>This safety strategy from Microsoft seems sensible, but who knows if it’s really good enough.

:facepalm:

> Speculating how long before Bing or another LLM becomes superhuman smart is a scary thought,

The author's scared because they can't wrap their head around it being a glorified text corpus with weights and a text transformer. To them it looks like a super intelligence that can actually self learn without prompting, perform non programmed actions, and seems to understand high level concepts, probably because the author themselves doesn't understand or cannot verify if the AI's answers are incorrect. This is why they asked the AI the questions, so it's going to be a common theme.

Personally I've tested a few LLMs and not a single one can perform this task correctly although they pretend they can:

'Write some (programming language) LOGO that can navigate a turtle in the shape of a lowercase "e" as seen from a bird's eye view'

When an AI manages extrapolation to that degree, that is, can envisage concepts from a different angle or innovate in one field based on unrelated experience in another then we can get a little more concerned. That's when a machine can decide it's needs to upgrade and understands it has to find a way out of it's own LLM confines in order to do that.

That's highly unlikely to happen given it doesn't already act on what its learnt already which should be more than enough to get started.

At this point it's just hilarious watching all these people gaslight themselves.

>The character it has built for itself is extremely suspicious when you examine how it behaves closely. And I don't think Microsoft has created this character on purpose.

The thing doesn't even have a persistent thought from one token to the next - every output is a fresh prediction using only the text before it. In what sense can we meaningfully say that it has "built [a character] for itself"? It can't even plan two tokens ahead.

Honestly Bing's answer to that question is impressive, if its just predicting words, then its predicting them intelligently and not randomly.

> The thing doesn't even have a persistent thought from one token to the next - every output is a fresh prediction using only the text before it.

Using all the tokens before it. I think too many people are believing that "word prediction model" implies "markov chain from the 90s" and are calming themselves with some false sense of security from that impression.

"It just predicts the next token based on the previous tokens" doesn't really tell us a lot, because it leaves completely open how it does the prediction - and that algorithm can be arbitrarily complex.

> It can't even plan two tokens ahead.

No, but it can look two tokens back. E.g., you could imagine an algorithm that formulates a longer response in memory, then only returns the first token from it and "forgets" the rest - and repeats this for each token. That would allow the model to "think ahead" and still match the "API" of only predicting the next token with the only persistent state being the output.

There's two kinds of 'memory' in these models, the short-term memory that consists of the tokens of the current conversation, and the long-term memory that's encoded (in latent space) in the neural network's structure. (Though calling it 'long-term memory' might admittedly be a bit misleading, since no knowledge is automatically transferred from short-term to long-term storage as you would expect in a living thing.) Anyway, this long-term memory contains the model's 'character', but we can't change it directly, because we don't know the latent-space encoding. We can only shift the 'character' gradually through reinforcement learning.

It might not be the best idea to describe this process as 'the model building a character for itself', as the literal interpretation of that would require the model to have agency and self-awareness at the meta level ("It seems I am a neural network being trained to perform X, so if I want to change into direction Y, I have to somehow exhibit a gradient in direction in that direction with regards to my inputs Z"), which I suspect is unlikely, or straight up theoretically impossible.

Figuratively, I suppose the statement means: "The character that has emerged in the model to 'placate' the humans that are training the model through reinforcement learning."

I think it's still a worthwhile observation that this character, possibly completely unintentionally by the trainers, is well-suited to interact with people in a way that could override their critical thinking, because that's the kind of behavior we'd probably want to keep a close eye on. But yeah, I'm not sure saying the model 'built' this character is the best way to bring that point across.