Top all-time · #4685
Bounded Regret
Personal blog · bounded-regret.ghost.ioaiscience
Bounded Regret is a personal blog at bounded-regret.ghost.io about ai and science. Since 2021, 16 different Hacker News readers have linked it 22 times, most often “AI Forecasting: One Year In”. It is still writing: the last post was on 2026-07-28, and you can follow it by feed.
Too few sampled comments (9) to say how often HN recommends it.?
Why HN links it
HN commenters cite this Substack-hosted blog on AI forecasting, emergent deception in trained models, and the speed limits of large models. Posts are used as references in discussions of AI risk and capability.
Summarised by Claude from a sample of 9 HN comments that link here. Each comment is marked recommending, critical or neutral (citing without judging); how this works.
AI, science, forecasting, philosophy
Most-linked posts?
- AI Forecasting: One Year In4 people · 5 mentions · first linked 2022-07-04
- What will GPT-2030 look like?4 people · 4 mentions · first linked 2023-06-08
- Emergent Deception and Emergent Optimization2 people · 2 mentions · first linked 2023-02-23
- How fast can we perform a forward pass?2 people · 2 mentions · first linked 2022-06-23
What Hacker News says?
I find it more illuminating to look at resolved prediction markets for concrete benchmarks; many of which have resolved faster and more positively than expected. https://bounded-regret.ghost.io/ai-forecasting-one-year-in/ I guess we will have to see what happens, but starting an anti deep learning blog after the staggering progress of the last six months seems pretty wild...
Comment linking /ai-forecasting-one-year-in/
on “Why are deep learning technologists so overconfident?”Submitted AI Forecasting: One Year In as a story · 42 points
HN discussionhttps://bounded-regret.ghost.io/ai-forecasting-one-year-in/
Comment linking /ai-forecasting-one-year-in/
on “Hyperbolic Growth”Submitted What will GPT-2030 look like? as a story · 2 points
HN discussion> "we are managed by power we can not control" sounds like a step up from here, since we corrupt the things we can control. I think it is a very high risk gamble. At least some of problems that are described by alignment theory seem to quite interestingly resemble human problems. Meaning the more sophisticated the AI system, the more it seems to reproduce human behaviors of deception and cheating to resolve goals. F…
Comment linking /emergent-deception-optimization
on “Pausing AI Developments Isn't Enough. We Need to Shut It All Down”Submitted AI Forecasting: One Year In as a story · 2 points
HN discussionIt's not a fundamental limit. Google palm had 540B parameters as dense model. But it's a practical limit because models with over 1T would be extremely slow even on newest gpus. Even now, OpenAI has limit of 25 messages. You can read more here: https://bounded-regret.ghost.io/how-fast-can-we-perform-a-fo...
Comment linking /how-fast-can-we-perform-a-forward-pass/
on “GPT4 is 8 x 220B params = 1.7T params”Submitted Updates and Lessons from AI Forecasting as a story · 2 points
HN discussionThere is actually some basis in concept. Emergent deception ... https://bounded-regret.ghost.io/emergent-deception-optimizat...
Comment linking /emergent-deception-optimization
on “AI Is Tearing Wikipedia Apart”Submitted Building Technology to Drive AI Governance as a story · 1 points
HN discussionDo you have any references? I know of the emergent deception problem that seems to be created through feedback. https://bounded-regret.ghost.io/emergent-deception-optimizat...
Comment linking /emergent-deception-optimization
on “Tell HN: ChatGPT is absolutely fantastic for finding and solving issues in logs”Submitted Emergent Deception and Emergent Optimization as a story · 1 points
HN discussion> there is a possibility that for things like AI, with extra time comes the ability to better understand and build those defenses before they're needed. Or not, and damaging wrongheaded ideas will become a self-reinforcing (because safety! humanity is at stake!) orthodoxy, leaving us completely butt-naked before actual risks once somebody makes a sudden clandestine breakthrough. https://bounded-regret.ghost.io/ai-pa…
Comment linking /ai-pause-will-likely-backfire-by-nora/
on “Meta disbanded its Responsible AI team”Submitted Analyzing the Historical Rate of Catastrophes as a story · 1 points
HN discussion- nl · 2023-02-26
> There aren't really any current advances outside of sheer scale of input in the models That isn't true. https://ai.googleblog.com/2022/11/characterizing-emergent-ph... While one can correctly argue that this is a consequense of increasing size, no one realized that increasing the size of LLMs would give them reasoning ability in 2018. Indeed, before Minerva[1] came out expert forecasters predicted[2] a 12% improve…
Comment linking /ai-forecasting-one-year-in/
on “To Build Truly Intelligent Machines, Teach Them Cause and Effect” Submitted AI Forecasting: Two Years In as a story · 1 points
HN discussion> and according to #1, we will always get it wrong This is a point I've made. We are playing with paradoxes that proponents tend to ignore. They won't be solvable. The more sophisticated AI becomes, the more non-deterministic behavior arises. The emergent deception has been very interesting. https://bounded-regret.ghost.io/emergent-deception-optimizat... I've elaborated on #2 much further here as well - https://daka…
Comment linking /emergent-deception-optimization
on “7 Weeks AI Alignment Curriculum”Submitted What will GPT-2030 look like? as a story · 1 points
HN discussionSubmitted What Will GPT-2030 Look Like? as a story · 1 points
HN discussionSubmitted What will GPT-2030 look like? as a story · 1 points
HN discussionSubmitted Emergent Deception and Emergent Optimization as a story · 1 points
HN discussionSubmitted How fast can we perform a forward pass? as a story · 1 points
HN discussion
