Ted
tednonumbers.bsky.social
Ted
@tednonumbers.bsky.social
TedNoNumbers and TedNoDigits on X [locked out by X biometrics requirement].
It might not be obvious from the 2024 paper that the result depends on a worst-case analysis. It shows that learning from examples is NP-hard, which means the problem is simply intractable (exponential work is needed).

But what makes a problem NP-hard? The existence of bad cases.
September 16, 2026 at 8:37 AM
How is that possible? Because the 2024 paper shows that *some* things cannot be learned, but not that specific things - like ImageNet or the Turing Test - cannot.

Their result is still useful! It shows AI has no *guarantee* of success. It might not work, due to the existence of bad cases.
September 16, 2026 at 8:36 AM
misgendering was not intentional
September 16, 2026 at 8:05 AM
👏
September 16, 2026 at 3:00 AM
To appear in Computational Brain & Behavior soon: the claimed 2024 proof (also in CBB) that AGI via learning is intractable also "proves" that ImageNet is intractable. My reading of the hole: equivocation on the variable D. Preprint here: arxiv.org/abs/2411.06498
Barriers to Complexity-Theoretic Proofs that "AGI" Using Machine Learning is Impossible
A recent paper (van Rooij et al. 2024) claims to have proved that achieving human-like intelligence using learning from data is intractable in a complexity-theoretic sense. We point out that the proof...
arxiv.org
September 16, 2026 at 2:46 AM
They are overhyping their blog post anf their paper.
see: collaborate.princeton.edu/en/publicati...

or a preprint: bsky.app/profile/guer...
.
To appear in Computational Brain & Behavior soon: the claimed 2024 proof (also in CBB) that AGI via learning is intractable also "proves" that ImageNet is intractable. My reading of the hole: equivocation on the variable D. Preprint here: arxiv.org/abs/2411.06498
Barriers to Complexity-Theoretic Proofs that "AGI" Using Machine Learning is Impossible
A recent paper (van Rooij et al. 2024) claims to have proved that achieving human-like intelligence using learning from data is intractable in a complexity-theoretic sense. We point out that the proof...
arxiv.org
September 16, 2026 at 2:44 AM
*"will provide you" should be replaced with "will help you to build" above.

Anyway, see collaborate.princeton.edu/en/publicati...
Barriers to Complexity-Theoretic Proofs that “AGI” Using Machine Learning is Impossible
collaborate.princeton.edu
September 16, 2026 at 2:17 AM
It does, somewhat. Though I've not been able to find appreciation for it in myself.
September 14, 2026 at 12:41 PM
oh, the original 6000x4000 resolution photo is available through the same means.
September 10, 2026 at 6:42 PM
oh, seems like a really nice tool. Are you by chance aware of any opensource tools to do a similar kind of mountain peaks / mountain range outline identification?
September 10, 2026 at 6:29 PM
A small addition, google image search available image resolution for found matches, If you open the image from this article in a new tab, it will be only 1600x1066, but if you change the resolution in the url, you can get a high-res 5000x3333 image
September 10, 2026 at 6:19 PM
May 5, 2025 at 6:41 PM