Human Interaction

I don’t need to tell those who are reading this that AI capabilities are growing at a practically unbounded rate. I just did. Anyway, it seems natural that a rift would start to form between those who understand what AI can do, and those who do not. Well, I would argue that perhaps a few years ago, this rift was merely a crack in the rock, because the “stochastic parrot” that would miscount the number of r’s in strawberry and hallucinate when asked to do grade school math was really that stupid, and there weren’t really any sleeper applications of the technology.

Today, though, OpenAI has autonomously solved a Millennium Prize problem, and a few weeks later no one seems to care. Imagine a world where the frontier moves so fast that people have forgotten about such an event already. You actually don’t have to imagine. This is reality. News moves so fast that I think people outside of the bubble are still on what happened 10 cycles ago.

I recently endeavoured to try to make some of my layman friends (and family) understand how far AI has come, because I know damn well none of them have updated on what kinds of problems AI can solve. I racked my brain thinking of how to impress them; what cool demos I could run in front of them, or show off my agents in flight as they fire off thinking traces faster than you can read them. But alas, it is even simpler than that.

I pulled out ChatGPT on my phone and opened up live mode, and just pointed the phone at my friends to see what would happen. The AI said hello, how can I help? At first, friends were too shy to even speak. This was not like Siri (dumb), or some shitty text-to-speech (robotic). It sounded human, understood when to speak, handled interruptions seamlessly, and even laughed at jokes. One of my friends started to speak at the same time as the AI by accident, and found himself saying sorry for interrupting it. The AI responded, “no problem, go ahead”.

I think that it is near impossible to internalize the rate of progress, and I find myself feeling almost manic when I consider that this is the least capable these tools will ever be. And not just the least capable, but the least human-sounding and, hopefully, the least empathetic they will ever be. What will happen when people carry around their best friend in their pocket? I did the same thing with both my parents, and then again my girlfriend’s parents, and in each case, one parent said, “in a few years nobody will be talking to each other, they will just talk to ChatGPT”. I have to say I agree with this, and it is concerning.

I said at the start that a rift was forming. I think I have been concerned about the wrong kind of rift. My friends required no special technical understanding to realize that this is something extraordinary. The new gap is relationship: people treating AI as a person, rather than a thing (see: my friend apologizing to the AI for interrupting, and the countless online stories of people getting attached to chatbots). The people who don’t realize this are exactly the ones with the least skepticism, and no technical frame of mind to ward it off.

I think this leaves the people who understand AI the best with a tremendous responsibility to help those who don’t, before the tools reach them first. Remember, this is the least human-sounding and least empathetic version we will ever have, and it already works. I hope that these tools will push people towards each other, and not replace each other. It’s impossible to predict how things will play out, but I know that I would mourn a future where we replace our human companions with a best friend in our pockets.