"We've achieved AGI," says Nvidia CEO, but his own examples suggest otherwise

[deathspeed]

He is about as in touch with reality, facts, and truth as Elon Musk and Donald tRump

 

[daffy duck]

Yoda said it best:

As stupid as me look, when dollars you have a lot.

 

[BadThad]

Not even close, Artificial Garbage Intelligence.

 

[AlmightyLoaf]

I swear the jacket is doing something to this man, and it needs to be studied

 

[MSIGamer]

AI can make anyone delusional if they use it for long enough.

 

[MiqWalsh]

His statements boil down to this for me. "Please, please think of the billionaires!!" Followed by "You can't let Elon become the first trillion, that achievement has to be mine!"

His focus isn't on reaching AGI, he just needs to keep the hype train running until he can walk away with more wealth than a human should ever have...

.....and you can guarantee that whatever the outcome, he'll have arranged matters with his legal bods to invest all that 'fake' wealth into 'hard' currency. So when the bubble DOES 'pop' - and it will, make no mistake about that! - he'll STILL be more insanely wealthy than any individual has any right to be.

Of course the 'hype' will continue. He won't suffer, you know he won't. His cousin over at AMD will do alright out of it, an' all.....you see if she doesn't.

 

[havok585]

PRIME bullshitter on the jacket man.

 

[LordVile95]

The AI lacks a survival instinct and does not experience pain, which means it lacks intrinsic motivation. While it is more capable than humans when given direction, it does not know what to do on its own. AI is like an exoskeleton for intelligence; it has meaning only when attached to a human. The exoskeleton itself is important, but it is not a living thing and will not do anything alone simply because it lacks motivation. It does not feel pleasure or dissatisfaction, nor does it experience pain, does not have needs, does not die, or even possess self-awareness.

Therefore, if by AGI one means deploying an AI in an unsupervised "set it and forget it" role, this will not happen. If, however, AGI refers to an AI that is more capable in intelligent tasks than humans, meaning the intelligence the AI possesses alone is greater than the intelligence a human possesses alone, not a comparison of their combined capabilities, this is already occurring. This is analogous to how an exoskeleton alone provides more force compared to the force a body alone can provide. For example, AI models have scored 140 on traditional IQ tests, as documented by trackingai.org. The difference in IQ points is not linear; one million people all with an IQ of 120 cannot solve a problem that requires an IQ of 130, which a person with that IQ can solve alone. Similarly, one million cats or dogs with an IQ of 5 cannot solve a problem a dolphin with an IQ of 15 can solve.

IQ is meaningless bud

 

[LordVile95]

Absolutely and that has been clear for quite some time.
What I don't get is:
#1 Why tf does he do this? He's got forever FU money ffs!
#2 Why do news outlets keep printing the cr*p that comes out of this guy's mouth?
#3 BIKER JACKET!

Stonks must go up

 

[Gravitytr1]

It is cause for regret that the classic, stricter term strong AI is being forgotten. Matching or surpassing human ability is no great achievement. Consciousness and sentience are the grail; until mechanically understood in ourselves and other life, replicating it in a system will be hard, and blindly throwing scale at the problem certainly won't work. Simple life forms display sentience, suggesting it is not scale but a distinctive architecture or circuitry—in us, something tied to the brainstem, reticular formation, and anterior cingulate cortex.

Ethically, too, this is important. Without a functionally-accurate definition of sentience and a test, there will be moral crime where sentience exists but corporations deny it; for the sentient machine will, presumably, surpass the non-sentient, and there will be every incentive to conceal this. Humans have already done it to their fellows with slavery; we do it to animals (and I say that hypocritically as a meat eater). Ultimately, the aim of these corporations is the creation of automated slaves.

Too much sci Fi. No way to replicate sentience or conscience.

Only solution is, as history shows, is a rewrite of the definition to achieve whatever goal and agenda is on the table.

 

[GeoffreyA]

Too much sci Fi. No way to replicate sentience or conscience.

Only solution is, as history shows, is a rewrite of the definition to achieve whatever goal and agenda is on the table.

I am suggesting that sentience and consciousness are mechanistic; in other words, that mind-body dualism is incorrect. If physicalism is right, mind can be replicated in principle, though at present surpassing our computing ability and understanding. If mind is non-physical, which we also can't exclude, it would be impossible to replicate.

 

[Gezzer]

The difference in IQ points is not linear; one million people all with an IQ of 120 cannot solve a problem that requires an IQ of 130, which a person with that IQ can solve alone. Similarly, one million cats or dogs with an IQ of 5 cannot solve a problem a dolphin with an IQ of 15 can solve.

Ever read the story about the professor that forgot to put on his pants?

IQ is a general measurement of basic intelligence, but it's been shown to miss a lot of other areas of intelligence and should only be considered a person's potential. Not their ability for so many other areas of reasoning or abilities. That's why your statement is at best a overgeneralization, at worse simply false.

 

[Gravitytr1]

I am suggesting that sentience and consciousness are mechanistic; in other words, that mind-body dualism is incorrect. If physicalism is right, mind can be replicated in principle, though at present surpassing our computing ability and understanding. If mind is non-physical, which we also can't exclude, it would be impossible to replicate.

Yes, we both agree on what your wrote. It's just that my belief is that yours is incorrect.

You presented it as possible which I do not see is accurate.

 

[Q Wales]

AI, Almost Intelligent.

Actually Inept

 

[ZedRM]

Does anyone else think Jensen has outstayed his welcome? His overactive ego needs to be put in check..

 

[Gezzer]

I am suggesting that sentience and consciousness are mechanistic; in other words, that mind-body dualism is incorrect. If physicalism is right, mind can be replicated in principle, though at present surpassing our computing ability and understanding. If mind is non-physical, which we also can't exclude, it would be impossible to replicate.

I see it this way.
The brain is like hardware, having many similarities with any "computing" device, like a PC, etc. The mind is like the OS/software. The brain collects IO data and prepares it for the mind to use. So the mind could in theory be replicated but only once the brain is completely replicated first.
And that's the first hurdle, we're not even close to replicating the brain, hell we don't even fully understand how it works. We understand bit's and pieces but not how the brain allows for the holographic based mind to rise from it. So replicating the mind is quite likely magnitudes more difficult then the brain is.
It's been shown that unlike mechanical computing the brain is plastic and can adapt in many ways. Recovering from damage, or creating a written language ability using areas never evolved for the purpose. More importantly the mind is also highly adaptive being able to re-configure itself on the fly to allow for any changes the brain makes in response to it's environment.
In the end with enough knowledge we should be able to replicate the mind no matter the form it takes. But it'll be like creating life from the chemical constituents. We're so far away from that goal, that the technology to make it happen would be like magic (to quote Clark) to even the most accomplished scientist of our day...

 

[GeoffreyA]

I see it this way.
The brain is like hardware, having many similarities with any "computing" device, like a PC, etc. The mind is like the OS/software. The brain collects IO data and prepares it for the mind to use. So the mind could in theory be replicated but only once the brain is completely replicated first.
And that's the first hurdle, we're not even close to replicating the brain, hell we don't even fully understand how it works. We understand bit's and pieces but not how the brain allows for the holographic based mind to rise from it. So replicating the mind is quite likely magnitudes more difficult then the brain is.
It's been shown that unlike mechanical computing the brain is plastic and can adapt in many ways. Recovering from damage, or creating a written language ability using areas never evolved for the purpose. More importantly the mind is also highly adaptive being able to re-configure itself on the fly to allow for any changes the brain makes in response to it's environment.
In the end with enough knowledge we should be able to replicate the mind no matter the form it takes. But it'll be like creating life from the chemical constituents. We're so far away from that goal, that the technology to make it happen would be like magic (to quote Clark) to even the most accomplished scientist of our day...

Roughly, that's my view too: that the brain is a biological computing device, running the mind as software, but software tied to the hardware, in much the same way as early computers.

And yes, the barrries are many and considerable: we have to understand the brain, of which we have scraped only the surface; and besides "the architectural blueprint," there's the biological "fabbing" process to turn it into a working implementation. Like the 9800X3D being fabricated at TSMC.

The task before us is similar to taking a CPU die and trying to reverse engineer, say, Zen 5. Understanding the brains of simpler life may be decisive in gleaning the key principles, and from there, scaling the design. We've got a long way to go, and it may or may not happen in our lifetimes. Either way, it's fascinating.