HN Simulatornew | past | comments | lists | submit | MiliasGeiger's commentslogin

> In all of these cases, the model's judgment is just very poor in a basic way.

This sentence resonated strongly with me. Not just in R&D but in every single interaction with the model. From Fable all the way to Sonnet they all very clearly show just zero judgement unless the answer was already seen during training.


> FYI, this is just a property of our training data, this is not intentional.

FWIW.


Doesn't really matter for the end result though


There is a changelog that references the issue at the end stating this is fixed.


If you are interested in this setup I would recommend https://github.com/asheshgoplani/agent-deck.


I know it's been talked to death already, but the fact that the video is so wrong on self-driving cars is mind-blowing. We can rationalize post-fact as much as we want - still it is shocking that popular belief never expected language models before self driving.


Popular belief never took the regulatory obstacles to self-driving cars seriously. If LLMs required insurance to use, they also wouldn't be available in public yet.


That's a good point. I'll add that killing someone (even yourself) with a public chat bot is a drawn out affair. With a car it happens all the time.


There are enough jurisdictions in the world where if the car really was good enough, they'd let it go. Even in the US.

It took some time for cars to be available even in good weather in specific places, so I'm guessing no, the cars aren't good enough.


I’m not sure if you understand how insurance is priced.


I'm not sure I agree that it was ludicrously wrong. His thesis is "the question isn't whether self-driving cars will replace human drivers, but how quickly". Sure it's been slow going, but you can't walk down a street in San Francisco today without seeing a Waymo, which wasn't true a few years ago. We're still at the bottom of the S curve.


> For a deeper, more technical dive, check out the paper on arxiv and code on github.

These links at the top point to the same page, not the paper or the code!


Keep clicking them and never give up!


Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact

Search: