I think it’s just fundamentally flawed to try and create something that “thinks” via a starting point of language.
Language is not the starting point of thought. There are very simple creatures that clearly experience, think and even problem solve at a level beyond that of existing llm “AI”.
The starting point for creating artificial intelligence therefore shouldn’t be “let’s make a talking human” but rather more simple artificial intelligence with the ability to problem solve without training data.
Machine learning via evolution algorithms has always been a more promising pathway to real artificial intelligence. LLMs have always been and can only ever be a facsimile.
AI will be a bunch of different tech working together. LLMs probably will be a part of it. It’s not dumb technology despite what everyone says. It’s not even new technology.
And it is useful. Being able to use natural language to talk to computers is pretty revolutionary and sci-fi. But it’s not AI and it can’t do everything people are trying to sell it as being able to do.
And like… all intelligence had “training data” before solving any issues. The only living being without any “training data” are newborns. And they can’t solve anything by themselves. You have to “train” them for years and years showing the right way of doing things until they finally get it.
I think it’s just fundamentally flawed to try and create something that “thinks” via a starting point of language.
Language is not the starting point of thought. There are very simple creatures that clearly experience, think and even problem solve at a level beyond that of existing llm “AI”.
The starting point for creating artificial intelligence therefore shouldn’t be “let’s make a talking human” but rather more simple artificial intelligence with the ability to problem solve without training data.
Machine learning via evolution algorithms has always been a more promising pathway to real artificial intelligence. LLMs have always been and can only ever be a facsimile.
The people who think LLM’s are sentient tend to be managers… and managers are a facsimile of sentience. Ergo a manager is an LLM
This is what you get when nobody actually reads or understands Turing but bases all their ideas on what they think he said.
AI will be a bunch of different tech working together. LLMs probably will be a part of it. It’s not dumb technology despite what everyone says. It’s not even new technology.
And it is useful. Being able to use natural language to talk to computers is pretty revolutionary and sci-fi. But it’s not AI and it can’t do everything people are trying to sell it as being able to do.
And like… all intelligence had “training data” before solving any issues. The only living being without any “training data” are newborns. And they can’t solve anything by themselves. You have to “train” them for years and years showing the right way of doing things until they finally get it.