News for hackers. Stuff that matters.
Re:Scaling will never get us to AGI by galangalalgol on Thursday April 11, 2024 @08:56AM (#40001592)
Driverless cars are here, without liability issues they would already be widespread. But scaling isn't what got them there. It was tons of labeled data, both from rules based implementations and shadowing human drivers.

The ability to generalize completely outside training data isn't that common among humans IME. That is a high bar. How many of us have done so without at least drawing an analogy to some other experience we have had? Truly unique thinkers aren't that commonplace. I myself could have probably just asked an LLM what I would say in this post and gotten pretty close...

Re:Scaling will never get us to AGI by parpfish on Thursday April 11, 2024 @09:19AM (#40001804)
> But scaling isn't what got them there. It was tons of labeled data, both from rules based implementations and shadowing human drivers.

Isn't the second half of that saying that scaling up the dataset was what got them where they are?

When I hear about "scaling up" models, I think there are two parts of it: 1) use the same architecture, but bigger 2) make the training data bigger to make use of all those new parameters.

So when I hear about something that's not scaling, it would require some sort of fundamental change to architecture or algorithms.

Re:Scaling will never get us to AGI by staticman2 on Thursday April 11, 2024 @10:17AM (#40002414)
> Driverless cars are here, without liability issues they would already be widespread.

In the sense that if you were allowed to run over anyone without legal consequences lots of tech enthusiasts would sign up for this experience?

Re:Scaling will never get us to AGI by rchaud on Thursday April 11, 2024 @09:12AM (#40001718)
> Driverless cars are here, without liability issues they would already be widespread

In other words, they're in a lab somewhere undergoing perpetual test runs so they don't kill people, so not generally available.

It is like saying that email was available in the '80s, you just couldn't message anybody outside of the university network.

Re:Scaling will never get us to AGI by naasking on Thursday April 11, 2024 @09:20AM (#40001827)
You can order driverless taxis right now in some cities:

https://www.axios.com/2023/08/29/cities-testing-self-driving...

Re:Scaling will never get us to AGI by loire280 on Thursday April 11, 2024 @09:46AM (#40002093)
However, they require constant human monitoring and interventions from those humans every few miles to work properly - and it's not clear that the need for those interventions will be solved with more training data.

https://www.theverge.com/23948708/cruise-robotaxi-suspension...

Re:Scaling will never get us to AGI by tivert on Thursday April 11, 2024 @11:10AM (#40003035)
> However, they require constant human monitoring and interventions from those humans every few miles to work properly

It's a little-known fact that AI means "Actually, Indians."

https://boingboing.net/2024/04/03/amazons-ai-powered-just-wa...

Re:Scaling will never get us to AGI by MatthiasPortzel on Thursday April 11, 2024 @09:43AM (#40002064)
That article is discussing Cruise. Since it was published, a Cruise car dragged a pedestrian 20 ft, it came out that they need remote intervention every few miles, the CEO resigned, and they stopped doing rides anywhere.

Waymo remains available to my knowledge.

Re:Scaling will never get us to AGI by minitoar on Thursday April 11, 2024 @09:23AM (#40001854)
No, I think he meant they are literally here. I see one everyday.