Leo, You are Living Under a Rock When It Comes to AI

[Archie]

@Leo Gura Given your understanding of what intelligence is, do you have any insights on how to create AGI? or models that are better than our current one? Do you think AGI is even possible?

[OBEler]

34 minutes ago, Leo Gura said:

No. AGI is not a vague term. And no, we do not have anything close to AGI.

AGI is human level intelligence. No LLM is capable of that.

AGI is able to understand things outside of its training data. LLMs cannot do that. Don't even try to act like they can.

We dont need human level intelligence. We need intelligence that is capable enough to do every task a human can do in a working environment. And AI is very capable of that. You should work in a company for a while, you will see how stupid mundane tasks are.

[Something Funny]

3 minutes ago, OBEler said:

We dont need human level intelligence. We need intelligence that is capable enough to do every task a human can do in a working environment.

But a post before that you've said "infastructure is not ready".

What's the difference between infrustructure and environment?

In my understanding, they are the same. So to me it sounds like you are contradicting yourself or saying that we need to adapt the working environment for AI just so it could be useful, when humans don't require that. They can adapt to the working environment instead. I guess that's the main difference in the capabilities. 

Hire a worker, often for low wage vs take a giant risk and spend millions overhauling your whole company infrustructure for AI that might or might not work and then be a slave to the AI company, paying it for every token you use.

[Something Funny]

Just now, Something Funny said:

But a post before that you've said "infastructure is not ready".

What's the difference between infrustructure and environment?

In my understanding, they are the same. So to me it sounds like you are contradicting yourself or saying that we need to adapt the working environment for AI just so it could be useful, when humans don't require that. They can adapt to the working environment instead. I guess that's the main difference in the capabilities. 

Hire a worker, often for low wage vs take a giant risk and spend millions overhauling your whole company infrustructure for AI that might or might not work and then be a slave to the AI company, paying it for every token you use.

@OBEler I think this is also what Leo means by AGI. 

AI that's capable of being put in a random job, reading working instructions file, and then figuring out how to do it with the existing tools, whatever they are

[something_else]

19 minutes ago, Something Funny said:

Hire a worker, often for low wage vs take a giant risk and spend millions overhauling your whole company infrustructure for AI that might or might not work and then be a slave to the AI company, paying it for every token you use.

It's more like "hire 2 senior devs who each have a team of 3 junior/mid-level devs working with them" versus "hire 2 senior devs who each get a 100-200$/m AI subscription that you can set up for them in 15 minutes"

Both of these probably have equivalent output just now, but 100-200$/m is much cheaper than 3 junior/mid-level devs.

Your final point is somewhat valid, you are kind of held hostage by the AI company. But that happens all of the time in business. Old developers were held hostage by Google or a JetBrains license. It's just a cost the company has to pay for a tool so that you can do your job at a competitive level.

Edited 23 hours ago by something_else

[Something Funny]

@something_else you are talking about software development. Most jobs don't involve you asking a chatbot to write a code for you.

Sure, AI is great for software development because this is literally an ideal task for it.

 

[Something Funny]

@something_else also, I don't think they are offering subscription for enterprise clients. It is usage based, which is much more expensive.

[Butters]

3 hours ago, Leo Gura said:

You guys are underestimating how profound intelligence is.

So what's your EXACT benchmark for intelligent AI? 

[something_else]

1 hour ago, Something Funny said:

@something_else you are talking about software development. Most jobs don't involve you asking a chatbot to write a code for you.

Sure, AI is great for software development because this is literally an ideal task for it.

 

A significant chunk of many white collar jobs can be done within the domain of language + software tool usage, which is what LLM models are good at.

Software is where it’s most impressive just now, and also where it’s been applied most readily because tech bros can adapt to this stuff quickly, but it’s certainly far from the only field it’s capable in

Quote

also, I don't think they are offering subscription for enterprise clients. It is usage based, which is much more expensive.

Up to 150 seats you can get the team plan which is $20/m for standard seats and $110/m for premium seats per dev.

Beyond 150 people you need enterprise where it’s metered, but 150 people covers the vast majority of companies.

Even when it’s metered, I suspect most devs would rarely go above $500/m in usage tokens, which is more than enough to allow them to be 4-5x as productive. It’s still very worth it for companies

[OBEler]

2 hours ago, Something Funny said:

But a post before that you've said "infastructure is not ready".

What's the difference between infrustructure and environment?

In my understanding, they are the same. So to me it sounds like you are contradicting yourself or saying that we need to adapt the working environment for AI just so it could be useful, when humans don't require that. They can adapt to the working environment instead. I guess that's the main difference in the capabilities. 

Hire a worker, often for low wage vs take a giant risk and spend millions overhauling your whole company infrustructure for AI that might or might not work and then be a slave to the AI company, paying it for every token you use.

With infrastructure I mean the operating system on the computer. Humans rn have more rights and accessibility . An i is not allowed to download and install software, to login into certain programs, to create orders on your computer etc.  It is totally capable of that but the infrastructure is not there yet. 

[Uddi]

Intelligence is from semantics, not syntax. AI is syntax only and since syntax has no intrinsic value, it therefore can never get true intelligence.

[Hojo]

Intelligence is like you fully programmed an AI deleted the AI, fully programmed it then deleted the AI, fully programmed it deleted it, then the Ai starts programming itself out of nothing. And you ask where is it coming from? Its programming itself from another dimension.

Life and you are so intelligent you randomly broke through to another dimension as if cheating death. If the AI was intelligent you could delete it and it could come back. Then once it programmed itself you could ask it how it did it and it would say I dont know.

Edited 20 hours ago by Hojo

[thenondualtankie]

Modern AI can program itself with literally zero training data. Because it can just keep sampling from the reality it's learning from, and keep getting a better picture of that reality.

That's how the AlphaZero engine was trained. LLMs use the exact same mechanism, but they can learn realities that are not so limited as Go.

[AerisVahnEphelia]

6 hours ago, Hafiz said:

LLM’s are a grift 

As much as google search was a grift.

"Data" is underestimated & overestimated at the same time. 

[something_else]

3 hours ago, Uddi said:

Intelligence is from semantics, not syntax. AI is syntax only and since syntax has no intrinsic value, it therefore can never get true intelligence.

This is just wrong. LLMs capture semantics extremely well or they would not be able to write a coherent sentence

 

[Elliott]

Ai data is only what humans have digitized. You know how they say a picture is worth a thousand words, AI has not tasted, smelled, or felt.

[OBEler]

5 hours ago, Uddi said:

Intelligence is from semantics, not syntax. AI is syntax only and since syntax has no intrinsic value, it therefore can never get true intelligence.

Oh no... That's wrong. 

[Eskilon]

The question is, can true intelligence be mechanical? If it can't, it needs consciousness, so unless AI becomes conscious it cannot be truly intelligent.

A computer is not intelligent, even though it is doing insane ammounts of calculations no human can. The intelligence of the computer is coming from somewhere else(humans). This is the same thing happening with AI, it is dead, it's intelligence doesn't come from itself, it comes from somewhere else.

Edited 15 hours ago by Eskilon

[Jodistrict]

1 hour ago, something_else said:

This is just wrong. LLMs capture semantics extremely well or they would not be able to write a coherent sentence

 

Right.  The LLM builds an internal contextual model of the sentence it is interpreting.   Natural language processing based on a syntactic grammar would only be able to interpret toy sentences.

[Natasha Tori Maru]

Intelligence requires emotion.