A Futurism article about "AI bros are having a meltdown because the Associated Press stylebook clarifies that AIs don’t have feelings" featured this guy's tweets in the first sentence. Turns out he was talking about dogs instead of AI.
For some reason I can't stop laughing about this.
Every neural network, including those in your own head, have sensory neurons that allow it to take in information from the environment. It can’t process anything without them. LLMs are trained primarily on symbols, but there is no reason you cannot attach these sensory neurons directly to things like cameras or tactical motor inputs, which many of times we do.
By this way of thinking, computers have been "taking in information from the environment" for decades, like when you type some text in a text editor. It's greatly stretching the meaning of what sensation and sensory perception is.
Modern LLMs are obviously self-aware by the very definition of the word.
By what definition? As far as I can tell, you didn't provide one. What I see is a couple of anecdotes on trying to trick a chatbot. Chatbots themselves aren't even an inherent form of an LLM (text completion/continuation is). If you get rid of the mechanisms that make an LLM stop before it responds for the user, it can easily continue the text as them. It doesn't actually know there is an it and there is a you. Much of what people take as traits of cleverness that are inherent to an LLM is a combination of the LLM itself and clever UI design / prompting by humans.
People primarily learning what LLMs are through chatbot harnesses like ChatGPT is honestly one of the worst things that could have happened to the field. It has truly damaged literacy about the tech and created a lot of wrong assumptions about what it is.
Indeed, you can take a screenshot of the conversation you are having with an LLM and feed it back into it, and ask it what that is, and it will recognize not only that it is a screenshot of a conversation, but a screenshot of this conversation that it is having with you right now.
Yes, it can label things. You can also (long before now) create a computer program that looks through a list of numbers and labels which are odd and which are even. You can create algorithms that do complex computation. Does this mean the algorithms are aware? Or is it only seeming like awareness now because it can simulate language?
Although, I probably know your response, you will just keep suggesting that it’s all some illusion or trickery of self-awareness and it’s not actually there, without justifying it either than explaining how it works by saying it’s “all linear algebra on GPUs” or something. So I don’t really need a response as I already know what you’re going to say.
I don't think you understand why I am so staunch about this. It's not that I think computation could never produce anything that qualifies as real intelligence, as similar to organic beings. It's more that basing the argument on how convincing a computer is at talking like a human is like building a machine called the Fool Me Machine and then getting fooled by it and saying that people who point out it's a Fool Me Machine are the problem.
It has to be investigated on a more foundation level than that. That's why I bring up the hardware. People need to see past the surface level appearances.