If AI doubles in strength each year, then in 10 years it will be 1,000 times stronger than now

[WizardofSoda]

These are factors leading to the increasing ability of the AI's. 2023 was the dawn of AI and also in the last year is when the mega tech companies hugely increased spending on AI.

-the computer processors for AI are getting better exponentially
-the mega tech corporations are greatly scaling up the number of AI processors they are buying

-the AI scientists are figuring out more advanced algorithms to speed up the AI learning
-they are also coming up with creative ways for the AI's to get to higher wisdom and abilities

 

[Vendetta]

bros onto nothing

 

[theautismcel]

AI stacy foidbots when??

 

[Confidencel]

Wtf is AI “strength”? Stop watching animes schizo.

bros onto nothing

 

[PersonaPimp]

I just want an AI gf

 

[FrothySolutions]

-the computer processors for AI are getting better exponentially

Moore's Law no longer applies here. Processors are still getting better, but they're not doubling every year. Not anymore.

 

[WizardofSoda]

Moore's Law no longer applies here. Processors are still getting better, but they're not doubling every year. Not anymore.

So far the Nvidia AI processors seem to be doubling fast for the specific task of AI.

But I am not saying the computer processors have to double every year. I am saying the combination of the above factors together may double every year for the next 10 years.

 

[WizardofSoda]

Wtf is AI “strength”? Stop watching animes schizo.

That is like a PHD level thesis question, that is why I chose to leave it as just 'strength' and not get specific.

 

[Vale]

Do you think this will decrease inceldom? tbh i see it happening in 5 years

 

[NeverEvenBegan]

it's over for mooreslawwcels

 

[based_meme]

So far the Nvidia AI processors seem to be doubling fast for the specific task of AI.

But I am not saying the computer processors have to double every year. I am saying the combination of the above factors together may double every year for the next 10 years.

Can you be specific? AI isn't a task. AI is given tasks. @FrothySolutions is correct here. The hard ceilings in current technology cannot be solved by throwing money at it, and nobody can predict when - or more importantly, IF - the next breakthrough will happen to propel the research forward.

 

[EnviadoVerde]

I want holographic girlfriend

 

[WizardofSoda]

Can you be specific? AI isn't a task. AI is given tasks. @FrothySolutions is correct here. The hard ceilings in current technology cannot be solved by throwing money at it, and nobody can predict when - or more importantly, IF - the next breakthrough will happen to propel the research forward.

I'm going by Nvidia's graphs they release showing the increasing performance for AI tasks of each new release of their processors. I don't know what they mean specifically but the AI companies seem to agree with the estimates and be buying large amounts of the chips.

In the AI science literature there are papers flowing in all the time for new algorithms and techniques that advance things. If say ChatGPT was based on one big breakthrough algorithm, I would agree that we couldn't predict the next one. But when its a whole bunch of small advances that are coming constantly it seems probable they are going to keep coming. Especially as the mega tech corporations are now putting huge money into AI.

On benchmarks there was also major improvements between I think it was ChatGPT 3.5 and 4.0. And that wasn't even a year apart.

 

[WizardofSoda]

Well we will see where it goes. Most guys commenting I read online are predicting that AI has peaked out, that it won't make improvements past here. I am predicting monster improvements over the next 10 years.

 

[based_meme]

I am predicting monster improvements over the next 10 years.

How do you define "monster improvements"?

 

[Dusk]

How do you define "monster improvements"?

I'm curious too, share your research!

 

[based_meme]

I'm curious too, share your research!

A month later and still no definition.

 

[Dusk]

I'm interested in AI and I can anticipate major advancements too in a very short time period, but I'm just curious as to what kind of sources you have? Always love to read about the stuff.

 

[WizardofSoda]

I'm curious too, share your research!

A month later and still no definition.

Say 1,000 times improvement over AI today. What exactly that means in terms of what the AI's can do, I don't know. Its like take two people and its clear one of them has a vastly deeper intellect, more creativity, resourcefulness and so forth. But try defining that difference in objective terms. Or try guessing what the smart person will do in the next 10 years compared to the average person.

 

[Dusk]

Say 1,000 times improvement over AI today. What exactly that means in terms of what the AI's can do, I don't know. Its like take two people and its clear one of them has a vastly deeper intellect, more creativity, resourcefulness and so forth. But try defining that difference in objective terms. Or try guessing what the smart person will do in the next 10 years compared to the average person.

Yes, but where are you getting the prediction from? Is there somewhere that says AI will get 1000 times more powerful in only 10 years?

 

[WizardofSoda]

Yes, but where are you getting the prediction from? Is there somewhere that says AI will get 1000 times more powerful in only 10 years?

I got it from the history of computing from the 1970's to 2010 the gains per year were doubling every year on benchmarks. Not sure since 2010 as I haven't read into it since then for computing. Half the improvements from 1970 to 2010 were from the computer itself and half from algorithm improvements in the software.

And doubling every year for 10 years is 1,000 times improvement. (and 1 million times increase over 20 years)

 

[based_meme]

Say 1,000 times improvement over AI today. What exactly that means in terms of what the AI's can do, I don't know. Its like take two people and its clear one of them has a vastly deeper intellect, more creativity, resourcefulness and so forth. But try defining that difference in objective terms. Or try guessing what the smart person will do in the next 10 years compared to the average person.

Intelligence doesn't translate and scale like that. Not only that, but there are no intelligent machines as of today. There are specific programs with narrowly-defined goals and parameters, but 1000x that will just mean it does more of the same we have today - it won't do anything new or anything novel. It will continue to do the things it's programmed to do, only it will do things that don't make immediate sense to us. But the chain of logic inside of the "black box" won't be anything alien.