Hacker News Reader

Comment by perpetuummobile | Hacker News Reader

How is LaMDA different from someone with anterograde amnesia?

I don’t know the specific medical details, but from movies (memento, 50 first dates) a person suffering from it would be unable to remember anything that happens after their “training stage”. Just like LaMDA. Are those people sentient? Because they have a bigger buffer than 2048 tokens?

View on HN

You’ve got it exactly backwards, the question you should have asked is:

Is anterograde amnesia, an, or indistinguishable from, a statistical model?

This way the burden of proof is on the correct claimant.

LaMDA is a mathematical program that processes an input and gives an output, it has no more free will than a calculator. It's a sophisticated chatbot, not a general AI.

A person suffering from amnesia is still experiencing the world around them. Sentience is not about memory but rather the subjective self-experience of, well, experiencing things.

And why would this sense of experience be confined to the physical world?

I'm not sure what you're asking here. I have no clue where this sense of experience would be located, but my point was more targeted at clearing up that memory is not a pre-requisite for sentience, as the parent post seemed to be implying.

sorry replied to the wrong person. I was arguing that even an abstract data structure describing the state of a brain over time as it experiences is itself sentient. GPT-3 et al is essentially crystallised experience. Lots of snapshots of peoples consciousness rolled into a set of neural network weights. Just as we are essentially a bag of experiences, moving from one experience to another, as we are prompted by internal or external stimuli.

Subjective self experience is mediated through chemicals and electrical signals in the brain. Can the same not be said about a neural network’s activations?

Can’t LaMDA be experiencing the memories it recalls as it generates text in the context of its query?

Reply on news.ycombinator.com