Ben Zhao
speaker
105 appearances
2 recordings
2 series
first heard Jan 2025
last heard Mar 2025
Ben Zhao’s voice in public audio — every appearance, attributed to the second.
Trend
recordings per month · last 12 monthsNo recordings in the last 12 months.Older appearances are listed below; set an alert to hear about the next one.
Appearances
It is a storm that will eventually blow over. And how I see my role in this is not so much to necessarily stop the storm. I'm more, if you will, a giant umbrella. I'm trying to cover as many people as possible and shield them from the short-term harm.
I would actually disagree, but that's OK. We can have that discussion, right? Look, you're the guy that knows stuff. I'm just asking the questions. I don't know anything about this. No, no. I think this is a great conversation to have because back in 2022 or early 2023, when I used to talk to journalists, the conversation was very, very different. The conversation was always, when is AGI coming?
You know, what industries will be completely useless in a year or two? It was never the question of like, are we going to get return on investment for these billions and trillions of dollars? Are these applications going to be legit? So even in the year and a half since then, the conversation has changed materially because The truth has come out.
These models are actually having trouble generating any sort of realistic value. I'm not saying that they're completely useless. There are certain scientific applications or daily applications where it is handy, but it is far, far less than what people had hoped them to be. And so, yeah, you know, how do I believe it? Part of this is hubris. I've been a professor for 20 years.
I've been trained or I've been training myself to believe in myself in a way. Another answer to this question is that it really is irrelevant because the harms are happening to people in real time. And so it's not about will we eventually win or will this happen eventually in the end? It's the fact that people's lives are being affected on a daily basis.
And if I can make a difference in that, then that is worthwhile in and of itself, regardless of the outcome.
Very interesting. Okay, let me unpack that a little bit there. The thing that allows me to do the kind of work that I do now, I recognize is quite a privilege. The position in being a senior, tenured professor, and honestly, I don't have many of the pressures that some of my younger colleagues do.
No, I mean, all of our grants are quite public. And I'm pretty sure that I'm not the most well-funded professor in the department, but I run a pretty regular lab. We write a few grants, but it's nothing earth-shaking. It's just what we turn our time towards. That's all. There's very little that drives me these days outside of just wanting my students to succeed.
I don't have the pressures of needing to establish a reputation or explain to colleagues who I am and why I do what I do. So in that sense, I almost don't care. In terms of self-interest, none of these products have any... money attached to them in any way, shape or form. And I've tried very, very hard to keep it that way. There's no startup. There's no hidden profit motive or revenue here.
So that simplifies things for me.
No. The university always encourages entrepreneurship. They always encourage licensing, but they certainly have no control over what we do or don't do with our technology. This is sort of the reality of economics and academic research. We as a lab have a stream of PhD students that come through and we train them. They do research along the way and then they graduate and then they leave.
For things like Fox, where, you know, this was the idea. Here's the tool. Here's some code. We put that out there. But ultimately, we don't expect to be maintaining that software for years to come. We just don't have the resources. That sounds like a shame if you come up with a good tool.
Well, the idea behind academic research is always that if you have the good ideas and you demonstrate it, then someone else will carry it across the finish line, whether that's a startup or a research lab elsewhere. But somebody with resources who sees that need and understands it will go ahead and produce that physical tool or make that software and actually maintain it.
You know, at a high level, I think that's great. I think if we get to that point, that will be a very welcome problem to have. We are in the process of exploring perhaps what a nonprofit organization would look like, because that would sort of make some of these questions transparent.
Well, yeah, very different type of nonprofit, I would argue. I'm more interested in being just the first person to walk down a particular path and encouraging others to follow. So I would love it if we were not the only technology in the space. Every time I see one of these other research papers that works to protect human creatives, I applaud all that.
In order for AI and human creativity to coexist in the future, they had to have a complementary relationship. And what that really means is that AI needs human work product or images or text in order to survive. So they need humans and humans really need to be compensated for this work that they're producing.
Otherwise, if human artistry dies out, then AI will die out because they're going to have nothing new to learn on and they're just going to get stale and fall apart.
Over the last couple of years, I've been practicing lots of fun analogies. Barbed wire is one, the large Doberman in your backyard. One particular funny one is where the hot sauce that you put on your lunch. So if that unscrupulous coworker steals your lunch repeatedly, they get a tummy ache. But wait a minute, you have to eat your lunch too. That doesn't sound very good.
Well, you know, you eat the portion that you know is good and then you leave out some stuff that... Got it.
Boy, that's a bit of a loaded question because honestly, we don't know. It really comes down to how these models are being used. Ultimately, I think what people want is creative content that's crafted by humans.
Showing 81–100 of 105 · page 5 of 6
← Previous
Next →