitive algorithm to overcome this tendency.

Realistic in what sense?

This proposal requires a majority vote by four state legislatures, and increases each state's influence in presidential politics.

A constitutional amendment requires a 2/3 majority of both houses of Congress and a majority in the legislatures of 3/4 of the states, and reduces the influence of about 20 of the states who are currently overrepresented in the electoral college.

Sully likes Claude Haiku 3.5 but notes that it’s in a weird spot after the price increase - it costs a lot more than other small models

The price is $1 per million input tokens (which are compute bound, so easier to evaluate than output tokens), while Llama-3-405B costs $3.5. At $2 per H100-hour we buy 3600 seconds of 1e15 FLOP/s at say 40% utilization, $1.4e-18 per useful FLOP. So $1 buys 7e17 useful FLOPs, or inference with 75-120B^[1] active parameters for 1 million tokens. That's with zero margin and perfect batch size, so should be smaller.