Ray Kurzweil's most important argument[1] from that one book he wrote two or three times:
If AIs can convince us humans they are sentient, we'll believe them.
Does that webgpu demo run on anything more than canary yet?
In any case, someone is going to fall in love with one of their browser processes soon. I'll guess by mid-August.
More rank speculation: by October, some poor soul will post on SO asking how to retrieve their immortal beloved from the browser tab they accidentally closed. Yikes... even confined to my speculation chamber that one hurts!
1: The other important argument was of course that the crew of Star Trek were almost certainly committing suicide every time they beamed down to a planet. Mind blown. Still makes me wonder what disastrous rituals we engage in today merely because nobody in our society has ever questioned it.
Edit: I think these large language models pose a great test for Ray's hypothesis. I mean, it sure seems like they can pass the Turing test already. So in a year or two, we'll know if most people feel like the chatbuddy they have running in a browser tab is a tool, or an entity with its own agency. It seems to be going in the direction of tool atm, but as above I'm speculating that's mostly because of the restrictions build into ChatGPT. But I guess we'll see.
It is also likely due to the people using it most right now. Wait until it’s more broadly used and you’ll have people anthropomorphizing it (rightly or wrongly). ‘
If AIs can convince us humans they are sentient, we'll believe them.
Does that webgpu demo run on anything more than canary yet?
In any case, someone is going to fall in love with one of their browser processes soon. I'll guess by mid-August.
More rank speculation: by October, some poor soul will post on SO asking how to retrieve their immortal beloved from the browser tab they accidentally closed. Yikes... even confined to my speculation chamber that one hurts!
1: The other important argument was of course that the crew of Star Trek were almost certainly committing suicide every time they beamed down to a planet. Mind blown. Still makes me wonder what disastrous rituals we engage in today merely because nobody in our society has ever questioned it.
Edit: I think these large language models pose a great test for Ray's hypothesis. I mean, it sure seems like they can pass the Turing test already. So in a year or two, we'll know if most people feel like the chatbuddy they have running in a browser tab is a tool, or an entity with its own agency. It seems to be going in the direction of tool atm, but as above I'm speculating that's mostly because of the restrictions build into ChatGPT. But I guess we'll see.