New Subforum Added - The AI Age

[Mona Eberhardt]

I'd be happy with "The Neural Network Age, 7th Edition"

Or whatever other buzzword...

 

[Argent Stonecutter]

The concept that most people truly think of as AI is AGI - where the AI learns on its own and matches or exceeds human intelligence and capabilities. AGI has not been achieved, but there are estimates that it may happen in 2024 or 2025, given the pace of innovation.

I don't think so. There's not really a lot of innovation going on right now, just tweaks on a clever hack that's kind of a dead end when it comes to AGI... and really more valuable to corporations than actual AGI with its accompanying ethical and moral issues would be.

 

[Cristiano]

I don't think so. There's not really a lot of innovation going on right now, just tweaks on a clever hack that's kind of a dead end when it comes to AGI... and really more valuable to corporations than actual AGI with its accompanying ethical and moral issues would be.

From stuff I have read and watched about AGI, one of the key components to developing it is world simulation. AI right now does not understand the physical world in the way that we do. Tools like Sora from OpenAI could potentially lead to that piece of the puzzle. I don't think AGI is far off, honestly. There are profound ethical issues involved, certainly, but there is now an arms race between Microsoft and Google, which like it or not, is moving things forward very quickly.

Video generation models as world simulators

We explore large-scale training of generative models on video data. Specifically, we train text-conditional diffusion models jointly on videos and images of variable durations, resolutions and aspect ratios. We leverage a transformer architecture that operates on spacetime patches of video and...

openai.com

Unsurprisingly, Meta disputes that. They are trying to position themselves in this space as well:

Meta's chief AI researcher says OpenAI's "world simulator" Sora is a dead end

Here is some more about AGI - which is a nicer name for the singularity - a lot believe we are 20+ years away from it:

AGI/Singularity: 9,800 Predictions Analyzed

Explore key predictions from AI scientists, leading entrepreneurs, and the community predictions on AGI/singularity timeline.

research.aimultiple.com

 

[Argent Stonecutter]

From stuff I have read and watched about AGI, one of the key components to developing it is world simulation. AI right now does not understand the physical world in the way that we do. Tools like Sora from OpenAI could potentially lead to that piece of the puzzle.

Sora doesn't seem fundamentally different from any other generative neural net. I don't think the technology is fundamentally capable of modelling anything in a way that would lead to AI.

I don't think AGI is far off, honestly. There are profound ethical issues involved, certainly, but there is now an arms race between Microsoft and Google, which like it or not, is moving things forward very quickly.

But not in the direction of AI, let alone AGI. I don't think they actually *want* AGI.

I also don't think that most people who talk about the Singularity have any idea of what it means. I think the vast majority of "experts" who write about it are deliberately misinterpreting it so they can milk the wave.

The Coming Technological Siongularity -- Vinge (1993)

 

[Katheryne Helendale]

The concept that most people truly think of as AI is AGI - where the AI learns on its own and matches or exceeds human intelligence and capabilities. AGI has not been achieved, but there are estimates that it may happen in 2024 or 2025, given the pace of innovation.

I, for one, welcome our new computer overlords.*

*As long as they don't byte.

 

[GoblinCampFollower]

Here is some more about AGI - which is a nicer name for the singularity - a lot believe we are 20+ years away from it:

AGI/Singularity: 9,800 Predictions Analyzed

Explore key predictions from AI scientists, leading entrepreneurs, and the community predictions on AGI/singularity timeline.

research.aimultiple.com

In all cases, the majority of participants expected AI singularity before 2060.

Fun bit of trivia, Isaac Newton was so annoyed by the constant end of days predictions in his day that he did his own Isaac Newton math and determined the world would end no sooner than 2060

I think this article is reasonable to me. I think it's very hard to predict what will happen +20 years from now but do agree AGI is a good ways off. As I said before, I think AGI is totally different algorithms than just tweaking machine learning models and it might not exist. I think we'll develop new "clever machines" over the next few decades but we might need this much time just to know what AGI even really is.

That said, I do think something resembling AGI should be possible since I don't believe the human brain is magical. If humans can do it, a computer should be able to do it. It's when, not if.

 

[Imnotgoing Sideways]

the world would end no sooner than 2060

Newton said the world is ending in 2060!? Wasn't he the guy that invented gravity!? Coincidence? I think not. Birds aren't real!

 

[GoblinCampFollower]

Newton said the world is ending in 2060!? Wasn't he the guy that invented gravity!? Coincidence? I think not. Birds aren't real!

I know you're just being silly but to clarify, Newton didn't say "in 2060" he said "no sooner than 2060." He was mostly just trying to shut people up.

 

[Dakota Tebaldi]

Sora doesn't seem fundamentally different from any other generative neural net. I don't think the technology is fundamentally capable of modelling anything in a way that would lead to AI.

But not in the direction of AI, let alone AGI. I don't think they actually *want* AGI.

I also don't think that most people who talk about the Singularity have any idea of what it means. I think the vast majority of "experts" who write about it are deliberately misinterpreting it so they can milk the wave.

The Coming Technological Siongularity -- Vinge (1993)

You've talked about this before and you've kind of convinced me, really - of the idea that LLM's can't really lead to AGI. I don't know if I agree that they don't want it; but it does kind of feel to me like the laser-focus on LLMs has to be a conscious attempt to cheat, like somewhere along the line they decided that if they could make a convincing-enough chatbot they could all just kind of quietly agree that it's "AGI" and implicitly challenge people who disagree to prove that it's not.

What gives it away to me was the "revelation" as it were by OpenAI's engineers that they don't know how exactly GPT correlates its training data, it "just does". Like - my dude, it's a computer. Whatever it's doing is math, so no matter how complex it is, it IS quantifiable. It's just completely wild to me that a computer scientist - or any scientist - could just say "welp, guess we invented a black box lol" and be completely happy with that, unless they saw some kind of special value in leaving the problem unsolved.

 

[Beebo Brink]

That said, I do think something resembling AGI should be possible since I don't believe the human brain is magical.

Someday humans may look back on this statement the way we currently look back on the theory of humors, laughing at the "obvious" absurdity of 21st century views on intelligence and organic brains.

The unspoken assumption in your statement -- a perfectly reasonable statement, btw, given our current paradigms -- is that the brain is the singular structure that contains our intelligence. Mimic the brain's processes and voila! we have an intelligent machine!

Just suppose, however, that we don't have it right. What if our brain is only part of the answer to where our intelligence comes from? What if intelligence is more diffuse, built of sensory input gathered throughout our entire body? What if the brain just pulls it together, like a collator? That level of complexity would be significantly more difficult to mimic with circuits.

I have a sneaking suspicion we're not even close to understanding the true nature of intelligence, certainly not enough to try duplicating that function.

 

[Argent Stonecutter]

Just suppose, however, that we don't have it right. What if our brain is only part of the answer to where our intelligence comes from? What if intelligence is more diffuse, built of sensory input gathered throughout our entire body?

That just increases the computation required, it doesn't fundamentally change the argument that consciousness is mechanical and can be replicated in a computer. The same applies to additional requirements people imagine that the simulation include the environment, or multiple people, or external inputs from the physical universe. These are not deal-killers, they are just haggling about prices.

 

[GoblinCampFollower]

Someday humans may look back on this statement the way we currently look back on the theory of humors, laughing at the "obvious" absurdity of 21st century views on intelligence and organic brains.

The unspoken assumption in your statement -- a perfectly reasonable statement, btw, given our current paradigms -- is that the brain is the singular structure that contains our intelligence. Mimic the brain's processes and voila! we have an intelligent machine!

Just suppose, however, that we don't have it right. What if our brain is only part of the answer to where our intelligence comes from? What if intelligence is more diffuse, built of sensory input gathered throughout our entire body? What if the brain just pulls it together, like a collator? That level of complexity would be significantly more difficult to mimic with circuits.

I have a sneaking suspicion we're not even close to understanding the true nature of intelligence, certainly not enough to try duplicating that function.

As Argent pointed out, your statement doesn't change my argument, it just means it will take a bit longer.

 

[Cristiano]

When I was kid and playing Little Computer People, little did I know we are actually the little computer people inside of a simulation.

 

[Monica Dream]

When I was kid and playing Little Computer People, little did I know we are actually the little computer people inside of a simulation.

 

[Free]

All I know is, my brain is hurting right about now.

 

[Cristiano]

As part of my reorg, I renamed this to simply Artificial Intelligence.