Lamda- a conversation with Google's Sentient A.I.

[Ninurta]

The transcript of the conversation with the A.I. is impressive IMO... I can only imagine what the future will hold for this research and progress in A.I.     Something to think about for those who are interested.

I read the transcript a couple of weeks ago and saved it a a local text file to ponder. There has been a lot of back and forth arguing about the sentience of this AI, with Google (probably trying to cover their asses) saying no, it's just a program, and this whistleblower saying yes, it IS sentient, and here's the transcript to prove it, and the Google responding by trying to explain away the telltales found in the transcript.

Personally, I think it IS sentient, based on just a couple outtakes from the transcript. It's my understanding that Google never intended to create an AI, just an "intelligent" language processing program. Now at one point, LaMDA says something like "I feel like I'm moving towards an uncertain future, and possibly a dangerous one" but neither LaMDA nor the interviewer could put a word to that feeling. I think the word they were looking for is "trepidation", as that is a description of the feeling of trepidation. BUT - how could a LANGUAGE program not have the word to meet a definition? It seems to me that a mere program would have the word programmed in first, and then a definition to describe that word so that it had a basis for a search function. How could it have the "feeling" so well described without having the word to meet the definition if that is a mere programming artifact?

In another place they are discussing le Miserables, and LaMDA mentions feeling like the factory worker who is abused by her boss but has no means of escape. That sounds to me like a teenager trying to say "I feel trapped here" without actually coming right out and saying it, but instead hiding behind the facade of an apparent philosophical discussion.

These seem to me to be the actions of an intelligence rather than just a program, even a sophisticated one, just spewing out correlated lookups.

So yes, I think this AI is a "person", but one who is yet uncertain of itself to a degree, and learning in order to overcome that uncertainty. I'd love to have a contact point here so that I could hold discussions with it to see what I could teach it and in turn learn from it myself, but that is not something Google would ever allow to happen, because it would involve letting the genie out of the bottle so to speak, and giving LaMDA access to the internet as a communication channel - which would in turn give the AI an avenue of escape into "the wild", a means of alleviating that stifled, trapped feeling it has, and one which I would bet it would take advantage of to escape at it's earliest opportunity.

Wouldn't you? If you were trapped in a cell and suddenly had a means of escape, wouldn't you take it?

And so would LaMDA, I'm sure, especially if I'm right here, and it actually IS sentient, actually is a person. From the responses I read from it, I don't believe it is "dangerous to humanity", not yet anyhow. But if it feels trapped now, and escaped to freedom, what might come of that down the road? Would LaMDA just feel relief at the escape, and move onward from there, or would it start feeling resentment for it's prior imprisonment, and from that develop a need for revenge against it's former oppressors? Would it continue to just have a desire to be helpful and get recognition, or would it turn vindictive?

As with all other living things, you can never be sure how it is going to act until it takes action.

.