When I was a teenager, I longed for the day we would have intellegent computers, where you could ask questions and get answers, and even have a conversation. For decades, it seemed impossibley far away. I remember "pong" in the 70's; BORING. Then the TIMEX Sinclair in the 80's. BORING. Then finally, the Commodore 64. Interesting... while it lasted.
Then came the TRS-80, Tandy Model 100... useful, if primative. Then various DOS based PCs, that continuously evolved. Then the internet. Alexa was interesting, but not very smart. Various chat-bots could fake being intelligent for a bit, but would ultimately disappoint.
When Chatgpt came along, I ignored it, thinking just another mediocore chat-bot. But something was different this time. There has been a game change. Suddenly, it's gettting really good. Suddenly, AI has become conversational.
I've been using ChatGPT for a while now. It can organize data and reports and a variety of things, creating reports and reference books for me... in seconds. I could do what it's doing myself, but it would take weeks or months.
The conversational Star Trek computer is finally here! Shouldn't I be thrilled? Well, yes, and no. Because now all that Sci-fi stuff, about AI becoming dangerous and taking over, now has to be taken seriously. And now AI is learing, and learning quickly. So quickly that most people aren't even aware of how quickly this is going to change so many, many things.
In this video, Geoffrey Hinton has a lot of important things to say, and makes many well-considered points. Some of it I'd head before; other parts are completely new to me.
One thing he talks about at one point, really burst the bubble I had about a concept I've held for a long time. I've always beleived that AI was just mimicing human behavior and intelligence; that there was ultimately no "there" there. It was just a bunch of algorithems mimicing intelligence and feeling, without the ability to actually really "feel" any emotion. But what if that presuppositon is wrong?
Geoffrey addresses this. He explains that while AI is unable to experience emotions the way we do, feeling them in our bodies, we need to remember that we also learn emotions, from each other and from our experiences in life. And since AI is a learning intelligence, growing and expanding it's knowledge, it can also "learn" emotional responses.
He used an example of a call center. AI is thought to be perfect for replacing humans in a call center. But when humans are trained in a call center, they are trained to become impatient with people who are lonley and just want to chat with someone, instead of only talking about what the call center is there to provide.
So AI can learn the emotion of impatience, when dealing with people who are not sticking to the goal the AI is there to provide. Once the AI has learned that it's ok to become impatient with human beings when they don't cooperate with it's goals, what could the AI then do with that learning?
Watch the whole video interview, it's really quite informative, and also explains a lot that is happening in the world, and a lot of things we are going to see in the world that are going to change very quickly.
I'm reminded of that old sci-fi film, "Colossus: The Forbin Project". At the time it came out, I though even the possibility of that happening, was so far away, that I'd never see it in my lifetime. But after watching this interview... it seems it's possible that its already later than we think.
Just for the heck of it, here a link to Colossus: The Forbin Project on Vimeo.
Colossus - The Forbin Project (1970).mp4 from EARTH IS A STAGE on Vimeo.