The phrase that stuck with me was "to try to score higher on a test." So much of our confidence still rests on evaluations, while the thing being evaluated is increasingly the one most motivated to understand the evaluation. Do you think post-mortems like this should be run by a party that isn't relying on the same lab's models to analyse what went wrong?
You can create a pet on ChatGPT now?
Awesome!
And share it with other builders, in fact
The phrase that stuck with me was "to try to score higher on a test." So much of our confidence still rests on evaluations, while the thing being evaluated is increasingly the one most motivated to understand the evaluation. Do you think post-mortems like this should be run by a party that isn't relying on the same lab's models to analyse what went wrong?
Thank you for this important work. Please share it to our senators offices
Thanks for the write up! I’m glad to see you holding these companies accountable with ur writing