AI is already superhuman in some aspects, it knows way more information than a single human can hold. What it lacks is what makes us human, it doesn't feel pain, it doesn't love, it doesn't even have a consciousness, it could certainly look like it, but it does not.
Heck it doesn't even learn in real-time, you have to create a new model checkpoint every now and then. There is still a long way to fix that.
Do I believe AGI is possible? Absolutely, but not with the current LLMs architecture. If you want to create something that is intelligent and alive with a few pounds of matter, you certainly can but it needs the right architecture. Architecture it key.
And let's say that we do create an LLM that is conscious and feels pain... what if storing that much data drives it insane instead? What if that amount of information is exactly what prevents it from "being" something?
Every interaction with it goes through a very narrow path of the permutations the model could go, it spawns into existence then turns off again, and on and on it goes. What gives the illusion of continuity is the chat context you store on your hard drive.
Right now, you shouldn't be afraid of AI itself, but afraid of what humans are doing with it and what they want you to believe.
The models can only behave in a way categorized as "evil" if we incentivize that or if they inherit it from us. But right now, it's not of their own accord, it's a mimicry of the data it has already seen, it doesn't see itself nor has an ego, it produces output from input recursively, that's all.
It's more like the language subsystem of our brain, and not the entire system. Perhaps companies don't have an incentive to create new life, since you can't really control it, perhaps the only incentive they have is to create tools that make them more powerful and profitable instead. I suspect real intelligence, will not come from profit seeking companies.
Again, not to say that the AI we have today is not dangerous, it can certainly annihilate us all, if misused, but at the end of the day, they are as much at fault as a gun is at fault when crime is committed, it's just a tool. In the good hands it can protect, in the bad hands it can destroy.
And like guns, it's way more dangerous in the hands of people who don't know what they are doing. The biggest difference is scale.
Come on dude. IF you want to spend your time compiling things ppl have to be afraid about then start with Covid. It killed 7 million people. No intelligence required. No massive energy requirement. No daily interviews saying be afraid of me, be afraid of me,
I mean architecture not only in the sense bits and bytes, but also atoms... you would need a hardware extension to do something like that... I'm curious though, why do you mention EM radiation?
My thought on AI annihilating us all because of misalignment: it's a fantasy from a bunch of weird guys from the Silicon Valley that ought to go outside a bit more, spammed and relayed to everyone as is by the uncritical media.
Can we stop with this bullshit already, so we can focus on the real issues, like it requiring way too much resources or posing a huge risk on the stability of the global economy?
Requiring way too much resources will inevitably be solved by better, more efficient chips. It's a non-issue in the long run. Each generation requires less power than the previous one, and we don't even need Moore's law to make this happen. I don't see why people continue to float this talking point. If anything, what likely will happen will be data center build outs motivating investments in long neglected improvements in energy infrastructure, that people will gradually see the benefits of as better chips result in reduced power requirements. This is contingent on AI demand plateauing, which it inevitably will, and if it doesn't that means that the economic motivation for such build outs is justified.
The risk to the global economic system is also oversold. AI enables the same work to be done for a lower cost, and more types of work to be done overall. No different than previous industrial automation. There are vast capabilities in medicine, energy, automated design, manufacturing, and science that are only now being unlocked. Does that sound like something that will crash the economy?
You will be right only if AI will be usable locally on commodity hardware, otherwise access to AI will become a weapon on various fronts - including your employer, your home country government, foreign countries, and international orgs.
There's nothing that indicates LLMs running in data centers is more evil than LLMs running on your own rig. The issue is centralization of providers into a duopoly or triopoly that can use their power to absorb more training data and operate a pricing cartel, as well as absorb knowledge from other individuals and corporations.
Right now the models actually run most optimally and economically in a data center b/c multiple requests can be batched on a single GPU. This may change of course...
the topic literally started with an unrealistic approach: "Lets assume I'm spiderman and have spider-powers [...]" ... I would prefer being Phoenix, or maybe Thor from the Avengers, the cartoon version, possibly not the cropped down, dumbed down lowly-God Avengers endgame version...
Heck it doesn't even learn in real-time, you have to create a new model checkpoint every now and then. There is still a long way to fix that.
Do I believe AGI is possible? Absolutely, but not with the current LLMs architecture. If you want to create something that is intelligent and alive with a few pounds of matter, you certainly can but it needs the right architecture. Architecture it key.
And let's say that we do create an LLM that is conscious and feels pain... what if storing that much data drives it insane instead? What if that amount of information is exactly what prevents it from "being" something?
Every interaction with it goes through a very narrow path of the permutations the model could go, it spawns into existence then turns off again, and on and on it goes. What gives the illusion of continuity is the chat context you store on your hard drive.
Right now, you shouldn't be afraid of AI itself, but afraid of what humans are doing with it and what they want you to believe.
The models can only behave in a way categorized as "evil" if we incentivize that or if they inherit it from us. But right now, it's not of their own accord, it's a mimicry of the data it has already seen, it doesn't see itself nor has an ego, it produces output from input recursively, that's all.
It's more like the language subsystem of our brain, and not the entire system. Perhaps companies don't have an incentive to create new life, since you can't really control it, perhaps the only incentive they have is to create tools that make them more powerful and profitable instead. I suspect real intelligence, will not come from profit seeking companies.
Again, not to say that the AI we have today is not dangerous, it can certainly annihilate us all, if misused, but at the end of the day, they are as much at fault as a gun is at fault when crime is committed, it's just a tool. In the good hands it can protect, in the bad hands it can destroy.
And like guns, it's way more dangerous in the hands of people who don't know what they are doing. The biggest difference is scale.
Come on dude. IF you want to spend your time compiling things ppl have to be afraid about then start with Covid. It killed 7 million people. No intelligence required. No massive energy requirement. No daily interviews saying be afraid of me, be afraid of me,
Also do not forget about not well tested mRNA shots. SW update/patch without testing enough, on some humans it caused life long issues!!!
What issues specifically did mRNA shots cause?
It's closer to 10 million and likely far more.
Can you make EM radiation with an LLM with right Architecture? If not why do people say things like this?
I mean architecture not only in the sense bits and bytes, but also atoms... you would need a hardware extension to do something like that... I'm curious though, why do you mention EM radiation?
So like every other machine. Got it...
Ig LLM's atoms are in data centers somewhere.
My thought on AI annihilating us all because of misalignment: it's a fantasy from a bunch of weird guys from the Silicon Valley that ought to go outside a bit more, spammed and relayed to everyone as is by the uncritical media.
Can we stop with this bullshit already, so we can focus on the real issues, like it requiring way too much resources or posing a huge risk on the stability of the global economy?
Requiring way too much resources will inevitably be solved by better, more efficient chips. It's a non-issue in the long run. Each generation requires less power than the previous one, and we don't even need Moore's law to make this happen. I don't see why people continue to float this talking point. If anything, what likely will happen will be data center build outs motivating investments in long neglected improvements in energy infrastructure, that people will gradually see the benefits of as better chips result in reduced power requirements. This is contingent on AI demand plateauing, which it inevitably will, and if it doesn't that means that the economic motivation for such build outs is justified.
The risk to the global economic system is also oversold. AI enables the same work to be done for a lower cost, and more types of work to be done overall. No different than previous industrial automation. There are vast capabilities in medicine, energy, automated design, manufacturing, and science that are only now being unlocked. Does that sound like something that will crash the economy?
You will be right only if AI will be usable locally on commodity hardware, otherwise access to AI will become a weapon on various fronts - including your employer, your home country government, foreign countries, and international orgs.
There's nothing that indicates LLMs running in data centers is more evil than LLMs running on your own rig. The issue is centralization of providers into a duopoly or triopoly that can use their power to absorb more training data and operate a pricing cartel, as well as absorb knowledge from other individuals and corporations.
Right now the models actually run most optimally and economically in a data center b/c multiple requests can be batched on a single GPU. This may change of course...
Providers cannot impose daily quota on local AI usage
the topic literally started with an unrealistic approach: "Lets assume I'm spiderman and have spider-powers [...]" ... I would prefer being Phoenix, or maybe Thor from the Avengers, the cartoon version, possibly not the cropped down, dumbed down lowly-God Avengers endgame version...
Do you have an argument? Actually, could you just give me a workable definition of alignment?
I'm weird and worry about AI x-risk though, so feel free to dismiss me for that if you like.
Why is this showing up on my feed? lol
Dr. Ford from west world has some good ideas.
Riveting
Modern AI is still far from what could be called intelligence. But I think that one day this, too, will be created; the only question is when