AI Acceleration - Page 2

On May 10 2023 18:19 Loco wrote: A.I. has no chance of replacing genuine human creativity. Its only real use is in improving corporate profits, surveillance, those kinds of things. It will never be "more intelligent than humans" because it is still determined by computing laws, which is not how human brains work. It's old outdated thinking that still has people believe that the mind is essentially a computer and all that matters is how much bandwidth it has and how many things it can compute at the same time. It's very naive and misinformed. The kind of thing that Elon talks about is a dystopian fantasy. Only people who are really disconnected from reality can entertain the thought that a computer is more dangerous than nukes and climate change. A computer cannot do shit and be "let free" on the world. It can't rebuild itself, it can't even plug itself into an outlet when it's out of power (or near out of power). It does not have that kind of intelligence, it constantly needs support from human beings for its very limited existence. A robot can be sent to autocharge but it doesn't know that it's going to charge itself, it cannot sense this in its environment unless it is programmed to interact with platforms that can charge it. In this dystopian world there would need to be a lot of nearly indestructible charging platforms everywhere for the robots to recharge while they wage war on us, the mental image of that is absolutely ludicrous. Part of me wants to give them credit and say "they know that they are full of shit and they're just distracting the populace from how much power they have for as long as they can" but that'd be granting them too much intelligence and self-awareness which they likely don't have.

You are a cornucopia of single minded bad takes and terrible predictions, I love it.

We like to pretend our creativity regarding art is endless and its process so mysteriouis yet the most popular songs are a few chords regurgitated by music producers that know what people like, easily mimicked by A.I. and thinking that only applies to the masses that a refined taste like yours can only be tickled by the highest quality of artistry well think again, it won't take that long for AI to write a masterful symphony that will delight our snobby ears.

That being said I dont think it will erradicate the arts, we will discriminate in favour to human made art for the same reason I stated above, we think too highly of ourselves.

On May 13 2023 10:00 Baalim wrote: We like to pretend our creativity regarding art is endless and its process so mysteriouis yet the most popular songs are a few chords regurgitated by music producers that know what people like, easily mimicked by A.I. and thinking that only applies to the masses that a refined taste like yours can only be tickled by the highest quality of artistry well think again, it won't take that long for AI to write a masterful symphony that will delight our snobby ears. That being said I dont think it will erradicate the arts, we will discriminate in favour to human made art for the same reason I stated above, we think too highly of ourselves.

It is self-evident that human beings have infinite creativity, though. That's not some sort of anthropocentric worldview. It's just an observation. Take something basic like natural language. Many sentences people use in daily life have never been used before. In terms of art, yes, human beings have constraints on what they can create, just like other animals do. But within those constraints, you still have infinite generative capacity.

It's possible that an unregulated internet will become fake enough that people won't have the privilege to discriminate between human music and robot music. That shouldn't be ruled out.

It's not that an AI can't create a creative product that will match my or other people's preferences perfectly well. I'm sure the more time that passes and the more practice and feedback they get, the better they'll be. It's already plenty good at that, anyway.

But at least to me, appreciating art is actually about appreciating the creative human genius behind it, not just about appreciating the product. I'm not worried about AI 'overtaking the world' or whatever, I'm worried about AI becoming better than humans at all the various fun human activities that are fun to engage in, to the point where attempting to become 'great' at them ends up feeling kinda meaningless because even if you spend three decades mastering something, some jerk can just ask a computer to make something better.

If a computer can do it better then it should, and it's evidence that it was meaningless to begin with. That's what computers do. They don't deal with meaning. They can recreate popular patterns that a lot of people find pleasing in music, sure, but they cannot create thematic art that is actually meaningful. It can't have depth. It can certainly be programmed to use a lot of poetic words, and even superficially seem like it's saying something that matters, but it never will create a piece of art that can really touch and change people. It's not in the code and can never be. With computers and humans we are dealing with differences of kind, not degrees. It doesn't matter how advanced it gets because it's not about the complicatedness of the code. The complicatedness of the code cannot compare in kind with the complexity of hundreds of millions of years of self-organizing, evolving biological life. We can express that complexity through sound because we have an inner need for it, a need that a computer will never be able to understand and therefore won't reproduce. No matter how advanced it gets it would never write anything that matters. But even if it did, like in the case of the infinite monkey theorem (a monkey with a typewriter given enough time will eventually write all of Shakespeare), would it actually have the kind of impact that Shakespeare had? Seems impossible because a work of art is appreciated in part because it is an attempt to communicate with/express something towards somebody. We often love to ask questions to artists, get to know them so we can know the work better, etc. If it came from a computer that experience would be gone.

On May 14 2023 01:26 Loco wrote: The complicatedness of the code cannot compare in kind with the complexity of hundreds of millions of years of self-organizing, evolving biological life.

Yeah, that's basically why I said I don't consider chatGPT to be a step forward for AGI. It made no scientific advancements in understanding how the mind works, even if it's useful for engineering. Chatgpt is not nearly as complicated as a single-cell organism. So that should give an indicator as to how far off we are. The algorithms that A.I researchers use to find meaning in text basically just count how often words co-occur together. That's a simplification, but that's really what it comes down to.

The neurons in biology are beyond complex, they themselves contain millions of computers processing DNA and all that. The ones in Artificial neural networks are simply nodes that take numbers in, apply a function, and spit numbers out. Then adjust those numbers by looking at the mistakes they make, until the numbers are just right. The more recent ones add a few bells and whistles to each neuron, but are still doing basic math calculations. That should tell you the difference in complexity compared to biology and A.I.

Also to Drone, yes, A.I. did make poker boring.

On May 14 2023 01:26 Loco wrote: but it never will create a piece of art that can really touch and change people.

That seems empirically testable. We could do a blindfold test for human music and A.I. music and check the results.

On May 14 2023 02:51 Stroggoz wrote: Show nested quote + On May 14 2023 01:26 Loco wrote: The complicatedness of the code cannot compare in kind with the complexity of hundreds of millions of years of self-organizing, evolving biological life. Yeah, that's basically why I said I don't consider chatGPT to be a step forward for AGI. It made no scientific advancements in understanding how the mind works, even if it's useful for engineering. Chatgpt is not nearly as complicated as a single-cell organism. So that should give an indicator as to how far off we are. The algorithms that A.I researchers use to find meaning in text basically just count how often words co-occur together. That's a simplification, but that's really what it comes down to.

I told my coworker about ChatGPT and that it can make certain school documents that she needs for work, like an IEP for a student. She said it did good, but she had to modify a few things.

This is was really coming even before chatgpt but all chatbots will definitely become 100x smarter. These types of jobs will be the first ones to go. BPOs and offshore customer support types will probably stick around for another year or so but who knows

https://www.wsj.com/articles/wendys-g...on-order-taker-an-ai-chatbot-968ff865

On May 13 2023 22:56 Liquid`Drone wrote: It's not that an AI can't create a creative product that will match my or other people's preferences perfectly well. I'm sure the more time that passes and the more practice and feedback they get, the better they'll be. It's already plenty good at that, anyway. But at least to me, appreciating art is actually about appreciating the creative human genius behind it, not just about appreciating the product. I'm not worried about AI 'overtaking the world' or whatever, I'm worried about AI becoming better than humans at all the various fun human activities that are fun to engage in, to the point where attempting to become 'great' at them ends up feeling kinda meaningless because even if you spend three decades mastering something, some jerk can just ask a computer to make something better.

It's almost certain that will be the case, but I think we will make strong differentiation between AI/human created and place differnet value on it.

I mean, Magnus Carlsen's play isn't any less impressive only because an app in a phone can play much better and we still watch weight lifting in the olympics even if a forklift can lift 3 tons.

On May 14 2023 01:26 Loco wrote: If a computer can do it better then it should, and it's evidence that it was meaningless to begin with. That's what computers do. They don't deal with meaning. They can recreate popular patterns that a lot of people find pleasing in music, sure, but they cannot create thematic art that is actually meaningful. It can't have depth. It can certainly be programmed to use a lot of poetic words, and even superficially seem like it's saying something that matters, but it never will create a piece of art that can really touch and change people. It's not in the code and can never be. With computers and humans we are dealing with differences of kind, not degrees. It doesn't matter how advanced it gets because it's not about the complicatedness of the code. The complicatedness of the code cannot compare in kind with the complexity of hundreds of millions of years of self-organizing, evolving biological life. We can express that complexity through sound because we have an inner need for it, a need that a computer will never be able to understand and therefore won't reproduce. No matter how advanced it gets it would never write anything that matters. But even if it did, like in the case of the infinite monkey theorem (a monkey with a typewriter given enough time will eventually write all of Shakespeare), would it actually have the kind of impact that Shakespeare had? Seems impossible because a work of art is appreciated in part because it is an attempt to communicate with/express something towards somebody. We often love to ask questions to artists, get to know them so we can know the work better, etc. If it came from a computer that experience would be gone.

You are talking two completely different things, first you say machines will never surpass humans in lets say music, and cannot create a song that touches and changes people I think this is completely wrong, they will be able to even mimic artits that you wouldn't be able to tell if the artist created it or it was AI.

And it is not that I understimate the depth of human art at all, the complexity of the interaction of the art and the feelings and thoughts evoked is massive, however AI due to its processing capabilities is very good at knowing what we like and what we don't and copy it, and as much as we'd like to believe we can tell a genuine from a "copy" we can't.

I do agree that the experience hence the capacity of art to move and change people is mostly gone when you know it didn't come from a human, but that is only the case when you wan't that experience most of art is low-level and mass consumed, nobody will care about AI generated electronic music for example, or AI generated posters/ADs.

The scary thing about all this is this particular leap is available to everyone now. Coding alone vastly reduces the need for huge teams and for the malicious guys in the space this is a godsend. Imagine this thing being available is barely even in it's 6th month and each day there's some crazy disruption or illegal activity going on. It's impact is actually harder to contain than the internet or cryptocurrency.

https://www.nytimes.com/2023/05/01/te...le-chatbot-engineer-quits-hinton.html

guy who helped create this thing even resigned. Elon Musk who helped fund it is even upset that it's turned into a for-profit model when in fact it was designed to not be one.

You could probably implement a type of barcode into every A.I.- generated character and image by generating a sequence of slightly different rbg values into several pixels that are not human-detectable and are a distinct pattern. Then governments could regulate the use of algorithms to enforce these barcodes, at least for A.I. generated images and text. This would allow machines to easily detect whether something else is made by a machine. If people copy pasted text from chatgpt they would be copy pasting the barcodes. It would need strong legal enforcement since you could also make machines that would erase the barcode. That's my idea.

People who think AI will never be as smart or creative or well rounded as humans: you're not as smart as you think you are, and you're orders of magnitude less smart than true AI will be.

Other than someone bioengineering unstoppable deadly viruses, this is the biggest threat to our existence as a species.

On May 30 2023 16:16 Daut wrote: People who think AI will never be as smart or creative or well rounded as humans: you're not as smart as you think you are, and you're orders of magnitude less smart than true AI will be. Other than someone bioengineering unstoppable deadly viruses, this is the biggest threat to our existence as a species.

The implication behind thinking we can achieve AGI anytime soon is thinking that we humans are so unbelievably intelligent that we are going to figure out things like free will, consciousness, and other things we have made virtually no progress on for the last thousands of years. Either that or the view that humans can be explained with gradient descent, cross-entropy loss, and backpropagation, which is pretty stupid. You cannot explain much less complex animals with deep learning algorithms. So yes, thinking we are close to AGI is pretty arrogant, on top of being an irrational view, and there being a highly inaccurate record of prediction in this research area.

On May 30 2023 16:16 Daut wrote: People who think AI will never be as smart or creative or well rounded as humans: you're not as smart as you think you are, and you're orders of magnitude less smart than true AI will be. Other than someone bioengineering unstoppable deadly viruses, this is the biggest threat to our existence as a species.

Facts

On May 30 2023 16:16 Daut wrote: People who think AI will never be as smart or creative or well rounded as humans: you're not as smart as you think you are, and you're orders of magnitude less smart than true AI will be. Other than someone bioengineering unstoppable deadly viruses, this is the biggest threat to our existence as a species.

If we're not as smart as we think we are, why do you trust your own judgment of what is the biggest threat to our existence? That's one hell of a leap to make. How would you even weigh the possible damage of intentional harm from bioengineering vs what could come unintentionally in our unnatural world?

There is only one group of people who thinks AI is the biggest threat to our existence, and it's people working in tech. Those people are not the "smartest" people in the world. They have a specific kind of intelligence and view the world from a limited perspective. Usually these people are pretty well off financially, which tells you something.

It amazes me that people can believe in this ideology when it has no obvious way to be an issue without us interfering to include it in our lives. AI can't be the biggest threat to our world simply because it is going to be in our hands at all times, unless it can become embodied, which, by the time it could, this planet will already be one big fireball. AI is dependent on architecture that receives inputs from us. It can always be turned off. It is a threat for other reasons than existential ones.

What is even a good example of creativity from an AI? Has an AI ever produced something completely new and original? Has it been an overt influence on people's thinking? Does AI have ideas that philosophers work with in philosophy departments?

will the self driving tesla taxis be out before ai is gonna kill us all? maybe it's right after we colonize mars...

On May 30 2023 16:38 Stroggoz wrote: Show nested quote + On May 30 2023 16:16 Daut wrote: People who think AI will never be as smart or creative or well rounded as humans: you're not as smart as you think you are, and you're orders of magnitude less smart than true AI will be. Other than someone bioengineering unstoppable deadly viruses, this is the biggest threat to our existence as a species. The implication behind thinking we can achieve AGI anytime soon is thinking that we humans are so unbelievably intelligent that we are going to figure out things like free will, consciousness, and other things we have made virtually no progress on for the last thousands of years. Either that or the view that humans can be explained with gradient descent, cross-entropy loss, and backpropagation, which is pretty stupid. You cannot explain much less complex animals with deep learning algorithms. So yes, thinking we are close to AGI is pretty arrogant, on top of being an irrational view, and there being a highly inaccurate record of prediction in this research area.

Just like pretty much every invention since the dawn of time we will stumble upon it while studing it, we dont need to "understand free will and consciousness" to create it, nobody in the field has even mentioned something like that, if it happens it will be from a self-improving mechanist just like we improved our own consciousness as species through time.

On June 09 2023 03:25 Loco wrote: Show nested quote + On May 30 2023 16:16 Daut wrote: People who think AI will never be as smart or creative or well rounded as humans: you're not as smart as you think you are, and you're orders of magnitude less smart than true AI will be. Other than someone bioengineering unstoppable deadly viruses, this is the biggest threat to our existence as a species. If we're not as smart as we think we are, why do you trust your own judgment of what is the biggest threat to our existence? That's one hell of a leap to make. How would you even weigh the possible damage of intentional harm from bioengineering vs what could come unintentionally in our unnatural world?

Daut: We overestimate our intelligence
Loco: If we are so dumb, how come you can think

lol

There is only one group of people who thinks AI is the biggest threat to our existence, and it's people working in tech.

If virologists working bioengineering viruses told me that these viruses are the biggest existencia threat to mankind I'd fucking listen.

Usually these people are pretty well off financially, which tells you something.

They are also probably mostly CIS straight white men, shees... don't listen to them.

It amazes me that people can believe in this ideology when it has no obvious way to be an issue without us interfering to include it in our lives. AI can't be the biggest threat to our world simply because it is going to be in our hands at all times, unless it can become embodied, which, by the time it could, this planet will already be one big fireball. AI is dependent on architecture that receives inputs from us. It can always be turned off. It is a threat for other reasons than existential ones.

Yeah it will take 5 billion years to get embodied AI, It's not like there are already experimental military drones running on AI or millions of electric vehicles running on crude AI and will have billions of them in the upcoming decades.

Yeah I guess if it goes out of control we can turn the internet off, destroy all our satellites cut off the intercontinental cables, if that doesnt work we can pull the electricity lerver off, and if they somehow harvest their own energy then we can go full animatrix and block the earth from the sun, theres always an off button lol.

The whole argument is that when we realize is too dangerous it will be too late, it will be too smart and far too integrated with us, how can you miss the main point of the danger, wich is that when it "switches on" there is no turning back, we rolled the dice.

What is even a good example of creativity from an AI? Has an AI ever produced something completely new and original? Has it been an overt influence on people's thinking? Does AI have ideas that philosophers work with in philosophy departments?

"But does it create" hahaha the same quality of argument of "But does it have a soul".

Cant wait for AI to keep absolutely demolishing the creative areas before it even can open a soda bottle while leftoids are debating wether if the AI's masterwork is original or a collage of images, because only men can conjure from the ether

Descartes's definition of a soul is actually fairly scientific at the time. He basically says it's this thing that gives humans the ability to reason, whatever that thing is. So to say we have a soul could be interpreted as saying "Do we have a thing that allows us to reason", and so you could very reasonably argue that human beings have souls, since they can reason. The issue is that it's a very vague definition. This however might be acceptable for the time period.
Descartes was a genius but he was a self-conscious bootlicker and had megalomania, and was also a sadist who enjoyed torturing animals. He may also be partly responsible for the climate crisis due in part to entrenching Western philosophy with materialist views. Francis Bacon is another culprit.

I'm not sure if there is an off-button. Cloud Storage uses something called sharding which replicates and distributes your data to servers on 3-4 continents, to dramatically reduce the chance of it ever being erased from a war or a massive flood, ect. The only off button would be to simultaneously destroy all data centres in the world. I don't believe this Terminator thing is anything more than science fiction at this point though heh.

Also, people need to stop associating working in "tech" with being a silicon valley billionaire douchebag.

A.I. is glorious technology imo, and of great benefit to everyone. I refuse to shift blame for social credit score calculations from the CCP onto something like "A.I."