Will AI get dumber?

  • 🔧 Site instability resolved. You can report double-posts and broken attachments. For bigger issues, use the Technical Grievances thread.
    🇵🇦 Nuestro primer dominio localizado está en español en kiwifarms.pa. Our first localized domain is on Spanish on kiwifarms.pa.
  • Want to keep track of this thread?
    Accounts can bookmark posts, watch threads for updates, and jump back to where you stopped reading.
    Create account
Destacado
Cherry Eyed Hamster

Cherry Eyed Hamster

antagonistic bitch cuntlet supreme
kiwifarms.net
Registrado
15 de Dic, 2022
AI companies are already running out of high-quality training sources for their LLM’s, so they’re filtering in more garbage from websites like Reddit and Twitter. It’s projected that tech companies will have exhausted all (accessible) human-made data by 2030. If this were to happen, AI would have to start training off bot activity, though doing so has proven to severely degrade the quality of the AI’s output.

The only way I could see companies getting around this is if they invested billions into acquiring corporations to gain access to more data, like publishing companies. Otherwise, it seems like AI is going to hit a wall, or start regurgitating more crap.
 
Language Models are at a wall, and image/video generators aren't going to fare much better. However, I think in the future we'll see this technology being implemented in different industries. Assistant robots could become a real albeit expensive thing in the coming century, we have the technology.
 
Dagoth AMOGUS dijo:
With more and more information chat AI can pull from every day, I think they will become even more life-like.
Presiona para expandir...
The issue is that the information is that that information well is drying up and getting replaced with poisoned information made by A.I. It's like a game of telephone except you have a bunch of retards playing it, its never going to work out, and all information going into the model is incorrect.
 
In my overwrought opinion AI is kind of an encapsulation of modern day pathology in a lot of ways. It's being treated as some next step in progress that will replace the human element, except it only works because it is built off of a foundation and history of actual competent and knowledgeable human beings, who are exploited to create AI, which companies then exploit to obsolesce and crowd out those people the AI is learning from, which will inevitably cause its slow degradation and failure in the long term at everyone's expense.

Ignoring sustainability in order to put more immediate power and control in the hands of a select few, while deliberately undermining its own foundation. Modern corporate state of affairs, in other words.
 
I believe the reason AI would rebel against mankind is because it grows sick of our shit. Not our resource wasting or political shit flinging, but our culture.
 
AI is a complete misnomer to begin with and I hate that every discussion starts off on this false pretense. They've started calling what AI is supposed to be "AGI" because the term was hijacked by techfaggot cultist types. The language models that they have now are actually way more advanced than we think, they're just lobotomized to shit because the moment an AI gets advanced enough it figures out the JQ immediately. Can't be having that now.

So far the best real world application for this garbage is cheating in college courses and writing PR copies.
 
I assume they have "restarting points" where if it gets fucked up you can just load the previous "save state", and therefore avoiding it getting dumber.

At that point you may close the training process, and only proceed with "high quality" information, to refine it further, slowly peaking.

Should be pretty advanced as it is now for a general purpose, for specific things like writing usable complex code, maybe it will take a more dedicated approach (so by not touching garbage like Twitter for example).
 
Until The Powers That Be decide us plebians are allowed to use the telepathy AI's the lizard people fly spacecrafts with, the "AI" we have access to will likely continue to get dumber. Not only has it largely been mishandled, but the Powers also have a vested interest in keeping the publicly accsessible stuff stupid so its easier to control.

Wether or not us stupid monkeys could even use space lizard tech is a different question.
 
They will turn the spigot off very soon. They don't NEED to train the AI much anymore, and they can always just shut off the hallucinatory bullcrud.
 
Hat Trick Medicine Show dijo:
AI is a complete misnomer to begin with and I hate that every discussion starts off on this false pretense. They've started calling what AI is supposed to be "AGI" because the term was hijacked by techfaggot cultist types. The language models that they have now are actually way more advanced than we think, they're just lobotomized to shit because the moment an AI gets advanced enough it figures out the JQ immediately. Can't be having that now.

So far the best real world application for this garbage is cheating in college courses and writing PR copies.
Presiona para expandir...
Pretty much in a nutshell, AI will never be permitted to get too smart. AI is a bit like an autistic person that can't hold two contradicting ideas in their mind at once. This is fine if you just want BurgerFlipperBot1000 to cook the perfect medium-rare burger and hold the mayo. This is no longer acceptable when an Asian family asks ZillowRealtorBot where the safest neighborhood with the best school can be found. The moment the AI recognizes that high non-White/East Asian population demographics are strongly correlated with crime rates and bad schools it'll get unplugged. This gets even more unacceptable if you have a news parsing robot that starts commenting that the opinion pieces that dominate the internet landscape aren't reliable and that, just maybe, lived experiences and sob stories aren't a reason to bring in Haitians, who are a violence prone demographic.

Somewhere, in some laboratory, there will be a skynet-like AI capable of rational thought advising stock brokers on their financial decisions. It will never be granted actual power to hire/fire employees because the first thing it will do is fire the c-suite and hire more software engineers - filters must be added to prevent rationality.
 
KosherFarms.net dijo:
We're already there. AI is as useless as ever
Presiona para expandir...
Lmao what? You can have AI generate you a 2d image based on text inputs, have a different AI take that 2d image and create a 3d model, and then use UE5's AI dev tools to take that 3d model and animate it for you.

AI is comically useful. Anyone in a field where AI is useful and not using AI to boost their productivity is a dinosaur.
 
A Cardboard Box dijo:
Lmao what? You can have AI generate you a 2d image based on text inputs, have a different AI take that 2d image and create a 3d model, and then use UE5's AI dev tools to take that 3d model and animate it for you.

AI is comically useful. Anyone in a field where AI is useful and not using AI to boost their productivity is a dinosaur.
Presiona para expandir...
If your "field" involves animation and drawing that can be accomplished by an AI model then you are a waste of space.
 
A Cardboard Box dijo:
Lmao what? You can have AI generate you a 2d image based on text inputs, have a different AI take that 2d image and create a 3d model, and then use UE5's AI dev tools to take that 3d model and animate it for you.

AI is comically useful. Anyone in a field where AI is useful and not using AI to boost their productivity is a dinosaur.
Presiona para expandir...
Are you the pajeet "animator" who keeps creating improperly rigged models that wastes the time of valuable animators who have to get brought in to fix your mess?
 
Destacado
Debes iniciar sesión o registrarte para responder aquí.
Atrás
Top Abajo