Very interesting research. Reminds me of the trade off studied by the Johnson O’Connor Research Foundation between structural visualization and abstract reasoning. Do some people see the car because they think structurally, and others see the strawberry because they think abstractly?

We don’t know yet which applications are going to be broadly useful. I think it’s still possible that smaller, cheaper LLMs will end up being the most useful. Both the business cases and the bro-talk are based on magical and wishful thinking right now.

Most also don’t get to pick where they live. They go to the team that drafted them.

Anytime a meaningful counterfactual is utilized is a step forward. There are better and worse ones, but the contrast illustrates something valuable.

Congrats on the pub in Science. It’s a big honkin’ deal. You’re going to get a lot of attention, and that’s mostly good.

Agree that rational goal seeking is as much used as habit. Another form of choice behavior is reflex. The autonomic nervous system doesn’t give us time to rationally goal seek or engage habitual decision making but it does make quick decisions.

To be absolutely clear, it is impossible to vote for a party that cannot bring itself to unequivocally defend the rights of people living and working legally in this country. Because arbitrary removal of legal rights, or sudden varying of terms, for some is ultimately a threat to everyone.

Just like the .com bubble, there is some real value in generative ai, but the unsubstantiated hype is far larger. Signal to noise, and noise is currently winning.

A smart thing would be to work together in secret against this type of threat. Some CEOs may not consider it a threat, but fascists want power and corporations have it. It is likely that the administration will look to seize corporate power like they have with academic institutions.