askjdfksdbfhk

I agree that for average web dev tasks the open models are already good enough. I've had good experiences with both DeepSeek and GLM. And these models are just better for anything security related since they don't throw massive hissy fits.

However I do actually have a project where I need the frontier models--I'm working on a deep learning project of moderate complexity (something novel/state of the art within its domain, adapting a known approach from published research in a related domain). The difference from Opus 5 -> Opus 5.5 was huge for my project. Opus 5 was struggling, Opus 5.5 is doing really well.

I think the demand for frontier models will continue to be there, at least for a subset of tasks, although I agree that it is probably going to shrink as the non-frontier becomes more and more capable.

You seem like a bit of an asshole so I'm not really interested in continuing further, but I will say that a large amount of that activity is financed by demand from students in one shape or another.

I don't really see much "reasoning" that they are presenting. Their contributions to this thread have been closer to sealioning than any kind of positive argument or thesis. What do you see in their comments that the rest of us are too stupid and dishonest to engage with, exactly?

>As someone who did math professionally for a long time, I agree with you. I think it has as much value as chess (a game we made up which we play for fun).

I mean, I don't see how this is much different from what I argued, no? Nobody really cares about computer chess all that much at this point except as a benchmark of technological achievement, and except to the extent that it helps human players better understand the game and improve their own play. We still play chess, even though computers can do it better than we can, because we find it interesting.

I have no idea what you're trying to say. Once again, if you see value in it then feel free to explain why.

The only reason most pure mathematics research exists right now is because it is interesting to the mathematicians who conduct it. I think a black box is not interesting to mathematicians and therefore has little worth. That's the argument.

There's a weird thing going on at 22nd and San Jose in the Mission, where it doesn't seem to understand that you can simply continue down 22nd, but instead takes a hook down San Jose before backtracking. I see some other weird discontinuities like this as I click around.

I have to say that the Claude UI has made many inexplicable decisions here, including (but not limited to) the mysterious color coding, the confusing continuous slider over a discrete set, and the weird positioning of the height labels in the altitude graph.

I think I already answered this in my comment. I claimed that I don't see much value in a magical black box theorem-proving oracle whose proofs could not, and would not, be understood by humans.

If you disagree (as the GP does, apparently), I ask you--what value do you see in it?

>AI understanding of math is millions times more important than human's.

This is a persistent misconception that I see on HN (although not usually framed as abrasively as you have chosen to frame it).

Math is not an engineering discipline. Pure mathematics research is not done with an eye towards practical applications in other fields. You should think of it more like the humanities--it is done because it enriches the human experience (and it's not very expensive).

You could give us a black box oracle that could tell us, with 100% certainty, whether the 50 most famous and important open conjectures in math were true or false, and it would be more or less worthless. What does it actually get you? This isn't a rhetorical question.

AI in math is valuable only insofar as it ultimately helps drive human understanding.

There is another side of math that is more driven by practical applications: statistics and "applied math" and whatnot. This is what most people think of when they try to imagine math research, because it is much closer to the math that they have learned in their school education. A lot (but not all) of this math would get more value out of black box results, yes. But don't conflate this with the entire field of mathematics as a whole.

>LLMs have allowed double-dummy solvers to move from ~0.1sec to microseconds (if you allow 99.9% accuracy)

I don't know what this is referring to. The ubiquitous dds library, which is AFAIK basically the only double dummy solver, has not seen any real improvements. Neither of those numbers look right to me, I think it's more in the range of ~10ms.

There's a crank who was claiming some magical improvements a little while ago that really just boiled down to AI psychosis (and having absolutely no understanding of what he was claiming). I hope that's not what you're referring to.

Monte Carlo with a double-dummy solver is fundamentally insufficient for good single-dummy play because it is incapable of understanding information. It won't take discovery plays (lines aimed at discovering more information about the opponents' hands before choosing a line of play) and will systemically overvalue positions which are good double dummy but require a guess. It doesn't understand falsecarding (because double dummy, it doesn't matter).

I agree that many games could make progress if people were actually inclined to try.

I think it will get a bit better in the coming decade thanks to continued hardware improvements & powerful LLM coding agents making it more feasible for amateurs to tackle these things at home. Personally I've been working on a game AI project for the last month at home based around published techniques for a similar game, using my 5090 for training and Opus for implementation and orchestrating tasks and so on. It's going quite well and it looks like I'm on track for a SOTA, superhuman AI at the end. Doing this ten years ago would have been incomparably harder.

>There is nothing that, in principle, makes MTG different from poker or bridge, and we have superhuman engines for both.

We don't have superhuman play for bridge.

Poker and bridge are quite different from each other in terms of solving them. Among other things, the hidden information space in poker (at least, in hold'em) is far smaller than in bridge (or Stratego, for that matter, as discussed in the linked paper). This makes hold'em solvable using CFR, an algorithm which essentially optimizes play by considering all the possible holdings than the opponent might have and their best strategy with each one. Even going from two to four hidden cards per player (Omaha) requires a slightly different approach although you can still use CFR as the basis for the search algorithm.

Bridge has 13 hidden cards per player which makes CFR basically impossible to apply, at least in any obvious way--just way too many states. Similarly you see it's not used at all in this Stratego paper.

>So well written I would have clocked it as AI

Still a few humans writing on the internet... at least for now ;) I'm sure the green username doesn't help either, I just switched to a new account a few days ago.

>This is standardized at tournaments (I'm sure you know that.)

Right, in human play we have convention cards (pieces of paper that are essentially big forms for specifying common agreements), although these don't cover every situation.

What I was getting at was more along the lines of some kind of general schema that would allow fully describing any individual bid. But then there's also a problem where such a schema inherently limits the creativity available in constructing a bidding system, if all the bids must fit into what is expressible by this schema.

My personal approach would be to try to decompose it into two subproblems: learning a set of agreements, and optimizing results given a fixed set of agreements. Then you could try to solve the former problem using the results from the latter one. But even playing well under a fixed set of agreements isn't so easy to solve.

I would really love to see a serious research effort take a crack at contract bridge. Bridge, like Stratego, is an imperfect information game with a big hidden information space. Bridge also adds another wrinkle of explainability which is, I think, very interesting.

Bridge is played as a pair vs pair game, with North/South and East/West being the two pairs and seated around the table in these compass directions. A bridge hand consists of two phases: there is first an auction phase, where players go around the table bidding on contracts (agreeing to take a certain number of tricks with a certain trump suit) until a final contract is decided. Then there is the cardplay phase, where the player who won the auction is the declarer, their partner is the dummy, and the other pair are defenders. The dummy's hand is placed face up on the the table and the declarer controls which cards are played from dummy, so the cardplay phase is effectively played by only three players now, with each of the three knowing one common hand (dummy) and one private hand (their own) and not knowing the other two hands.

In both the auction and (for the defense) the cardplay phases, it is important for players to exchange some information about their hand to their partner. However, any information you exchange about your own hands also helps your opponents. You might naturally conclude that you want to come up with some secret scheme to exchange information which your opponents don't know (and it is even possible to exchange encrypted information which your opponents can't know--if the defense is known to hold a certain card, but declarer doesn't know in which hand it is, the defense could say that a signal means one thing if the card is in one defender's hand, but means a different thing if it's in the other defender's hand).

But it turns out that this ends up being very uninteresting to play, so instead, when playing bridge, there is an important rule: all of your partnership agreements must be public. If a certain bid that I make promises that I have at least 5 spades in my hand, it is the opponents' right to know that this is our agreement. You must be able to explain the information which your action provides, and you must be able to use the information that the opponents give you themselves.

This poses several problems for self-play reinforcement learning. First, a naive self-play approach will produce agreements that cannot be explained to a human. What really needs to happen is that your partner, when determining what hands you might have as part of search, must not do so simply by sampling its own system (ie by asking what it itself would have done with hand X or hand Y). The information and possibilities really need to be mediated by some kind of intermediate, rules-based description, which can be provided to the opponents as well.

You also need to be able to encode and ingest the opponents' agreements, and to use this information to inform your own decisions. And you need, in particular, to be able to handle a wide variety of agreements from your opponents; it's not enough to force them to play the same system as you.

You must also account for deceit. If, for example, I have a bid which promises that I have at least 2 cards in every suit, it's perfectly legal for me to lie and make this bid when I only have 1 card in some suit--as long as my partner is in the dark about this just as much as the opponents. So if you make this bid, and your machine opponents assume there is a 0% probability of you having lied about your hand, it is possible that they will make gross errors by not accounting for this possibility (for example, they may be in a position where all of their actions are equivalent if you told the truth, but where one action is clearly better if you didn't--a human player will naturally take this action, but a robot may just select an action randomly).

It's an interesting game and a very interesting AI challenge.

Yes, you're right, although there was a lot of churn in SvelteKit when it was in beta--maybe they had that in mind. Of course, it was marked as a beta, so it's hard to blame them too much. But I was using Kit heavily during that period and ended up getting frustrated by the whiplash, and then the addition of runes (which was announced in late 2023 even if it wasn't released until 2024) was what made me lose interest in Svelte. I still have some fondness for the framework though.

>Ok, so does this mean prices in the US will go down? I keep seeing stuff about countries shifting away from US goods but prices seem to continue to rise.

The actual answer is yes, at least when it comes to wine (although I will note that the trade war stuff is only part of the picture--it has certainly hurt a number of wineries that previously relied on exports, but the wine industry is experiencing a much broader downturn than just this).

Not all wine, of course, but some wine is indeed seeing prices drop. There have been some great deals to be had in the past few years in certain corners, both big and small. Costco has had increasingly good deals on wine lately and other large retailers are having more and more as well.

There's a lot of price stickiness with wine, with wineries not wanting to lower sticker prices (since price is closely tied to branding and market positioning), so there's also been this fascinating trend of wine increasingly being bulked out to online retailers that operate on a rotating flash sale model (companies like Last Bottle or Wine Access), or to third parties who relabel it and sell it under an NDA on the original brand. These segments of the market are only going to continue to grow in the coming years as the glut of wine continues.