qnleigh

6 ms·

91.
by qnleigh 7mo ago
You might say "but wait, haven't we just proven that it's true? So isn't that also a contradiction?" This would be a disaster, because it would prove that the axioms of arithmetic are inconsistent! Now 1+1=3 for all
92.
by qnleigh 7mo ago
I feel like it's nice to get the gist before diving into the gory details. The proof works by showing that just within the axioms of arithmetic, you can formally state the sentence "this sentence is unprovable." This has some
93.
by qnleigh 7mo ago
Yeah, I think it would be better to first explain the liar's paradox to give the broad brush strokes, and then go into the details of Gôdel numbering. It seems like most expositions of Gödel's incompleteness theorem go into a surp
94.
by qnleigh 7mo ago
I don't understand your question. Can you elaborate?
95.
by qnleigh 7mo ago
This comment is wildly inappropriate and violates the community guidelines here. I suggest you delete it.
96.
by qnleigh 7mo ago
This is a good question, and currently the answer is no. Quantum computers can only run very short, simple algorithms right now, because the qubits they're built out of are noisy. You need a lot of error correction, which the community
97.
What Gödel Discovered (2020)(stopa.io)
97 points by qnleigh 7mo ago|37 comments
98.
by qnleigh 7mo ago
Where? If you mean that a computation was performed without creating any entropy, I'm skeptical. Surely some energy input was needed, and some energy was dissipated.
99.
by qnleigh 7mo ago
I haven't finished reading this yet, but I don't think the author is saying that the Landauer limit for erasure is wrong. They're saying that there are other limits in computing beyond erasure. I think this makes sense; altho
100.
by qnleigh 7mo ago
I did mention RL as a valid counterargument in my comment. I agree that in verifiable domains RL systems should be able to blow past human performance, and this might already be happening. There's another interesting question as to how
101.
by qnleigh 7mo ago
In the paper, they give part of their system prompt: > * After EVERY exploreXX.py run, IMMEDIATELY update this file [plan.md] before doing anything else. * No exceptions. Do not start the next exploration until the previous one is docume
102.
by qnleigh 7mo ago
> That will help save enormous amounts of power: up to 48 percent on a single charge, Why does refresh rate have such a large impact on power consumption? I understand that the control electronics are 60x more active at 60 Hz than 1 Hz,
103.
Building superconducting and neutral atom quantum computers(blog.google)
2 points by qnleigh 7mo ago|0 comments
104.
by qnleigh 7mo ago
I've found notebookLM summaries to be too high-level and oversimplified to be useful. Hopefully in a few years they can go deeper.
105.
by qnleigh 7mo ago
This is why I put 'learned' in quotes. They started from a state of not being able to solve algebra problems or produce basic steps of scientific reasoning to being able to. Operationally, that is what I mean by learning and they
106.
by qnleigh 7mo ago
Actually here's an even better list of progress on a number of open math problems, with plenty of caveats and exposition: https://github.com/teorth/erdosproblems/wiki/AI-contribution...
107.
by qnleigh 7mo ago
Where can I read about how LLMs have changed epistemology? Is there a field of philosophy that tries to define and understand 'intelligence'? That sounds very interesting.
108.
by qnleigh 7mo ago
> I think "novel" is ill defined here That's exactly my point. When people say "LLMs will never do something novel," they seem to be leaning on some vague, ill-defined notion of novelty. The burden of proof is th
109.
by qnleigh 7mo ago
I might as well answer my own question, because I do think there are some coherent arguments for fundamental LLM limitations: 1. LLMs are trained on human-quality data, so they will naturally learn to mimic our limitations. Their capabiliti
110.
by qnleigh 7mo ago
Yup, I agree with this. So based on this, where do you draw the line between what will be possible and what will not be possible?
111.
by qnleigh 7mo ago
> It doesn't discern between them, just looks for the best statistical fit Of course at the lowest level, LLMs are trained on next-token prediction, and on the surface, that looks like a statistics problem. But this is an incredibly
112.
by qnleigh 7mo ago
I am kind of amazed at how many commenters respond to this result by confidently asserting that LLMs will never generate 'truly novel' ideas or problem solutions. > AI is a remixer; it remixes all known ideas together. It won&#
113.
by qnleigh 7mo ago
That sets a vastly higher bar than what we're talking about here. You're comparing modern AI to one of the greatest geniuses in human history. Obviously AI is not there yet. That being said, I think this is a great question. Did E
114.
by qnleigh 7mo ago
Do you know that from reading the proof, or are you just assuming this based on what you think LLMs should be capable of? If the latter, what evidence would be required for you to change your mind? - Edit: I can't reply, probably becau
115.
by qnleigh 7mo ago
Their 'Open Problems page' linked below gives some interesting context. They list 15 open problems in total, categorized as 'moderately interesting,' 'solid result,' 'major advance,' or 'breakthr
116.
by qnleigh 7mo ago
But human researchers are also remixers. Copying something I commented below: > Speaking as a researcher, the line between new ideas and existing knowledge is very blurry and maybe doesn't even exist. The vast majority of research p
117.
by qnleigh 7mo ago
Speaking as a researcher, the line between new ideas and existing knowledge is very blurry and maybe doesn't even exist. The vast majority of research papers get new results by combining existing ideas in novel ways. This process can l
118.
by qnleigh 7mo ago
I got Gemini to find a polynomial-time algorithm for integer factoring, but then I mysteriously got locked out of my Google account. They should at least refund me the tokens.
119.
by qnleigh 7mo ago
For number of mathematicians familiar with and actively working on the problem, modern mathematics research is incredibly specialized, so it's easy to keep track of who's working on similar problems. You read each other's pap
120.
by qnleigh 7mo ago
I hope any hold-outs who aren't convinced yet will be after reading this comment! Did you wind up sticking with Windows (or Mac) for a long time after this? How long until you tried again?

More ›