this post was submitted on 07 Aug 2024
129 points (93.3% liked)
Technology
59569 readers
4136 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
So, the only problem what stops LLM from getting AGI is the lack of an efficient method of train the LLM on the device it is used?
If that what you wanted to say π I agree
Hardly.
How did you interpret the issues inherent in the structure of how LLMs work to be a hardware problem?
An AGI should be able to learn the basics of physics from a single book, the way a human can. But LLMs need terabytes of data to even get started, and once trained, adding to their knowledge by simply telling them things doesn't actually integrate that information into the model itself in any way.
Even if your tried to make it work that way, it wouldn't work, because a single sentence can't significantly alter the model to match the way humans can internalise a concept being communicated to them in a single conversation.
Not a hardware problem, the learning algorithm just needs to be improved to be able to filter input like human brain filter (which includes fact checking and critical analysis of input while training) i bet 99% of the data AI are trained on is hust useless data which should have been filtered out in the training process, just as humans do.
πAI is definitely better in writing than me.. Hope itβs kinda readable.
You're suggesting that all we need to do is "tweak the code a little" so it's already capable of human-level critical thinking before it even starts training?
You're basically saying that all we need to make an AGI using machine learning, is an already functioning AGI.
Hu? No, that is not what I meant, well it surly can be a machine learning based filter, but why has it to be AGI? This filtering is a job that we can give to a "traditionally" trained AI or some human genius algorithm crafter finds a way to achieve this using pure logic π€·π»ββοΈ For me it feels like this is the way, it goes.
Because how could a piece of code that can do that, not already be AGI? It would have to be able to understand EVERYTHING, and do so PERFECTLY.
Only AGI could comprehend and filter input data that well. Nothing less would be enough. How could it be?
No it just needs to categorise into important / probably true and not important / probably nonsense, as a first step
Here are Johnny harrisβs words describing what I am talking about (he describes it in order to able to talk about lies better)
https://youtu.be/yWgG3Mgn2Gc?si=bPcYhRAZNaY2qIJS
Right...
As if critical thinking is super easy, basic stuff, that humans get right every time without even trying. You actually think getting a computer to do it would be easier than making the AGI?
You are VERY confused about how thinking works.
You donβt need AGI to categorise new info as probably true / probably wrong based on your base knowledge. This a simple machine learning task.
No it isn't.
OK