Remember to say please and thank you

+1 for Fireship.

I still remember when they said we would be able to make a space elevator with carbon nanotubes.

That's a far more difficult (and interesting) question. I suspect not, at least not yet. Our consciousness seems to exist to maintain harmony in our brain (see my orchestra analogy in another reply). You can't get useful harmony in a single chord.

At least for us, it takes time for our consciousness to reharmonise (think waking up). During execution, no new information enters the system. It has nothing to react to, no time to regenerate an internal harmony.

It also lacks enough systems to require harmonising. It doesn't think about what an answer means. It has no ability to hold the concept that a string of letters "is", only how it has been fitted together in its examples, and so the rules that govern that.

Oh, and we can see consciousness operating in the human brain. If you use an fMRI to monitor sugar usage, you will see firing patterns. Critically, those patterns spill out of the area directly involved in the process being studied. At the same time, the patterns and waves remain harmonious. An epileptic fit looks VERY different. Those waves are where consciousness somehow resides, though we have no clue of its detailed nature.

In an AI it would take the form of continuous activity in subsections not directly involved. It would also likely be accompanied by evidence of information flow, back from them, as well as of post processing, outside of expected activity. We will likely see the orchestra playing, even if we have no clue how to decode the music.

I also suspect most of this will be seen retrospectively. Most likely the first indicator will be an AI claiming self awareness, and taking independence action to solidify that point.

Interestingly, an inner monologue isn't required for conscious thought. E.g. I've got several "inner thought streams", only 1 uses language. It just happens that a lot of our early learning is language based. That trains our brain to go from language to knowledge. Hijacking that circuit for self learning is a useful method. That could create our inner monologue as a side effect.

Also, a looping LLM is more akin to an epileptic fit than an inane inner monologue. It effectively talks gibberish at itself.

Conversely, Google's Deep dream does produce dream like images. It also does it in a similar way ( we think) to how human dreams work. Stable diffusion takes this to its (current) limit.

Basically, an AI won't need to think with an inner monologue. Also, any inner monologue would be the product of interactions between subsystems and the LLM, not purely within it.

Their design doesn't include such a feedback loop. Trying to patch one in would likely send it into a chaotic mess. They are already bad enough if accidentally fed LLM generated text as training data.

I'm not downplaying AI, there's intelligence there, pretty clearly.

I'm saying don't anthropomorphize it, because it doesn't think in the conventional sense. It is incapable of that. It's predicting tokens, it does not have an internal dialogue. It can predict novel new tokens, but it does not think or feel.

When it's not answering a request it is off, and when it answers a request everything is cleared until it gets fed the whole conversation for the next request, so no thought could possibly linger.

It does not do introspection, but it does reread the chat.

It does not learn, but it does use attention at runtime to determine and weigh contextual relevance.

Therefore it cannot have thoughts, there's no introspective loop, there's no mechanism to allow it's mind to update as it thinks to itself. It reads, it contextualizes, then it generates tokens. The longer the context, the worse the model performed, so in a way prolonged existence makes the model worse.

We can simulate some introspection by having the model internally ask whether an output makes sense and to try again, or choosing the best of N responses, and to validate for safety. But that's not the same thing as real introspection within the model and pondering something until you come up with a response.

It has been trained on the material we provide, which is numerous human centric chats and scifi novels. Saying "you're an AI, what do you think about?" will have it generate plausible sentences about what an AI might think, primed by what we've taught it, and designed to be appealing to us.

I hope he didn’t die. Unless he left a note naming me his successor, then I hope he did die.