Those are fair points, but I think you're massively underestimating how dangerous human drivers are
I spend most of my time in a city where there are no Waymos and I encounter situations almost weekly where human drivers take dangerous actions without realizing the peril they're creating
Let's also consider trend lines: we have every reason to believe that computers will be safer drivers 1/5/10 years from now. Can't say the same for human drivers.
@Beylin There's pretty extensive data that Waymos are far safer than the average human driver per mile driven
The NYTimes isn't exactly tech-forward, but even they are publishing op-eds making this case https://t.co/YXE8P8QiOh
@DicksonWuML Generalized frontier is going to be good in those areas, so unless they figure out how to prevent jailbreaks entirely, I’m not sure how they can do that segmentation
Unfortunately, with access to an abliterated DeepSeek V4.1 model, there is little value to KYC holding back the frontier models now for cyber.
In fact, making it hard to secure systems w extra friction in American Enterprise is hurtful to national security.
@mosayeri Oh 100% definitely not a model problem
Just saying that if they somehow leapfrog to a crazy level of RSI, that could be enough to fix their myriad of product issues
@mosayeri Yeah, I'm not completely counting Google out from figuring out a way to leapfrog with an insane frontier model, but this is the first time I've ever been long-term bearish the company