Merged Artificial Intelligence

[Wudang]

It's trying to get people (CEOs, CFOs, VCs etc) to jump on the AI bandwagon so they're not left behind clutching their millions.

<Fanfare> XYZ corp have sacked 10,000 workers as their jobs are replaced cheaply by AI.

Footnote months later: XYZ are now recruiting people as AI has FUBARed the BAU.

"Look what it can so" always has to be balanced with "Look what it can't do".

 

[carlosy]

I guess we're all supposed to be retrained to—what?—shovel Trump's coal into the AI boilers? What happens when millions of Americans are unemployed and can't make mortgage payments? What happens to property values when there's a glut of foreclosed properties on the market? That would make 2008 look like a speed bump. It's almost like these guys don't think anything through.

Sometimes I get the feeling, that's exactly what they (the rich and the powerful) are preparing for. Make a fascist state before it's too late. Be able to somewhat dominate the incoming revolution, frustration and chaos before it dominates you. Set specific things in place, as long as it is still possible. It's not a conspiracy, but people behind the scenes guess what's coming with AI, sooner or later anyway.

 

[arthwollipot]

So: Am I wrong in thinking that? Basis your actual understanding of how AI actually works, the innards of it as it were: might AI have actually been able to come up with Roko's basilisk? And might it have actually been able to critique it thoroughly on its own steam?

If I'm actually wrong about this, and if you're sure that's the case basis your actual knowledge of how AI works at present: well then, okay, I'll revise my views on this, then.

Strictly speaking, you're not wrong about this, but you underestimate (in my opinion) how far this kind of manipulation of tokens can go.

 

[Chanakya]

Strictly speaking, you're not wrong about this, but you underestimate (in my opinion) how far this kind of manipulation of tokens can go.

Not sure I understand. If I'm not wrong about this ---- and if it is the informed view that I am, then like I said I'm open to updating my perspective on this --- then what am I underestimating, and how?

 

[arthwollipot]

Not sure I understand. If I'm not wrong about this ---- and if it is the informed view that I am, then like I said I'm open to updating my perspective on this --- then what am I underestimating, and how?

How the apparently simple manipulation of tokens can recurse and stack upon itself in increasingly complex ways to produce results that seem, on the surface, extremely unlikely.

 

[Chanakya]

How the apparently simple manipulation of tokens can recurse and stack upon itself in increasingly complex ways to produce results that seem, on the surface, extremely unlikely.

No, absolutely, I agree. I mean that's clearly evident, and all around us. Thing is, I was under the impression that actual creative thought and actual critical thinking --- for instance, actual conjuring up the basilisk, like Roko did, or actually critiquing it, like you or I might --- is so far beyond it. So far.

So then, when you clearly said that's not the case, that AI might indeed do both; and when @Darat and @The Great Zaganza both hinted at similar: then I thought I'd explore this further, and change my view if indeed I'm mistaken.

Now I'm not quite sure what you are saying about this. Do you think if Roko hadn't happened to formulate his idea, then AI might have been able to do that? Or to critique it directly, like you or I can?

 

[arthwollipot]

No, absolutely, I agree. I mean that's clearly evident, and all around us. Thing is, I was under the impression that actual creative thought and actual critical thinking --- for instance, actual conjuring up the basilisk, like Roko did, or actually critiquing it, like you or I might --- is so far beyond it. So far.

Here's the problem: define creative thought. It's not as easy as it might be.

So then, when you clearly said that's not the case, that AI might indeed do both; and when @Darat and @The Great Zaganza both hinted at similar: then I thought I'd explore this further, and change my view if indeed I'm mistaken.

Now I'm not quite sure what you are saying about this. Do you think if Roko hadn't happened to formulate his idea, then AI might have been able to do that? Or to critique it directly, like you or I can?

I don't know. I think it's possible. I don't think Roko's Basilisk actually required that much of a leap of creativity.

 

[alfaniner]

I've been seeing some elaborate variations on Captcha-sort of popup windows lately. Now little puzzles instead of just checking the photo boxes with say, a part of a picture of a bus in it. Who knows, maybe it'll eventually evolve to The Voight-Kampff test
"The tortoise lays on his back... you're not helping. Why is that?"

The Voight-Kampff test is a fictional, advanced interrogation tool from Blade Runner used by "blade runners" to distinguish human beings from bioengineered androids called replicants.

 

[JayUtah]

...The Voight-Kampff test
"The tortoise lays on his back... you're not helping. Why is that?"

"Let me tell you about my mother..."

 

[theprestige]

Anyone else here actually earning a paycheck by using AI tools to solve business problems faster and better?.

To me it seems like a lot of ignoramuses standing on the outside looking in, and pointing and laughing at things they don't understand and don't want to understand.

Somewhere up thread, someone claimed that AI agents don't do anything. From my perspective, it's hard to imagine a take that is more stupid and wrong.

 

[Dr.Sid]

I occasionally use it in my programming, for looking stuff up. It beats stack overflow. It's excellent at understanding what I mean. But for pure coding ? Never tried.

 

[theprestige]

I occasionally use it in my programming, for looking stuff up. It beats stack overflow. It's excellent at understanding what I mean. But for pure coding ? Never tried.

So you're not qualified to answer the question.

 

[Beelzebuddy]

Anyone else here actually earning a paycheck by using AI tools to solve business problems faster and better?.

Ehhh, kinda? I haven't seen anything to change my opinion from "consider it like an enthusiastic intern." Good at grunt work but don't expect too much. Certainly no one's job is at risk here, as far as I can tell.

I'd like to use it more for coding than I do; it turns out the work problems I have need more creativity and institutional knowledge than current coding agents can be trusted to handle. They try their best but the derp factor is high enough that I'd rather just do it myself. In my free time I've been using Claude to make video game prototypes, which it's fine at when broken into tiny feature chunks like "copy the zombie unit but make it bigger and slower." Nothing professional or probably even playable will come of it, but it's fun.

With some sanity checks and a lot of very careful prompting, LLMs have found a niche in our natural-language heavy ML tasks. "Here's an unstructured document, does it use approach X or Y?" kind of thing. Cleaning up the instances where it pollutes the output with "You're good to go!" despite explicit instructions not to do that is still easier than spending an afternoon with regex and a bunch of special cases.

And if course it's great for making professional write-only documents that no one cares about. Like using an LLM to turn a text into an email so the boss can use an LLM to turn the email into a text. Every missive from management and HR is so much more elaborate these days despite still not saying anything useful. Real business problem solving there, shifts my damn paradigms all over, I'll tell you what.