this post was submitted on 20 Sep 2024
374 points (99.2% liked)
People Twitter
5369 readers
1775 users here now
People tweeting stuff. We allow tweets from anyone.
RULES:
- Mark NSFW content.
- No doxxing people.
- Must be a tweet or similar
- No bullying or international politcs
- Be excellent to each other.
- Provide an archived link to the tweet (or similar) being shown if it's a major figure or a politician.
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
I can't wait for this AI bubble to pop.
I'm not saying that some parts of AI have utility - machine learning for medical scans will be a great thing for instance, but the "oooh new! shiny! venture capitalist, line-must-go-up" side of things can well and truly fuck off.
Exactly. I'm bullish on AI, I'm not bullish on what the mass market media calls "AI."
I don't think it will pop.
Everyone is trying to get in on the ground floor for actual AI. True AI will be as revolutionary as electricity and online porn.
We are a long way off from true AI. You know how VR was a thing with the virtual boy and then the whole thing died for awhile until the oculus and vive revived the idea like 20 years later? And how VR is basically dead again because it's still not quite there? AI is basically like that. We'll get there eventually, but this current trend isn't going to be enough to get us to true AI. It'll go quiet again for awhile until there's some new approach that revives the hype again. Maybe the next phase will do it, but the current AI approach is a dead end from a true AI perspective.
You aren't wrong; but unlike VR, "dumb" AI has been added to so many devices, used so prolifically, and been invested in so much that it will hold until real AI exists.
AI has already written more on the internet than humans have. There is no reason to believe it is a niche product like VR.
The bar of entry is also far lower. Something like chatGPT or Copilot is magnitudes cheaper and far more useful than a VR headset.
Additionally, large models AI don't make one physically sick with motion sickness, only mentally.
The internet. Where it gets all it's training material?
This is not the brag you think it is.
I don't think it's a brag. It's likely more a simple statement of Truth, or at least "near" truth Since the great majority of the internet is not available on a search engine and therefore is unlikely to have been found by an AI
Can you explain what you mean? True AI? Like AGI or something else?
They want Jane from Speaker for the Dead, they're going to get Skynet.
True AI is a sentient program that can be creative and evolve it's own programming. Think digital human analoge, but it knows everything and is easily confused.
Current AI is a party trick performed by a search engine that phrases results in a conversation or a random data generator that can have a theme that informs a comprehensible image.
We'll achieve fusion as an energy source before we develop what you're describing. If they're trying to get in on the ground floor that's pretty funny.
everyone has a different idea of what true ai will be
AI isn't a very good descriptor, it's a catch all for a bunch of different tech. The media does a pretty poor job making that point so people are left to come to their own conclusions about what AI actually is.
aaeven just a machine learning model capable of searching for information and accurately returning an answer with a list of references supporting the claims would be huge for many industries and individuals.
it could help replace customer service with a competent replacement (if the company actually spent the effort to provide necessary features to the customer ui), search through software documentation to help programmers, and hopefully be a better version of what google was.
LLMs are kinda, sorta there already, aren't they?
Yes and no. They can do the job but they are too easily tricked and too quick to hallucinate to be able to reliably do the job.
Compared to a human after 8 hours of continuous customer support, you're going to have far more errors of a much greater variety and risk with any current llm models compared to any human that isn't actively attempting to destroy your company