Palo Alto is a (mostly) wealthy suburban town next to Stanford U. Without traffic it is 30-40 minutes from SF -- twice as long when traffic is bad. There's a train that is lame by European standards but fairly good for the US, which takes around an hour, sometimes as little as 48 minutes.
This talk is about how generative large language models like ChatGPT are a new (since 2017) kind of AI, and they've turned out to develop surprising emergent capabilities like developing a theory of mind (currently equivalent to a 9-year-old human) without anyone noticing for a couple of years. This is a transformative technology, and developers need to look at the new responsibilities that come with it or we will watch it destroy our reality.
One precondition for sentience would be some kind of continuity. Does the AI operate separately from its tasks, or does it instantiate, take care of some request, and then end, as a ChatBot process does? Is it actively processing while waiting for input, or is it only alive and active when it is processing some input and preparing some output?
I think the lack of this continuity is why people who feel they know how ChatGPT works feel confident it is not sentient. Sentience isn't simply responding to stimuli -- although we don't really understand consciousness, it clearly involves some kind of continuing sense of self, some kind of "dial tone" that provides the basic sense of continuing existence, and that isn't there in a bot that simply takes input and builds a response based on rules and data.
The reminds me of an incident that made me (kind of unfairly) boycott Bed Bath and Beyond for the last few years -- I lost my wallet and didn't have credit cards for a bit, and when I tried to pay at BB&B with a check, they used a machine that told them I was too much of a risk, though I had an 800+ credit rating and thousands in my checking account. I complained and researched and eventually realized that basically all retailers use this check verification system that uses AI that considers your checks too great a risk if you don't normally pay by check -- so the idea that you can use a check in an emergency is really an idea from the last century. This bad decision by eBay seems like the same sort of thing -- machine learning teaches the AI that people who don't regularly buy and sell are far more risky than people who do (which of course they are), so new or occasional buyers/sellers look like a huge risk, and get randomly bounced.
I think the reality is that a lot of people really die from an overdose of morphine at the end. Once you are in hospice, the advice is to give enough to make the person comfortable, and eventually that leads to death.
Neal Stephenson's "The Diamond Age" has a fake AI "character" -- a book that appears to be AI but is really operated by low-paid humans in 3rd world countries (reminiscent of the Mechanical Turk, a supposed chess playing automaton that actually contained a small person).
There's an old observation from Arthur C. Clarke, that sufficiently advanced technology seems like magic. One thing that learning how magic tricks are performed taught me is that magicians typically do their fake magic by doing an unreasonable amount of work behind the scenes -- "magic" in the real world is often just doing a large amount of work that people don't realize is happening.
Given all that it seems appropriate that the new "real world magic" -- ML systems imitating intelligence -- really rest on a lot of hidden work by human beings. Just like magical devices like iPhones exist due to a lot of surprisingly cheap labor. Imagining otherwise is like imagining that the delicious food from a 3-star kitchen just appears from the chef's mind, without the help of all of the low-paid kitchen workers, farm workers, etc. that in reality do most of the work.