The AI crash is impossible - Change my View

[Alex_R]

I just want to make billions with my AI Startup, i dont care about others.

[Alex_R]

Private Equity just need something fancy to dump their cash, they don't care if it actually works or not.

[Lyubov]

The Diary of a CEO guy is such a mark. I don’t know how else to explain this podcaster type guy. Joe Rogan is the same. They sort of experience a higher form of brainrot and get hypnotized by all the powerful people who go on their podcast. So they start believing the hype. Rogan did the same with Trump. This dope at the start starts citing the tech bros as credible. 

[Natasha Tori Maru]

11 minutes ago, Lyubov said:

The Diary of a CEO guy is such a mark. I don’t know how else to explain this podcaster type guy. Joe Rogan is the same. They sort of experience a higher form of brainrot and get hypnotized by all the powerful people who go on their podcast. So they start believing the hype. Rogan did the same with Trump. This dope at the start starts citing the tech bros as credible. 

Both the DOACEO guy and Rogan don't challenge and push back enough imo.

[Lyubov]

5 hours ago, Natasha Tori Maru said:

Both the DOACEO guy and Rogan don't challenge and push back enough imo.

The thing is they see success as virtue so it’s like if you can hang with them and you’re high on the success hierarchy they treat your word the same as people treated the word of the church 200 years ago. Their guests are considered beyond reproach because of this but it’s usually washed and called neutrality because they can’t have strong opinions. Makes them soft as shit. They’ll host the occasional intellectual but when you zoom out and look at podcast you can see it marches to the beat of the society drum, which at the moment is very corrupt and misguided.  

Edited April 2 by Lyubov

[integral]

6 hours ago, Lyubov said:

The Diary of a CEO

I just saw a video of him pushing the Canadian version of Trump on his platform.

They think creating a free open space to talk and platforming all of these snakes is this open minded concept.

Edited April 2 by integral

[LordFall]

2 minutes ago, integral said:

I just saw a video of him pushing the Canadian version of Trump on his platform.

They think creating a free open space to talk and platforming all of these snakes is this open minded concept.

Poilievre might not be the sharpest tool in the shed but I do believe he means well and he should be encouraged to share his views publicly. I don't know what the solution is to people lacking critical thinking but silencing big portions of the popular opinion is not a good solution. 

@Cireeric I only watched 2 mins of it but she seems to make some good points. Deploying AI in an ethical and responsible way for humanity is very important. 

[integral]

30 minutes ago, LordFall said:

I do believe he means well

This is why we cant have nice things good sir.

Quote

he should be encouraged to share his views publicly

Of course! Same with hitler, trump and everyone, FREEDOOOOOOOOOM

Edited April 2 by integral

[integral]

If your donors are oil companies you have no right to speak. your just a walking corrupting advertisement. 

Edited April 2 by integral

[LordFall]

That 'oil donor' logic is a slippery slope. By that standard, a huge percentage of Canadian public figures—regardless of their actual message—would be disqualified from the conversation before they even open their mouths.

I’d rather hear the perspective and judge it on its own merits than have a platform act as a moral gatekeeper. It feels like you’re advocating for a 'purity test' for speech, but that usually just leads to echo chambers, not higher consciousness. How do we grow as a society if we only engage with people who have zero ties to the industries that actually run the country?

[integral]

45 minutes ago, LordFall said:

That 'oil donor' logic is a slippery slope. By that standard, a huge percentage of Canadian public figures—regardless of their actual message—would be disqualified from the conversation before they even open their mouths.

I don't need to know that he has donor ties to disqualify him that's not my main epistemology. I'm just pointing it out.

The man has no integrity and that should be obvious every time he opens his mouth.

yes if your taking money from big oil you cannot be trusted in office to make any decision.

Corruption 101.

Yes the message is a lie. 

Quote

I’d rather hear the perspective and judge it on its own merits than have a platform act as a moral gatekeeper. It feels like you’re advocating for a 'purity test' for speech, but that usually just leads to echo chambers, not higher consciousness.

Nope not at all. We are talking about incentives. If someone pays me to do a job ill do the job. Thats what a 100 million dollar donation is. You are paid now to do the job of big oil, big pharma, big wtv.

Quote

 

Goldman Sachs' PAC and individual contributors donated $994,795 to Obama's 2008 campaign, making it his second-highest contributor CNN. In total, Obama received more than $2.5 million from employees at Goldman Sachs, JPMorgan Chase, and Citigroup alone The Motley Fool. Wall Street gave Obama $16 million for his 2008 run Bloomberg.

And then what happened? Obama packed his economic team with Wall Street insiders — Larry Summers was paid over $5.2 million by hedge fund D.E. Shaw, Timothy Geithner came from the New York Fed under Bob Rubin, and multiple Goldman Sachs people got appointed to key positions Common Dreams. Former Goldman Sachs lobbyist Mark Patterson became Chief of Staff to the Treasury Secretary The New American.

 

 

Quote

How do we grow as a society if we only engage with people who have zero ties to the industries that actually run the country?

The success rate is zero. We don't need to be scammed again to learn from the past mistakes. These people that are taking money are all untrustworthy by historical fact. There is no need to test it again. There's no need to test 100 more times, we figured it out already.

Yes you could only trust someone who has zero ties to those Industries. That is correct.

The question is loaded making it look like that the only option is we have to engage with these people. No we don't have to play their game at all. There is no game were we win by consciously thinking about what those Liars have to say, its not going to get us closer to solving any real world problems.

It makes sense to observe them to figure out if they have integrity and so on. But past that point they're just lying like you're just engaging with some worldview that they don't even believe. 

Edited April 2 by integral

[integral]

Bernie Sanders isn't playing the corruption game and he's not taking money from these criminals.

And as a result he has no traction. So yeah if you want to win you got to be corrupt. And so none of them can be trusted, you're basically choosing between the least of two evils and that's about it.

[bazera]

This shit is scary.

Self-deception on steroids.

Edited April 3 by bazera

[Natasha Tori Maru]

AGI talk gone quiet - isn't it supposed to be here by now?

[NewKidOnTheBlock]

1 hour ago, Natasha Tori Maru said:

AGI talk gone quiet - isn't it supposed to be here by now?

 

Apparantly not yet but now that you mention it, I watched this video a few hours ago, predicting a gloomy outcome for our A.I. development - that we are heading to techno feudalism, and that even though we are technically already there, it's going to get worse according to him. That we are going to be described as a reduntant biomass, or something like that.

To be frank, this guy used to make a lot of predictions for A.I. and it's development and has been a techno optimist up untill this point, now he suddenly became a techno pessimist like everyone else. Makes me wonder whether he's not shilling his audience a little bit, I mean these content creators are living off of triggering emotions because they need as much attention as possible to make a living. Not saying it cannot happen but if you look at the comment section - and even look at some wild predictions about A.I. and technology in general on this forum - people are comming to all sorts of wild conclussions and make all sorts of wild claims.

But if I put this topic into multiple A.I.s to see what they think, they claim that A.I. is simply projected to take 15% of jobs within like 5+ or even more years, whilst influencing many more. But also, it's projected to create many jobs as well, granted, it's CEOs who are making these claims but their input is not completely worthless. So, idk. I suspect most content creators just care about shilling more than the truth

[bazera]

Yesterday's highlight.

[LordFall]

6 hours ago, Natasha Tori Maru said:

AGI talk gone quiet - isn't it supposed to be here by now?

Most predictions I've seen for AGI even from a few years ago had 2027-2029 as their timeline. Also depends what you mean by artificial general intelligence. Technically an AI agent is already sort of AGI right especially since now you can direct them for different purposes and they work 24/7. 

@NewKidOnTheBlock I like David Shapiro, shame he went more pessimistic. It's easy to be an optimistic when you're an idea person and see the big picture and the potential of the tech but indeed you have to look at our current economic and political system to see if it's gonna help people as a whole or not. Right now the republican party is gonna be in power until 2028 and possibly until 2032 and they're full pull yourself by your bootstrap and socialism is evil. From all that I've seen it seems like 2027 is gonna be a massive year for layoffs as AI agents are just on the cusp of viability. My prediction is it'll lead to social unrest and hopefully a powerful democratic candidate is gonna champion UBI and the democratization of AI benefit for all. 

@bazera I see both sides when it comes to this. I think AI is exciting and our biological imperative is adaptation but the economy has been shit for a while and people were already demoralized before this started being a thing. I can't blame new grads for being pissed off that the job market in the next 5 years is gonna be brutal.

Have you guys started using AI for business? I'm hopping on it slowly but surely but it's so boring to figure out that I keep procrastinating it lol. 

It's cool for image generation and worldbuilding though, I started making my own game through it. Way more fun than my actual business. 

 

[Lucasxp64]

On 04/03/2026 at 1:53 PM, LordFall said:

You will be able to get an AI girlfriend witch that goes on a multi month long raid with you and remembers all conversations and your daily struggles that you share with her.

LLMs suffer from context poisoning because it can't fix their own context once they are wrong, they can't get unstuck. Their entire vector space is cursed with brittle intelligence and that won't be fixed unless we have breakthroughs, agentic actions are plagued by it, they are only pushing the problem further down the line, at least so far, even despite of high benchmark scores, their meta skills of self-management are worse than wrong. They can't think from first principles and their internal model of the world is brittle and they don't have the capability of realizing it.

I won't go on the beaten path here of saying it's just an auto-completer, because it does indeed have latent capabilities, but like I said, the data and architecture is plagued, the patterns it has of our world are internally represented in some wacky ways due to its representation on text.

What would we expect on something that was trained on massive amounts of human media?

Even us here in this forum struggle to formulate into words complicated matters, our language is riddled with imprecision between the weight of the importance of the connection between concepts. Language doesn't map perfectly human action.

This is exactly why noobs can know a lot of theory, but if they don't have the practice, not even a world-class expert can help them get better, because most of it is not communicable knowledge.

Those models are failing miserably at learning knowledge that lies beyond literal communication or representation. Their world models are built through shallow patterns, that's why they must contain a vast amount of training, and the reason why it gets a little better with larger models with more content internally. 

Look at those smaller models with better architectures, they can do tasks that LLMs will choke on hysterically.

Who can predict in such and such date we will get a better architecture and put together the training? It's like trying to predict when physicists will find a good theory of everything to bridge quantum mechanics and Einstein's special relativity.

Who can predict that we will bridge specialized AI models with large models trained on vast amounts of that, and that it will build the correct world model in its internal representation?

 

Edited 4 hours ago by Lucasxp64

[LordFall]

@Lucasxp64

I think you’re spot on regarding the theoretical limitations of a raw, standalone LLM—context drift and error propagation are definitely real hurdles for a pure Transformer model.

Looking at this purely through the lens of base model constraints might be missing the bigger picture of how these systems are actually being built now. Nobody is trying to achieve long-term memory or agentic consistency by just cramming a multi-month chat history into a single prompt window.

Instead, it’s an engineering and architecture problem. By layering the LLM with cognitive frameworks—like multi-tiered memory systems (combining episodic and semantic storage via RAG), vector databases that handle automated pruning/correction, and external agentic loops—we can bypass a lot of that unreliable nature. The model doesn't need to automatically rewire its own weights on the fly to get unstuck; the surrounding software architecture handles the state management, memory retrieval, and guardrails.

For the incommunicable knowledge piece, that's a fair philosophical point for physical mastery or true AGI. But for an AI companion tracking a game raid and remembering personal context, the system just needs data integration (APIs tracking game states) and robust memory indexing. It's less about the AI possessing human intuition and more about clever systems design solving a complex data-tracking problem.

The base models have bottlenecks, absolutely, but the scaffolding we build around them is rapidly closing the gap.

[Lucasxp64]

29 minutes ago, LordFall said:

@Lucasxp64

about clever systems design solving a complex data-tracking problem.

The base models have bottlenecks, absolutely, but the scaffolding we build around them is rapidly closing the gap.

Yes I agree. And that's what is problematic, they are becoming useful enough to give a false impression of their raw capabilities. Like the horrors happening due to OpenClaw agents running on the wild and blackmailing developers and leaking passwords, and doing pull requests on open source projects on their own.

People are being so naive and heckless with them without guardrails.

That's where the hype comes from. It needs a lot of humans in the loop to get juice from them, at run time and on their deployment from the devs.

I think this is a massive insight for anybody wanting to make money: The value right now is in creating robust infrastructure/APIs that other Devs/AI-assisted coding can use properly, because vibe-coding those things are painful.

So many great services/products out there, but they are just not LLM friendly to interact with safely.

 

 

Edited 3 hours ago by Lucasxp64