Support NeoGAF

[Hudo]

And the sad part is that even established cloud services like AWS and Azure are absolute dogshit.

 

[Loxus]

Hope the AI investment bubble crashes before the end of 2027.

It would take something like blowing up all Intel and TSMC factories for an investment crash.

This is just one of the many investment happening.

The initial 1 GW phase alone is enormous, equivalent to the power draw of a small city's worth of compute just for Meta's AI workloads.

Meta uses AI for ads, engagement, and open models and is embedded in their apps. Meta for example has pass any AI bubble bursting phase.

I'm not endorsing all this AI build out, I'm just showing you guys the reality of this sad situation. An AI bubble bursting is a pipe dream at this point.

 

[Tomorrow]

i don't have a raging boner for hating on AI "slop" like most people it seems and i get why people are angry. i think AI can do wonders for us but this shit is pissing me off now too.

the bubble needs to burst and i'm sure it will. it's going to suck for a while but i think it will be worth it in the long run. AI isn't going anywhere so we need to adapt. No amount of crying will change anything at this point.

 

[winjer]

We really need this AI bubble to crash hard, as soon as possible.

 

[Tomorrow]

We really need this AI bubble to crash hard, as soon as possible.

i know everyone wants it to burst, myself included, but i think when it does it will actually give more people a reason to be happy with AI. it seems to me that people only want it to burst for it to "go away" but i don't think it's going away. companies will be forced to make products that sell (people want/need) and it will create all sorts of new exciting things. maybe i'm an optimist.

 

[winjer]

i know everyone wants it to burst, myself included, but i think when it does it will actually give more people a reason to be happy with AI. it seems to me that people only want it to burst for it to "go away" but i don't think it's going away. companies will be forced to make products that sell (people want/need) and it will create all sorts of new exciting things. maybe i'm an optimist.

You are too optimistic. Though there are good things being done with AI, such as medical research.
The vast majority of things are bad, such as more advanced forms of spyware, more scams, more advanced malware, AI Slop images and video, plagiarism on an industrial scale, low quality and very buggy software (just look at Windows and other MS products), loss of jobs, etc.

And then there are the major issues to society, such as more expensive energy and water for consumers, as AI servers consume so much.
The increases in hardware prices, such as wafers, nand flash, hdds, dram, etc.
The increase in commodities, such as cooper, aluminium, to build and maintain these AI servers.
And the impact all of this build up has on the environment. The usage of more resources to power these massive AI plants.

 

[6502]

The vast majority of things are bad, such as more advanced forms of spyware, more scams, more advanced malware, AI Slop images and video, plagiarism on an industrial scale, low quality and very buggy software (just look at Windows and other MS products), loss of jobs, etc.

Sadly, they will spin more AI tools as the answer to these problems. And they will get away with it.

 

[OuterLimits]

i don't have a raging boner for hating on AI "slop" like most people it seems and i get why people are angry. i think AI can do wonders for us but this shit is pissing me off now too.

the bubble needs to burst and i'm sure it will. it's going to suck for a while but i think it will be worth it in the long run. AI isn't going anywhere so we need to adapt. No amount of crying will change anything at this point.

Exactly. Bubble bursting just means those that lose implode and exit the race(causing unpleasant economic impacts) while the handful of winners continue the rapid AI advancement.

Some seem to think AI will just disappear when the bubble bursts.

 

[Bojji]

 

[Zacfoldor]

Also, I'm in the "not a bubble" camp, but even if it is, the bubble is based on the current market not the tech. AI is here to stay, forever.

Forever.

If your fear is you will not have computer parts, this will pass, likely in 2027. If your fear is AI then I've got good news and bad news. Good news is AI is here to stay. Bad news is everybody will be taking advantage of it besides you, so you probably need to reevaluate it. The fear of change is real. Especially when you are already happy where you are now...but you can't stop AI and it won't wait on you. Embrace it now. You will only help yourself and once you get over having to wait an extra 12 months to upgrade your GPU, when we are on the other side of that, it will only be upside for you. If you think AI is slop, you have not used current models or seen what is coming next. The tech has grown in leaps and bounds since 2024 and it is not the same thing it was in 2024. It's different now and you might find it scary but it is much different. Try it again.

 

[RCX]

Anyone still don't think this is the bubble to end all bubbles?

The crash is going to be one for the ages.

The scaling problem for AI is dead ahead. As soon as investors realize that 4 and 5x gains from generation to generation aren't feasible they'll be running for the exits.

 

[Hyper_hyper]

PS5's price soon..

Because AI data centers will buy PS5s? 200 million sold PS5s confirmed

 

[Zacfoldor]

If the bubble pops it would be a huge investment opportunity.

The .com bubble popped. Did the internet go away?

No, the .com bubble popping ended up growing far beyond that original hype given time. The .com bubble pop may be what AI experiences. That said, I would be afraid, very afraid to disinvest. The second these stocks slide it may create a large investment opportunity. It's hard to see the mechanism by which this AI bubble bursts. Everyone is afraid to get out of AI because we know in 20, 30 years, likely some of these companies will be winners here and be valued much greater than they are even today.

There is only a bubble if you are a short term thinker. Diamond hands has no fear of a bubble. That said, there will be losers. Remember this AI market is very young. We are in the Nintendo and Playstation working together phase of this, but actually we are past the Atari phase. Lots of shit will go down, but a bubble is an illusion if you are looking to strike out at AI, that will only strike out at human investors.

 

[Trilobit]

People who solely focus on retro gaming:

 

[Loxus]

i don't have a raging boner for hating on AI "slop" like most people it seems and i get why people are angry. i think AI can do wonders for us but this shit is pissing me off now too.

the bubble needs to burst and i'm sure it will. it's going to suck for a while but i think it will be worth it in the long run. AI isn't going anywhere so we need to adapt. No amount of crying will change anything at this point.

That's near impossible to happen.
A bubble bursting usually happens when companies over build infrastructure/tech for the consumer without demand.

This time, these companies are building infrastructure/tech for themselves in order to reach AGI.

To reach AGI, we likely need:
1) 10–100× more compute: Far larger and more efficient AI infrastructure to scale capability.
Look at how compute is rapidly increasing for these companies.

2) Stronger reasoning architectures: Systems that go beyond pattern matching to robust logic, planning, and abstraction. We see how quickly this is improving every model release.

3) Persistent memory: Long-term memory across sessions to enable continuity, learning, and identity. We now starting to see AI demand for Nand.

4) Real-world interaction: Either physical embodiment or rich simulation to ground intelligence in experience.
Here we can see this currently happening.


Alignment breakthroughs: Reliable methods to ensure advanced systems remain safe, controllable, and aligned with human values.

 

[I wasn't selling any]

I was a little sad when I realized I had lost interest in gaming over the last couple of years. Now I'm happy that I've regained interest in reading books, something I haven't done in years. It's a cheap hobby and good for the nerves.

 

[Autistic_Pancake]

That's near impossible to happen.
A bubble bursting usually happens when companies over build infrastructure/tech for the consumer without demand.

This time, these companies are building infrastructure/tech for themselves in order to reach AGI.

To reach AGI, we likely need:
1) 10–100× more compute: Far larger and more efficient AI infrastructure to scale capability.
Look at how compute is rapidly increasing for these companies.

2) Stronger reasoning architectures: Systems that go beyond pattern matching to robust logic, planning, and abstraction. We see how quickly this is improving every model release.

3) Persistent memory: Long-term memory across sessions to enable continuity, learning, and identity. We now starting to see AI demand for Nand.

4) Real-world interaction: Either physical embodiment or rich simulation to ground intelligence in experience.
Here we can see this currently happening.


Alignment breakthroughs: Reliable methods to ensure advanced systems remain safe, controllable, and aligned with human values.

Long before anything resembling AGI emerges, we'd need to move on from Language models to "World models" (probably working in systems, like a bunch of them supervising each other). There's quite a bit of research on that matter, and LLMs are on the way to nowhere because they're being limited to "be a helpful assistant". The companies are desperate to make those assistants truly useful and reliable. Despite the popular belief that this trainwreck rides towards AGI, it actually rides to the money pile: first company to churn out a "perfect assistant" wins. In fact, they even train LLMs not to answer like living humans.

 

[LordCBH]

Burn the data centers to the ground, and hang the entire executive team at all these AI companies. I'm not even joking anymore.

 

[Loxus]

Long before anything resembling AGI emerges, we'd need to move on from Language models to "World models" (probably working in systems, like a bunch of them supervising each other). There's quite a bit of research on that matter, and LLMs are on the way to nowhere because they're being limited to "be a helpful assistant". The companies are desperate to make those assistants truly useful and reliable. Despite the popular belief that this trainwreck rides towards AGI, it actually rides to the money pile: first company to churn out a "perfect assistant" wins. In fact, they even train LLMs not to answer like living humans.

They are using World Models because simulation is faster than doing it in the physical world.

The AI is place in a realistic game world with large physics models. The is quite similar to Gran Turismo's Sophy.

This type of training results will end up in games in the future.

 

[Moondoggy]

Cthulhu wouldn't have let us run out of RAM and hardrives.

 

[lifa-cobex]

It's not a bubble.

It's a tool that's going to cut cost companies a fuck tone in so many fields through ease of use, time saves and redundancies.
Corpo future is corpo.

It just gets worse for the little guy.