4 ms·

>That doesn't mean these kinds of systems can't be better and even more useful in 1-2 years. It's also faulty to assume that improvements will happen linearly.

>That doesn't mean these kinds of systems can't be better and even more useful in 1-2 years. It's also faulty to assume that improvements will happen linearly. We had a self-driving car mania in 2013-2017 with billions poured into it and people just forgot about it. Also consider that if easy linear improvements were possible, they would've delayed launching it. AI (software in general?) seems to be the kind of thing where once R&D hits the exponential difficulty wall, they make it generally available.

sva_ 4y ago
> The first 90 percent of the code accounts for the first 90 percent of the development time. The remaining 10 percent of the code accounts for the other 90 percent of the development time. - Tom Cargill, Bell Labs
kenjackson 4y ago
Self driving cars have a failure downside that is very high for basically all use cases. On improvements— one of the things that has impressed is how much Chat GPT has improved since it was released. I don’t think it’s going to be a perfect AI in two years. But it already is working better for some use cases for me.
Hyption 4y ago
It feels slow to us but from past performance we still progress on this topic and have achieved already impressive results. And more and more puzzle peaces are falling in place. We can see were it is going. And when you follow Nvidia with their digital twin topic, we are converging.
CuriouslyC 4y ago
Not so much. Making it available allows OpenAI to evaluate and iterate on the chatbot layer, which has had much less data and engineering invested. You equate lack of ubiquitous self driving cars with lack of progress in self-driving technology, but even if we've had super-linear progress in terms of self-driving "driving skill," if that skill is below the threshold where the sales from pushing the software are less than the potential legal liability, companies aren't gonna push it.

sva_ 4y ago

> The first 90 percent of the code accounts for the first 90 percent of the development time. The remaining 10 percent of the code accounts for the other 90 percent of the development time. - Tom Cargill, Bell Labs

kenjackson 4y ago

Self driving cars have a failure downside that is very high for basically all use cases. On improvements— one of the things that has impressed is how much Chat GPT has improved since it was released. I don’t think it’s going to be a perfect AI in two years. But it already is working better for some use cases for me.

Hyption 4y ago

It feels slow to us but from past performance we still progress on this topic and have achieved already impressive results. And more and more puzzle peaces are falling in place. We can see were it is going. And when you follow Nvidia with their digital twin topic, we are converging.

CuriouslyC 4y ago

Not so much. Making it available allows OpenAI to evaluate and iterate on the chatbot layer, which has had much less data and engineering invested. You equate lack of ubiquitous self driving cars with lack of progress in self-driving technology, but even if we've had super-linear progress in terms of self-driving "driving skill," if that skill is below the threshold where the sales from pushing the software are less than the potential legal liability, companies aren't gonna push it.