It is worth investing a little more on nuclear in the interest of energy security. It isn't always a race to the bottom.
A diverse and robust energy grid is composed of various generation sources that differ in capacity factors, environmental impact, stability, availability, etc.
The official openai-cookbook (https://github.com/openai/openai-cookbook) used to have an explicit, but buried, call out that instruction-following models like `text-davinci-003` were "Less diverse; less creative; sometimes harder to steer tone, style, etc." as opposed to base completion models like `davinci`.
It stood out to me because it seemed to be an internal admission that this training narrowed the potential of the models.
> Two-headed worms made this way reveal a permanent revision of the target morphology: subsequent rounds of regeneration in plain water, long after the reagent is gone from the tissue, continue to make two-headed worms.
Great work. It would be nice to see the interactions wrap to make a more seamless experience. The particles themselves do seem to wrap around the edges, but the forces do not.