Why did we not see ChatGPT coming and why are we still baffled by the advancement of this technology? Read link.springer.com/article/10…

Solving alignment could be even worse than not solving it, because it would mean an ASI could end up loyal to very bad humans. But I’m not too worried about it.

I bet we’ll also have the announcement of a breakthrough in medical research before the midterms.

Interpretability is a dead end. The cause of an AI’s behavior is not in the weights. It is in the training history.

The real shock of AI may not be what machines can do. It’s the possibility that intelligence was simpler than we believed.

We built AI before understanding it. Maybe because we looked at it through the wrong psychology. The framework that fits is not cognitivism. It’s behaviorism, the school declared obsolete 60 years ago.

Claude

Gemini

DeepSeek

It could be the only credible lever in a negotiation with an ASI.

Hi @pkghosh99, jokes aside, do you see any methodological issues in these papers that would undermine their conclusions? pnas.org/doi/10.1073/pnas.24… openaccess.thecvf.com/conten…

I thought having access to GPT-4 was so crazy that it would become illegal. How come we have free access to a monster like Gemini 3?!

Meaning will always relate to your hierarchical status among your conspecifics and your role in caring for and protecting those you love. AI could make the game harder or easier, but won't fundamentally change it.

About the AI bubble...

While cognitivism dominates universities and has permeated society, it was the behaviorist approach that, a century ago, theorized that complex behaviors emerge from simple associative processes. This perspective is essential for understanding today’s AI breakthroughs.

Associative learning (reinforcement learning) plays a central role in recent AI advances; paradoxically, this field is neglected in contemporary psychological research.