I think when people think about CRTs and NES/Atari, they aren’t thinking about _monitors_ they are thinking about the low quality TVs and hooked up through the antenna adapter, which absolutely did wash pixels out to some extent.
LLMs need a lot of help to get to any kind of complicated proof, really. And yes, they get in moods. I spent 3 weeks trying to prove something and frequently had to try and convince Claude that it wasn't impossible and that it could really do it.
Yeah I think to calibrate you should probably look to Native American words that made it into English — place names, local foods and wildlife, cultural items. Those are going to be your substrate borrowings for most languages.
Also, “not having a known etymology” is weak evidence for borrowing, full stop, let alone for borrowing from a substrate. There are lots of very common words in english (dog, pig, boy, girl) that have no known etymology which are incredibly unlikely to impossible to have come from a substrate language.
Languages invent new words and sound patterns endogenously all the time.
English dog has had an etymology for the last 20 years or so (Gasiorowski's) that many find convincing. But with regard to Greek lexicon, it is not just the lack of etymology that suggests substrate but the non-IE phonology of many of these words.
I sort of have this suspicion that the kinds of phenomena that we can even apprehend are pretty closely related to what we can reason about mathematically, and the reason that math seems to work so well to model behavior that we understand says more about how our brain works than how the universe works.
If a phenomenon isn’t model-able by relatively simple math it doesn’t even look like a phenomenon to us, it looks like a chaotic mess.
Which is to say that unless something has mathematical coherence it is more or less invisible to our experience.
The thesis is that Science is made for efficiency and hence does not include everything (which is infinite). It only looks for a hierarchy of interesting facts and focuses on simple foundational recurring phenomena from nature. It then uses the language of Mathematics to economize and impose order on the complexity to make it tractable.
Finally, our subconscious prefers aesthetic attributes and hence we often find/create harmony/symmetry/beauty in our mathematical products.
Perhaps symmetry is overrated and it is not the beauty that works, but it's economy. If something is symmetric depending on how symmetrical you can remove half or more of the required work during calculation. It is also useful to work in abstractions where it does not related to the system 1:1 but is some transformed representation (e.g. phase space). And it may be useful when you find a way to transform it to some easily solvable (symmetrical?) state, solve it there, and reverse the transform. Think of working in Cartesian and polar coordinates. Just a change in coordinate representations can improve your quality of life immediately on some problems with different symmetry. There's no intrinsic reason other than its much simpler on one than the other. Beauty doesn't have to come in. If it's round it's round. If it's square it's square.
So I think symmetry might just be one of the tricks that keep working and keeps on giving back and so we love it and call it "beautiful".
Well there are counting arguments to say that a higher level intelligence that somehow achieves say, 100000× brain efficiency of humans, still can't do that much more work than humans, if humans found the "best abstraction". Think about computability for example - that means you could feasibly solve problem instances of n+20 relative to what a human can solve.
Of course, I think putting numbers to wishy washy meta-quantities like "how efficiently does a certain conceptual scheme help you" are super loaded and hard to properly talk about (incommensurability). I've been toying with trying to make a repository of all the possible "moves" one can make in this kind of abstract analysis - constrain the problem statement, argue something like "the system is what it does", dissolving, etc. but even that seems hard
If the goal is merely to "win at chess", then yes, an LLM using stockfish is better than any human alone at performing the task. When you are talking about what AI agents are capable of doing, there is no such thing as "cheating". They are as capable as the tools they can use effectively. The entire history of human civilization was driven by effectively using tools to achieve goals.
Sure, but the LLM is free to construct a representation of the chess board and update it as it goes along. It is not in any way banned from using a virtual board, or whatever representation of game state it pleases.
AFAIK, current models will still sometimes make illegal moves even if given the entire game state (e.g. in FEN notation), so it is not purely an issue with the models’ ability to keep track of sequences of moves.
It isn't. Stockfish running on your laptop can beat every human being on earth easily at chess. It's not intelligent _at all_ in any sense that matters.
Well it's not a perfect test so you need a bit of care in how you use it. If you have no idea what the subject is doing, then you don't know if you're measuring cognitive ability or something else (like cheating ability, or algorithmic sophistication or whatever). But failing the test is a pretty clear sign of certain cognitive abilities being poor.
I want you to consider how relevant this is in any practical sense.
First -- most _people_ cannot do this, without having a physical board in front of them.
Second -- Claude Code is perfectly capable of downloading and running stockfish. People focus too much on LLMs by themselves as the entity of concern instead of the entire harness and all of it's capabilities together.
reply