The models are getting smarter, faster, cheaper, and safer with each iteration. The AI models you're using today are the worst you'll ever use, and the future will be very different from today.

The models currently have a lot of knowledge but lack the skill level to make significant discoveries, particularly in complex fields like biology.

The AI models you're using today are the worst you'll ever use for the rest of your life. Every two months, computers can do something they've never been able to do before, so you need to think differently about what you're doing.

There is ongoing debate in the AI community about whether we have reached the limits of scaling laws, as the latest models have become larger and more expensive, yet their capabilities seem to be plateauing.

We're in an interesting time where the capabilities of AI models are far ahead of the products people have built. There's a huge opportunity for new innovations, especially as the cost of AI continues to decrease dramatically.

The distinction between how models claim to perform tasks and how they actually do them can reveal significant insights into their capabilities.

The current advancements in AI suggest that we are not near a limit; the rate of change and new models indicates ongoing improvement.

We are no longer in a compute-constrained environment for our best models, marking a significant shift in our approach to model training and development.

The biggest barrier to increasing scientific progress with AI is that the models are just not smart enough yet.