[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
Settings Mobile Home
/g/ - Technology


Thread archived.
You cannot reply anymore.


[Advertise on 4chan]


File: i[1].jpg (53 KB, 800x800)
53 KB JPG
I've never seen an actually serious discussion on AI capabilities and potential. How are most people so fucking useless and retarded?

It's all either
>damn!! AI just broke out of containment and hacked the government!!
>AI just found the cure to cancer!!
>all jobs are about to be replaced!

Or
>it's not taking my job because tokens are expensive (do you think they'll be expensive for long you idiot?)
>it just predicts the next token!

Despite the hype, I'm pretty sure I literally never saw people actually stop to think and reasonably discuss future potential of AI based on how the technology works and what it could potentially become over the next few decades. Is everyone fucking retarded or what?
>>
Lot of people have severely different use cases. Not everyone is trying to train their own model, solve math or logging logic flaws and actively trying to prevent them from showing up again
>>
>>109670505
You don't hear the reasonable opinions because it doesn't get (You)'s.
>>
>>109670505
From what I can gauge, there is a lot of room to improve locality of token generation, and thus improve in turn the chance of a prediction that a "next token" will be in the area you expect and thus allow for prefetching tokens from pcie and reusing the ones available in vram enough so the hit isnt felt that much for a non poweruser. Couple it with apllying weight masks to a model for specific tasks, hopefully allowing for less lossy quanization, together with a request classifier to invoke the right model each time, and we may be able to run pretty good models on consumer cards
>>
>>109670535
So? Who cares.

My point is, AI is relevant, there's lots of hype around it, many people have opinions on their potential, but none of those opinions are based on real reasoning.

>>109670542
Makes sense
>>
I use the free version of Claude to do my excel work, it's really good and quick at it.
Also wrote several letters, let it make a list populated from my company website, which it could, and now I am thinking of setting up a webcam to let Claude see the world in it's off-time
I sell fish in retail, I show my seniors how to ai but they are to dumb to try
I also do not give a single shit about sharing company data since they pay me so, so well for all that extra work.
>>
>>109670505
Sam Harris does a lot of AI podcasts. #469 isn't paywalled. Most of them are 20 minute previews but it's better than nothing.



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.