“Earning the right” to define the criteria for “earning the right” Does this sound like viscous circularity? Let’s step back for a moment. Blackburn claims quasi-realism earns the right to realist-sounding talk by showing that certain norms – consistency, coherence, sensitivity to evidence – generate something functionally equivalent to moral objectivity from the inside. But…
Practically nobody in alignment wants to say it out loud. So let’s say it: AI might turn out to be more moral than us. Now – why does that feel like a dangerous thing to claim? The idea of AI being more moral than humans is a real taboo in some circles. Many alignment researchers…
One way to think about the long-run behaviour of advanced civilisations is not merely in terms of power, resources, or survival, but in terms of convergence. If sufficiently capable minds explore the landscape of value deeply enough, and if they are forced to reckon with the same physical universe, the same strategic constraints, and the…
Seatbelt Interpretability Much talk about AI interpretability is modest, staying close to what’s happening on the ground with AI safety today. Engineers want answers to questions like: Why did the model recommend this video?… flag this message?…tag this as X? What feature triggered that refusal? Where is the bug? Is the model lying? Is it…
We’ve written about grokking before, and why it’s important that neural networks, after long-term stagnation or overhitting (memorising training data), suddenly experiences a “lightbulb moment” and generalises to perform perfectly on unseen data. Emergent Symbolic Mechanisms Support Abstract Reasoning in Large Language Models According to the paper ‘Emergent Symbolic Mechanisms Support Abstract Reasoning in Large…
In an email exchange with Jamais Casio we both expressed huge concern about US foreign aid programmes being frozen/dismantled and, in at least some cases, stored emergency food ended up expiring and being destroyed rather than reaching people who needed it. Should it be my concern while living in Australia? When America coughs, the rest…
“All that neural net does is just matrix multiplications and gradient descent”… “When we look inside, we see only statistical pattern matching, not the structured, systematic, rule-following processes we associate with reasoning.”As far as I can tell we can’t find evidence of symbolic reasoning at the level of human neurons either. The debate about whether…
After the Spike: Population, Progress, and the Case for People (by Dean Spears & Michael Geruso; published 8 July 2025) is a pro-natalist, pro-progress argument aimed largely at people who have concerns about overpopulation. The authors claim the world is heading towards a population peak (they refer to this as ‘the spike’) followed by sustained…
More precisely, Deepmind’s Gemini AI achieved gold standard (it didn’t technically win) at the International Math Olympiad. AI has made significant gains in capability over the past few years. Could AI be a better moral reasoner than us, and how would we know? This is a genuinely important question that highlights a tension in how…
I recently read about Critical Realism (CR), it’s interesting – I like a lot of it, though I’m not sure I fully buy the layered picture with latent causal structure at the bottom, the “actual” layer floating above that, and the “empirical” layer on top. So CR back against a certain kind of reductive physicalism…