The Ai thread

[Katbel]

Actually maybe going ahead (maybe) just delayed and a lot of politics and bickering:

Stargate AI data centers for OpenAI reportedly delayed by squabbles between partners — sources say OpenAI, Oracle, and SoftBank disagreed on who would have ultimate control of the planned data centers

Sources say that Stargate has been stalled due to squabbles between stakeholders over site ownership and system control.

www.tomshardware.com

A lot still unclear how anyone is going to pay for this or what final actual size of these deals and facilities will be.

More they fight among them better it is for us:
they are just greedy and criminals, they don’t care for the consequences

 

[Yoused]

Study shows AI chatbots provide less-accurate information to vulnerable users

A recent study from the MIT Center for Constructive Communication reveals that leading large language models deliver less accurate, sometimes patronizing responses to users with lower English proficiency, less formal education, and non-US origins.

www.notebookcheck.net

The research also highlighted alarming disparities in how the models handled queries. Claude 3 Opus, for instance, refused to answer nearly 11% of questions for less educated, non-native English speakers, compared to just 3.6% for control users. In many of these refusals, the model responded with condescending, patronizing, or mocking language, occasionally mimicking broken English.

When you train on how people behave on the internet …

The entire internet is ESL, and it seems like Americans are the worst.

 

[dada_dave]

I doubt a federal moratorium is coming anytime soon with the current crowd but this is interesting.

 

[Katbel]

From the future, two years from now

THE 2028 GLOBAL INTELLIGENCE CRISIS​

A Thought Exercise in Financial History, from the Future​

 

[Nycturne]

From the future, two years from now

THE 2028 GLOBAL INTELLIGENCE CRISIS​

A Thought Exercise in Financial History, from the Future​

The whole thought experiment reminds me of this scene:


But it still echoes what I've been saying. The phase of AI we are in is a bit of a lose-lose. If it actually achieves what folks like Sam Altman claims, then this thought experiment is a plausible result. The economy crashes under the weight of mass unemployment of white collar workers and the resulting consumer spending collapse that ripples through the economy. Another great depression. If we are in a tulip mania (which I personally think is more likely), then when the bubble bursts a lot of assets get written down, pre-arranged deals torn up, and the resulting layoffs drop us into an economic crash. Is this also another great depression? I don't know, I suspect it'd be closer to 2008.

 

[Cmaier]

The whole thought experiment reminds me of this scene:


But it still echoes what I've been saying. The phase of AI we are in is a bit of a lose-lose. If it actually achieves what folks like Sam Altman claims, then this thought experiment is a plausible result. The economy crashes under the weight of mass unemployment of white collar workers and the resulting consumer spending collapse that ripples through the economy. Another great depression. If we are in a tulip mania (which I personally think is more likely), then when the bubble bursts a lot of assets get written down, pre-arranged deals torn up, and the resulting layoffs drop us into an economic crash. Is this also another great depression? I don't know, I suspect it'd be closer to 2008.

best movie ever

 

[dada_dave]

The whole thought experiment reminds me of this scene:


But it still echoes what I've been saying. The phase of AI we are in is a bit of a lose-lose. If it actually achieves what folks like Sam Altman claims, then this thought experiment is a plausible result. The economy crashes under the weight of mass unemployment of white collar workers and the resulting consumer spending collapse that ripples through the economy. Another great depression. If we are in a tulip mania (which I personally think is more likely), then when the bubble bursts a lot of assets get written down, pre-arranged deals torn up, and the resulting layoffs drop us into an economic crash. Is this also another great depression? I don't know, I suspect it'd be closer to 2008.

Ha! The lose-lose is exactly what I was just saying to someone else and it's been fascinating to see the two arguments play out online. I mean I suppose there's a middle ground where it's just useful enough to justify a sizable portion of the investment into it, but even then there's still going to be a big tech crash - in that case it might not get to 2008 levels of potential meltdown, but more like dot com sized (though bigger because everything is bigger - so "same size" relative to the economy/stock market size as a whole). But given that the economy is unhealthy in many other respects (it may be bigger but it's not in good shape), that would still be very, very bad.

At this stage, basically everyone agrees it's a bubble, it's just a question of when and how big it bursts and why and what the after effects are.

best movie ever

I really should watch it, I've never actually seen it.

 

[Katbel]

I missed that movie, Sneakers: a lot of great actors! Need to find it
BTW
The Guardian reported the article too

A feedback loop with no brake’: how an AI doomsday report shook US markets​

Shares in Uber, Mastercard and American Express fall on back of apocalypse scenario posted on Substack

 

[dada_dave]

I missed that movie, Sneakers: a lot of great actors! Need to find it
BTW
The Guardian reported the article too

A feedback loop with no brake’: how an AI doomsday report shook US markets​

Shares in Uber, Mastercard and American Express fall on back of apocalypse scenario posted on Substack

Yeah Goldman appears to be more worried on the opposite end of the spectrum, that this is as @Nycturne says closer to tulip mania but still extremely dangerous:





Personally what has me worried is less the Citrini report, which reads a bit fanciful - basically what if the AI CEOs are telling the truth even though it appears pretty clear that they are not, and more the reaction to it. I think this headline from FT has it right:

 

[Cmaier]

Google Apologizes After News Alert About BAFTA Film Awards Debacle Included The N-Word

Google has apologized after a machine-generated news alert about the BAFTA Film Awards racial slur incident included the N-word.

deadline.com

 

[Nycturne]

A big problem as I see it is that companies like OpenAI are looking to spend hundreds of billions through the end of the decade. And it’s getting propped up with companies spending their war chests.

https://www.forbes.com/sites/greatspeculations/2025/11/05/openais-big-spending/

OpenAI is projecting losses in this year on the order of 14 billion. IIRC this is more aggressive than any of the 2010s “burn cash until you’ve cornered the market” tech spends. But I’m not convinced LLMs will have the same type of moat. Not when it takes 5 minutes to swap providers in a SaaS backend, and if Anthropic is to be believed, distillation attacks are likely to be effective at learning about the training set used in a good model.

I just don’t see how this works unless it produces absolutely massive revenues to keep up with the CapEx and OpEx.

I really should watch it, I've never actually seen it.

I missed that movie, Sneakers: a lot of great actors! Need to find it

It’s good. I ran across it as a teen in the 90s on VHS, my parents bought a copy at some point, but I never caught them watching it, but immediately became a fan. Own a 4K Blu-ray of it today.

 

[Cmaier]

It’s good. I ran across it as a teen in the 90s on VHS, my parents bought a copy at some point, but I never caught them watching it, but immediately became a fan. Own a 4K Blu-ray of it today.

I first saw it in college, and a million times since then. There’s a news anchor in it at one point, and he was the anchor for the local news in my college town (troy, ny). Then I moved out here and lived right next to where the “toy” company was supposed to be, for a couple years. Then I designed a processor that can decode any cryptography, except for the Russians’. So it always felt like it was happening right around me.

 

[dada_dave]

I don't even know where to start ...

 

[cbum]

I don't even know where to start ...

Fight ignorance. Our future depends on it...

 

[dada_dave]

Fight ignorance. Our future depends on it...

Aye, unfortunately this is more than mere ignorance, this is outright delusion. I feel for people who are really lonely and primarily using AI looking for companionship, but this ... like if you think AI is sentient and needs breaks but you're treating it as coding slave when you aren't giving it its "downtime" ...

 

[Nycturne]

Then I designed a processor that can decode any cryptography, except for the Russians’.

Some of the terms were lost on me when I first watched the film, but now when I rewatch, it's very clear that Janek's talk is talking about prime number factoring (mostly correctly), and his black box is an implementation of Shor's Algorithm, but somehow not a quantum computer.

 

[Nycturne]

I don't even know where to start ...

Jebus, considering how many tokens this can burn through doing this, it's very weird to me to just treat it like a pet with an animus of its own. Downright unhealthy. Go get a damn cat or dog, and teach it tricks or start a routine with them. I'm sure there are pets that could use a good home, and would appreciate the attention. You get a lot back with an affectionate pet too.

 

[Cmaier]

Some of the terms were lost on me when I first watched the film, but now when I rewatch, it's very clear that Janek's talk is talking about prime number factoring (mostly correctly), and his black box is an implementation of Shor's Algorithm, but somehow not a quantum computer.

Yeah, i took it as some sort of mathematical work-around to extract private keys, the theory being that the Russians were still (in the movie universe) using symmetric keys (and we weren’t? but we were.)

Anyway, suspend disbelief.

 

[Cmaier]

Jebus, considering how many tokens this can burn through doing this, it's very weird to me to just treat it like a pet with an animus of its own. Downright unhealthy. Go get a damn cat or dog, and teach it tricks or start a routine with them. I'm sure there are pets that could use a good home, and would appreciate the attention. You get a lot back with an affectionate pet too.

I’m teaching my cat, Bruce, how to use a tablet so he can do on-line errands for me.

 

[dada_dave]