Investing General Discussion

  • Guest, it's time once again for the massively important and exciting FoH Asshat Tournament!



    Go here and give us your nominations!
    Who's been the biggest Asshat in the last year? Give us your worst ones!
Captain Suave

Captain Suave

Caesar si viveret, ad remum dareris.
5,561
9,578
Sanrith Descartes said:
[Grok:]

LLMs aren’t perfect. They can hallucinate facts
...
Google and Bing thrive on the free, ad-driven model—billions of users, monetized through targeted ads tied to search and browsing habits. It’s a machine that’s worked for 20+ years, pulling in $224 billion for Google alone in 2023.
Click to expand...

Well, at least it's honest. What's $14 billion here and there between friends?

1742399782351.png

(yes, I checked against Alphabet's earnings statements)

I don't see how anyone can trust these things for any purpose involving numbers.
 
Last edited:
  • 2Like
Reactions: 1 users
Sanrith Descartes

Sanrith Descartes

You have insufficient privileges to reply here.
<Gold Donor>
46,317
129,702
Captain Suave said:
Well, at least it's honest. What's $14 billion here and there between friends?

View attachment 578547
(yes, I checked against Alphabet's earnings statements)

I don't see how anyone can trust these things for any purpose involving numbers.
Click to expand...
What I dont get is its programmed. How hard is it to program it to say "I dont have the answer to that math/financial question". I should ask it why.
 
Flobee

Flobee

Vyemm Raider
2,761
3,211
Gen-X's version of believing everything on Fox News is going to be blindly relying on AI for answers. Its a useful tool in a number of circumstances but citing it as factual is foolish. Especially so when one pastes its responses in an argument lol.

More on topic I suspect market bounces soonish purely based on global liquidity uptick.
 
Arden

Arden

Blackwing Lair Raider
2,841
2,135
Flobee said:
Gen-X's version of believing everything on Fox News is going to be blindly relying on AI for answers. Its a useful tool in a number of circumstances but citing it as factual is foolish. Especially so when one pastes its responses in an argument lol.

More on topic I suspect market bounces soonish purely based on global liquidity uptick.
Click to expand...

Not sure anyone is suggesting that you should blindly rely on AI. You shouldn't blindy rely on anything- AI, TV, News, radio, books, conversations with people on the Internet.

AI is just a tool, and it happens to be really good (and fast) at aggregation info sources on the net and proving you with a succinct answer to your query.

If your AI is good, it will use typically reliable sources and provide you with citations so you can back check the facts it provides.

Being skeptical of ANY information source is wise these days, but automatically dismissing new tech just because it's new just makes you seem like a luddite.
 
  • 2Like
Reactions: 1 users
Captain Suave

Captain Suave

Caesar si viveret, ad remum dareris.
5,561
9,578
Sanrith Descartes said:
What I dont get is its programmed. How hard is it to program it to say "I dont have the answer to that math/financial question". I should ask it why.
Click to expand...

My wife runs an AI department (though not building LLMs). She says the correct solution is to identify specific factual elements of output and check those against a database of sourced links derived from traditional search, or refer to rigorous calculation engines a la Wolfram Alpha. This are already such products available in enterprise settings, within narrow domains. It's not part of general purpose public-facing ChatGPT/Grok/etc. for reasons of cost, complexity, and scale of engineering. They don't plaster caveats everywhere because it sells better. (I think Perplexity is working in this direction. I haven't tried it.)

I still think it's a major product design fail to have released in this state without guardrails. It's going to badly degrade future confidence even when precise implementations are eventually available. It'll be like asking people to trust self-driving cars after they've been marketed as functional and yet predictably crash. "Trust us this time, really!"
 
  • 1Like
Reactions: 1 user
Tmac

Tmac

Adventurer
<Aristocrat╭ರ_•́>
10,217
17,785
Sanrith Descartes said:
What I dont get is it’s programmed. How hard is it to program it to say "I dont have the answer to that math/financial question". I should ask it why.
Click to expand...

That’s not how they work. They don’t actually know how to tell it to do that so directly. Their only option is to train bias into it.

There’s entire departments at universities that are studying how LLM’s work trying to reverse engineer their internals so they can tell it stuff.

I have a buddy who’s getting his PhD in this stuff and is on the same floor as the British equivalent of OpenAI. He talks about LLM’s in scary ways.
 
Last edited:
  • 1Like
Reactions: 1 user
Blazin

Blazin

Creative Title
<Nazi Janitors>
7,320
37,033
Sanrith Descartes said:
Lets see if the S&P can bust through the 200.
Click to expand...
not ready for that, need some calming consolidation below before the move up. Does it too soon and it will just get smacked down. End of week is what matters. We may get the close at the 200d and wake up monday to either break out or break down.

Anything possible, gamma flip just up overhead. If it does break above it we would get some crazy breadth thrusts multiple 80% up days , I just don't see how declining 20d and the 200d aren't going to act as resistance for a bit.
 
  • 1Like
Reactions: 1 user
Jysin

Jysin

Ahn'Qiraj Raider
6,626
4,647
Largest OpEx in history coming on Friday. I would wait for Monday to really get a sense of direction.
 
  • 1Like
Reactions: 1 user
You must log in or register to reply here.