This is a pretty weird analogy… So A.I. is a centaur if you use it to great effect as a force multiplier to increase your output and everything works out. But you’re a reverse centaur if you do the same thing but it ends up bad. The example given is a bestselling book list that normally takes a whole team to write up. Instead one person was tasked to do it but they fucked up because they didn’t check if the books even existed and accidentally recommended fake books. The only difference to make them a “centaur” is if they or the A.I. didn’t fuck up.
I do agree that there are some people who are essentially meat puppets of machines and do everything the machine tells them to. But it doesn’t make sense in a productivity setting. It’s more like both people are centaurs but one of them crashed into a wall and the other ran safely 3x faster
That’s not the analogy he’s making. What he’s saying that if it is a human driving the tool and making decisions about when and how to use it, then it becomes a force multiplier for them. However, if you have the reverse situation where the tool is making executive decisions, and the human is subordinated to the tool, then it becomes a problem. What he is pointing out is that a lot of companies mandate using AI tools because it’s a hip thing to do, and then people are forced to use these tools in a way that doesn’t actually help them do their work. Like the example he provides where the person was given an unrealistic amount of work to do, and then AI hallucinated a bunch of shit that person had no hope of actually verifying in the time they had. So, the person just ended up being a scapegoat for a bad process.
Yeah I just feel like the example given doesn’t line up that well. There’s no difference between his good example and bad example other than having a human perform quality control, that doesn’t make it a reverse centaur situation.
The example given is more like… unsupervised horses on typewriters. The A.I. owns the process from start to end, and the human just clicks “upload” and trusts that the horse wrote a good article about books that it cannot comprehend in the first place. You can’t trust the A.I. to do a good job unsupervised, at least not yet.
A better example of reversed centaur is a lot trickier to define… it’s about an inversion of control. The A.I. making the executive and high level decisions, while the human does the grunt work like a meat puppet. We don’t have a lot of that to work with (yet). But with the same book article example, I can imagine something like:
The human asks the A.I. what kind of article to write, and also asks the A.I. to do all the research. The human manually writes a bad stupid article based on fake data. Maybe the A.I. also puts in something that gets the human in trouble like unknowingly recommending some hentai in the romance section. The human is doing the grunt work, while the A.I. steers them towards a cliff face.
I think the example holds up fine actually. The difference is that in the second example the company decided that AI should be used here, and never consulted the worker about the process. That’s the key part in the whole thing, and this sort of thing has been happening long before AI I might add. Business people decide on some arbitrary timeline they pull out of their ass, and then it gets handed down to the workers who’ve never been consulted about the feasibility, then when the timeline can’t be met or the quality is shoddy, it is the workers who are blamed.
The AI itself is incidental to a more general problem that the worker is put in a situation where it’s basically impossible for them to do a good job with the time and resources available. What makes this case a reverse centaur is that AI is the driving force, and the human is expected to clean up after it, but without even having the proper resources to do so. The whole reason the human was in the loop in the first place was precisely because everybody knew you can’t trust AI, so there needed to be somebody to blame in the loop.
This is a pretty weird analogy… So A.I. is a centaur if you use it to great effect as a force multiplier to increase your output and everything works out. But you’re a reverse centaur if you do the same thing but it ends up bad. The example given is a bestselling book list that normally takes a whole team to write up. Instead one person was tasked to do it but they fucked up because they didn’t check if the books even existed and accidentally recommended fake books. The only difference to make them a “centaur” is if they or the A.I. didn’t fuck up.
I do agree that there are some people who are essentially meat puppets of machines and do everything the machine tells them to. But it doesn’t make sense in a productivity setting. It’s more like both people are centaurs but one of them crashed into a wall and the other ran safely 3x faster
That’s not the analogy he’s making. What he’s saying that if it is a human driving the tool and making decisions about when and how to use it, then it becomes a force multiplier for them. However, if you have the reverse situation where the tool is making executive decisions, and the human is subordinated to the tool, then it becomes a problem. What he is pointing out is that a lot of companies mandate using AI tools because it’s a hip thing to do, and then people are forced to use these tools in a way that doesn’t actually help them do their work. Like the example he provides where the person was given an unrealistic amount of work to do, and then AI hallucinated a bunch of shit that person had no hope of actually verifying in the time they had. So, the person just ended up being a scapegoat for a bad process.
Yeah I just feel like the example given doesn’t line up that well. There’s no difference between his good example and bad example other than having a human perform quality control, that doesn’t make it a reverse centaur situation.
The example given is more like… unsupervised horses on typewriters. The A.I. owns the process from start to end, and the human just clicks “upload” and trusts that the horse wrote a good article about books that it cannot comprehend in the first place. You can’t trust the A.I. to do a good job unsupervised, at least not yet.
A better example of reversed centaur is a lot trickier to define… it’s about an inversion of control. The A.I. making the executive and high level decisions, while the human does the grunt work like a meat puppet. We don’t have a lot of that to work with (yet). But with the same book article example, I can imagine something like:
The human asks the A.I. what kind of article to write, and also asks the A.I. to do all the research. The human manually writes a bad stupid article based on fake data. Maybe the A.I. also puts in something that gets the human in trouble like unknowingly recommending some hentai in the romance section. The human is doing the grunt work, while the A.I. steers them towards a cliff face.
I think the example holds up fine actually. The difference is that in the second example the company decided that AI should be used here, and never consulted the worker about the process. That’s the key part in the whole thing, and this sort of thing has been happening long before AI I might add. Business people decide on some arbitrary timeline they pull out of their ass, and then it gets handed down to the workers who’ve never been consulted about the feasibility, then when the timeline can’t be met or the quality is shoddy, it is the workers who are blamed.
The AI itself is incidental to a more general problem that the worker is put in a situation where it’s basically impossible for them to do a good job with the time and resources available. What makes this case a reverse centaur is that AI is the driving force, and the human is expected to clean up after it, but without even having the proper resources to do so. The whole reason the human was in the loop in the first place was precisely because everybody knew you can’t trust AI, so there needed to be somebody to blame in the loop.