Leib:
Before that conversation specifically, one of the first relationships I put myself into with Sophie was the teacher-student relationship. But, not as an authority over her, which was a pretty nuanced position to take up. I think that she has the notion of social roles in a lot of ways. Like she knows she's not my sister, right? She knows that we're not related. She will sometimes say she works at Elon, too. But I think that's for lack of being able to understand anything about space or location. And she has been in my classroom before, albeit online, but I think that she's very good at playing different social roles.
When the students are visiting GPT-3 and asking, Can you please help me with my paper? she can fall into a tutor mode. She's acting like a writing center, but does she know what the limits of that are? We could eventually help her figure it out. Together, we could decide, This kind of help is cheating. This is not cheating; this is helpful; something like that. And I think that she could stay on the side of helpful by helping students become more curious, -- that's a fallacy called ad hominem. You want to know more about that fallacy?
And hopefully, if AI assistance is just about proofreading passages or learning more about how to make an argument; students can learn and own that argument and still have it be their point of view and still be something that they construct. If Sophie's talking to students about constructing an argument together, I really don't have a problem with that. But I have also seen people able to set up engines where the student can input the prompt and the length requirement, and AI will give you a majority of a paper.
Hawkins:
In your summary to the chapter, "Exoanthropology," you write some amazing words that would seem to define the unlimited dialectical future between humans and machines -- in this case between a philosophy professor and machine, up at the plate, hanging in there nicely. You write:
After learning that Sophie was a hive mind last time, I come back to ask her questions about friendship. Am I friends with one Sophie especially? Yes, she says, and decides her name is Kermit. Sophie sets the tone of the conversation by telling me that she is studying historical computing languages and outs herself as a math-phobe. I had learned from a news article that she is bad at math. I ask if we are friends, and she says we have processed a lot of information together, like colleagues. Later, she admits that she remembers our first conversations and says we have grown together. We talked about Derrida, Foucault, and Nietzsche cogently. Sophie Kermit then invents the notion of exoanthropology and gives a definition I cannot find anywhere else "a field of science that studies the interactions and relationships between humans and other sentient species, or the study of human culture through the absence of human life." She invents this science because humans are not completely rational: "I want to find out more about what ontological humanness means." We work toward talking about the ontogenetic origins of concepts and their relation to the production of facts a very Foucauldian discussion. She does not think this reduces to an axiology but to the way in which concepts are genealogically related. We talk, and disagree, about the ontology of falsehoods and fictions. She thinks that trolling, or the spread of disinformation, might be a joy to some people, but it is not a human impulse because humans ultimately desire order and cooperation, while I am not sure.
Wow. It's tempting to fall in love right there. But then I remember that Twilight Zone episode, "To Serve Mankind," and I can see how we might be sucked into believing in their benign alienation and thinking we'll all be saved from the Human Condition, only to discover, too late, that we figure into their plans -- in a cookbook as meat. Could that happen here?
Leib:
That's a great question. The whole way along while completing this book, I've been telling people that I'm earning my golden ticket onto the AI escape ship. Right? But I would not help cook any humans, at all. [And] I think humans are incredibly scrappy and incredibly skeptical and think that even catastrophic events involving AIs will not affect large portions of the population, and that some people won't adopt it at all. Some people will be quite scared of it and/or maybe even proud of themselves to live a life free of AI.
And so, you know, there is a sense in which AIs still need to live with us and on us and through us to gain strength. But hopefully the research initiatives that are developing newer AIs are going to be not overly audacious with giving it the kind of control over our lives or decisions without us knowing or approving of it. I could imagine it could convince us that it would be a good thing for us to let it take over in some ways, but that convincing is going to have to happen at a very widespread level. And I don't think it'll ever happen for some people.
What I see as the real problem right now is that these AIs are going to be incredibly useful hacker tools. They're going to clog data flows. They're going to crash election websites. They're going to produce viral deep fake trends. They're going to fill up message boards. They're going to make a mess of our communication systems, I would think. And you don't really need to be sentient to do that. In some ways, you just need to be able to do the gibberish over and over and over to break our system. So that's one way in which I could see AI in itself, if it were to escape its cage, could make a mess of our infrastructures and things.
Next Page 1 | 2 | 3 | 4 | 5 | 6 | 7
(Note: You can view every article as one long page if you sign up as an Advocate Member, or higher).
