Top all-time · #18452
Mikel Bober-Irizar
Personal blog · anokas.substack.comaiprogramming
Mikel Bober-Irizar's personal blog at anokas.substack.com covers ai and programming. Since 2024, 4 different Hacker News readers have linked it 5 times, most often “LLMs struggle with perception, not reasoning, in ARC-AGI”. It is still writing: the last post was on 2024-12-24, and you can follow it by feed.
Too few sampled comments (1) to say how often HN recommends it.?
Why HN links it
HN readers link this post on ARC-AGI puzzles and whether AI can do maths. One commenter cites its o1 example as underspecified, supporting a wider point about ARC and IQ-style tests.
Summarised by Claude from a sample of 1 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
Some musings on AI, gaming, security, machine reasoning and whatever else I'm working on.
Most-linked posts?
- LLMs struggle with perception, not reasoning, in ARC-AGI3 people · 4 mentions · first linked 2024-12-24
What Hacker News says?
Just a comment: the example o1 got wrong was actually underspecified: https://anokas.substack.com/p/o3-and-arc-agi-the-unsolved-ta... Which is actually a problem I have with ARC (and IQ tests more generally): it is computationally cheaper to go from ARC transformation rule -> ARC problem than it is the other way around. But this means it’s pretty easy to generate ARC problems with non-unique solutions.
Comment linking /p/o3-and-arc-agi-the-unsolved-tasks
on “Can AI do maths yet? Thoughts from a mathematician”Submitted LLMs struggle with perception, not reasoning, in ARC-AGI as a story · 5 points
HN discussionSubmitted LLMs struggle with perception, not reasoning, in ARC-AGI as a story · 2 points
HN discussionSubmitted LLMs struggle with perception, not reasoning, in ARC-AGI as a story · 1 points
HN discussionSubmitted LLMs struggle with perception, not reasoning, in ARC-AGI as a story · 1 points
HN discussion
