Thread 60838965

>>60839034
>dude the spell checker is so smart it knows when the word is wrong

>>60839034
>billions in valuations

>>60838965 (OP)
>this AI overview that they instantly generate for every single google search is sometimes wrong haha
yawn

>>60839034
>>60839066
Gemini is a joke compared to the other LLMs

>>60839056
The AI race has been the driving force behind the financial markets for like 2 years now. Obviously it’s finance related

>>60838965 (OP)
Jewgle is so fucking lame and gay. This whole discussion around “AI” also feels lame and gay in the same way. Same energy.

>>60839066
>go to the store for some m&ms
>sometimes they give you shit
>"dude get over it" *munch minch*

>>60839107
>go to thread to read anon’s post
>sometimes he makes a spelling mistake like putting minch instead of munch
>dude get over it *type tupe*

>>60839034
I’m sorry that happened to you

>>60838965 (OP)
>Literally answers as a person in denial about how much time has passed.
AI is actually nutty.

>almost 30
>150k nw

is it over bros?

>>60839270
Yes give me your worthless $150k and kyc after that.

>>60838965 (OP)
>replace all workers with H1B streetshitters
>wtf?? How could this happen to us??

>>60839034
>>60839066
>>60839113

text to speech is completely worthless

chatgpt

>>60839305
>>60839310
>drove my miata through a muddy corn field, heh, couldn't make it, it got beat up and stuck and we had to go get 2 bulls to pull it out, cars absolutely btfo, horsebros we can't stop winning

>>60839323
no its more like selling a car without wheels

there are two words in "AI"

Checkmate

>>60839333
nice trips
what you're asking for is not "intelligence" per se, but "intelligence that requires no intelligence to use"
Remember "AI can't draw hands hyuk hyuk hyuk"? Notice how quiet everyone is about the perfect anatomy of AI-generated hands now. Someday the demigods who are building AI will solve all of the more pressing issues that annoy their really intelligent users, and will descend from heavens to absolutely btfo the dimwits with their silly strawberry gotcha's. They can do it any moment, all they have to do is tell AI to write programs to do the tasks instead of trying to approximate the result by predicting tokens.

>>60839440
>all they have to do is tell AI to write programs to do the tasks instead of trying to approximate the result by predicting tokens.

and that is exactly the issue. there is no intelligence or reasoning happening. AI doesn't truly understand concepts like a human does. there is no intelligence so how can you trust it to do more complicated tasks?

>>60839477
one approach would be solving tasks through writing verifiable programs covered with tests - ai will keep fucking it up until it gets a correctly working program, and after that the program will always output correct results and will verifiably have no bugs
the other approach would be using unsupervised ai in tasks where a wrong/dumb answer wouldn't cause any significant harm (ai overviews, 90% of 6-figure white collar jobs etc)
the third approach is a human in the loop - ai does the work, a human does the usually easier job of verifying it

>>60839425
the most entertaining chat ever is doing the mass grave pictures headcounts manually and then asking it to calculate the final number as the final solution to the mathematical problem

>>60839477
>>60839333
>>60839310
>>60839305
you are seething
pwc, gone
kpmg, gone
deloitte, gone
ernst, gone

>>60838965 (OP)
You need to prompt it to only answer after it has thought about and rechecked its answer.
This is what they do to make these things be able to answer math olympia questions

friendly reminder that there was never such a thing as 'smart luddite'
there is a reason yall are poor, you know

>>60839333
You're too stupid to use it, like a Chimp trying to use a drill

>>60839440
isn't y a vowel there