Elon Musk and others urge AI pause, citing 'risks to society'

[redmunds]

I love my Tesla and I purchased FSD, so I'm not sure what to think about this:

Elon Musk and others urge AI pause, citing 'risks to society'

Elon Musk and a group of artificial intelligence experts and industry executives are calling for a six-month pause in developing systems more powerful than OpenAI's newly launched GPT-4, in an open letter citing potential risks to society.

www.reuters.com

 

[Midpack]

It's a nice idea, but it won't happen. Some entity(its) will cheat, put others at a disadvantage, 'the genie is already out of the bottle' like it or not.

 

[redmunds]

Yes, it's not easy to enforce.

But GPT and FSD are both machine learning, so it seems hypocritical for Elon to try to make a distinction between the two.

 

[diplomat33]

I find it kind of odd how Elon can seem both pro-AI and anti-AI at the same time. On one hand, he hypes up how Tesla is solving AI in autonomous driving and robotics but then he also says that we need to control AI and slow it down because it is going to destroy humanity.

 

[gtg465x]

Yes, it's not easy to enforce.

But GPT and FSD are both machine learning, so it seems hypocritical for Elon to try to make a distinction between the two.

It does seem hypocritical, but I kind of get his concern. AI that creates and disseminates information on the internet has much greater potential to influence our thoughts and fundamentally alter the direction of society than a car that (sort of) drives itself, in my opinion.

 

[nvx1977]

Yes, it's not easy to enforce.

But GPT and FSD are both machine learning, so it seems hypocritical for Elon to try to make a distinction between the two.

there is a distinction - the call is for a pause for the development of artificial general intelligence (AGI).

FSDb is very much still artificial narrow intelligence (ANI), and these are not the AIs that are going to hit the singularity and suddenly become super-human intelligent.

 

[davidski]

I think one can be both excited for the potential for AI to help humanity with all sorts of problems, and at the same time want us to proceed cautiously with this powerful technology. On the other hand, I have a strong feeling that a coordinated pause will be largely impossible.

 

[redmunds]

AI that creates and disseminates information on the internet has much greater potential to influence our thoughts and fundamentally alter the direction of society than a car that (sort of) drives itself, in my opinion.

AI creates information, but people post it on the internet. Same with FSD videos.

 

[EVRider-FL]

Elon has been voicing his concern about generalized AI for years, so it's not a new concern for him. I agree that it's completely different from using AI in FSD, so there's nothing hypocritical about it.

 

[glide]

I find it kind of odd how Elon can seem both pro-AI and anti-AI at the same time. On one hand, he hypes up how Tesla is solving AI in autonomous driving and robotics but then he also says that we need to control AI and slow it down because it is going to destroy humanity.

He only cares when it has the potential to impact one of his lines of business.

Just like everything else he does, this is ultimately about the bottom line.

 

[Krugerrand]

I find it kind of odd how Elon can seem both pro-AI and anti-AI at the same time. On one hand, he hypes up how Tesla is solving AI in autonomous driving and robotics but then he also says that we need to control AI and slow it down because it is going to destroy humanity.

It’s not odd to have an opposing view on the same topic; it’s called context. It’s like saying you’re for genetic engineering (when used to benefit society and saves lives, enhances life) vs against genetic engineering gone amok that is detrimental to society.

Elon simultaneously sees how AI can be used to help us AND destroy us. Not odd at all.

 

[voyager]

I was thinking the same thing: conflict of interests. Self-learning systems are supposed to bring Teslas to full autonomous driving.
At the same time, Musk says that in self-learning intelligence hides a considerable risk. Might it be that this way, a full functioning auto-pilot will never be realized?

This morning I heard a professor on the BBC comment that the risk is that an AI cannot reflect upon what's it doing the way humans can.

I find it kind of odd how Elon can seem both pro-AI and anti-AI at the same time. On one hand, he hypes up how Tesla is solving AI in autonomous driving and robotics but then he also says that we need to control AI and slow it down because it is going to destroy humanity.

 

[nvx1977]

sounds like many people in this thread need to do some basic research on AI before forming an opinion.

at a bare minimum, understand the difference between general AI and narrow AI, and how vastly different the approaches are, and the consequences.

 

[Mr. Michael]

It seems naive to take the word of a shill at face value.

 

[johnmodely]

sounds like many people in this thread need to do some basic research on AI before forming an opinion.

at a bare minimum, understand the difference between general AI and narrow AI, and how vastly different the approaches are, and the consequences.

There is no difference between ChatGPT AI and Tesla AI from a theoretical and data standpoint.

Both feed it enormous amount of real world data, and create neural networks which humans have limited knowledge on how they work in both cases.

One spits out a bunch of text based on the likelihood the next letter should be there, and the other does the same thing based on visual prompt data instead of text prompt.

One can argue that Tesla's AI, especially when decision neural nets go live, is a much more dangerous general AI than Chat GPT (and even musk has stated several times that vision and driving requires general AI to be solved), and especially when its making devisions in the real world with a couple ton vehicle it controls.

 

[HVM]

I put this here for info:

And paper: https://arxiv.org/ftp/arxiv/papers/2302/2302.02083.pdf

 

[HVM]

 

[nvx1977]

There is no difference between ChatGPT AI and Tesla AI from a theoretical and data standpoint.
Both feed it enormous amount of real world data, and create neural networks which humans have limited knowledge on how they work in both cases.

I think both are clearly in the narrow AI realm, for now.

One can argue that Tesla's AI, especially when decision neural nets go live, is a much more dangerous general AI than Chat GPT (and even musk has stated several times that vision and driving requires general AI to be solved), and especially when its making devisions in the real world with a couple ton vehicle it controls.

Here I would say the pen is mightier than the sword. Sure an AI controlling a big thing like cars can hurt people, but it is very limited in its damage through safeguards in its design. Natural language processing when abused (coupled with other narrow AI tech like deepfakes) can be used to wreak havoc via misinformation. I can easily imagine a different outcome of the Cuban missle crisis in a world of advanced language AI coupled with good deepfakes.

 

[zoomer0056]

All this because Elon said "AI stresses me out". Leading to conclusions fof more add revenue.

 

[nvx1977]