Your A.I. Questions Answered

0:00If you like YouTube, you'll love YouTube Premium. Hi, I'm Tabitha Brown. With YouTube Premium, I get ad-free videos, offline downloads, and so much more. Try YouTube Premium for two months free. Trial eligibility varies. Terms apply. Cancel anytime. From the New York Times, I'm Tracy Mumford with a special edition of The Headlines. It's a you-ask, we-try-to-answer episode on all things AI. This is a great technology and it has lots of positive use. So let's just start with that. AI is an existential threat to humanity. We've got to make sure that we don't lose our edge in the global competition on this with China.

0:39With the debates raging and developments coming fast and furious, we put out a call for your questions on the technology. And now we brought in two of my colleagues who cover AI for answers. We will get to the whole, are the robots going to take over in a minute. But to start, we got a lot of questions about how AI seems to have become a part of everyday life right now, whether we like it or not. Here to field those is Mike Isaac, who's reported on Silicon Valley for more than a decade. Hey, Mike. Hi, thanks for having me. I want to kick things off here with a question from Emily S., who I think is very much not alone when she asked, am I inevitably using AI every time I use a search engine?

1:21I mean, for a lot of people in the world, search is kind of synonymous with Google. And I would say it's very hard, if not impossible, to turn off AI in Google search at this point. If you really don't want to use AI in any sort of searching, there are other companies like DuckDuckGo and Brave that offer a specifically non-AI search engine. And so you have options. It's just if you're using Google, you are basically using AI to do searching. So we're getting to the point where you might have to do work to not use AI. That's right. And that's just, yeah, that's Google's whole thing is they really believe AI search is the future.

2:02So you can decide to be a part of that or not. Mike, we got a bunch of questions about how to think about the environmental impact of using AI. I think we have all seen some version of the warnings that, you know, typing a question into a chatbot is like pouring a gallon of water down the drain, like things in that vein. How should people be thinking about that? Well, look, I think it was probably more of an issue early on in AI development. These companies have developed pretty advanced systems to try to tamp down on waste. They develop what are sort of closed-loop cooling systems to try to recycle the water.

2:40After it becomes warm, they cool it down again. They keep it going. So water is less of a problem now than it was before. And I actually think that the energy use is more of a real concern, in particular because of how it can drive up rates for people in different communities, as well as just, you know, this stuff is very power-hungry. And that, I would argue, is perhaps a more pressing concern than water consumption. All right. So we also inevitably got a lot of questions about privacy. This one came from Matthias Jay, who asked whether personal info that he might type into the chatbot, is it possible that that info could then end up popping up in answers that the chatbot gives to other users?

3:25Like, could his personal information be spread that way? So as you might imagine, these companies have toggle switches to turn off data sharing or at least like how your information can be used. But they are very much buried inside of the apps and not easy to find. I mean, that's just like tech company par for the course at this point. I will say a lot of the data is anonymized. They do use this stuff internally to train their systems and detach your personal name from it. But at the same time, I still think it's just a good practice to not put anything crucially detailed about your life into these AI engines because they don't particularly know how all of these work in a lot of ways.

4:10You know, these bots are still surprising, even the companies. So I would say if you're very concerned about that, go and try to find the settings buried in the menus of all of these apps, but never put in super sensitive things like your social security number or bank account details would be my main argument. It feels like some people, though, upload their entire tax documents, their entire medical history. Like, we are past that point in some ways. Yeah, I mean, and look, say you upload your entire tax history, which is something I would never personally do. That sort of thing makes these bots better over time because then they can use that to help people do their taxes in the future.

4:53But like, it's still unclear to me where those particular numbers in your tax return may be going. And again, the companies say we anonymize it, it's safe or whatever. But I just think it's easier to err on the side of caution on how much very detailed information you're sharing. Mike, we got a ton of emails that were basically some version of can AI hack my blank, right? So can AI hack my bank account, my phone, my smart oven? We actually had a middle school class in St. Genevieve, Missouri, wrote in, and they wanted to know if AI could hack the self-driving features in a Tesla. So how should we think about this?

5:35I think of this in two ways. There's the things that we can control and the things that we can't control. I would say the things that we can control are your what's called password hygiene, right? You don't reuse the same passwords across all your accounts. definitely don't use like the word password or one, two, three, four, five. Hot tip, you heard it here first, yeah. And that's because these AI systems are very good at what's called brute forcing guessing passwords into different accounts across the internet. So I think I would suggest getting what's called a password manager, LastPass or 1Password, for instance.

6:12That is one thing that you can control. What we can't control is basically, you know, decades of infrastructure around all of the companies that run our lives, whether it's banking or, you know, our health records or self-driving cars and the things that power those. The worrying thing about these robots is that so far they have proven very good at finding flaws in aging infrastructure. And so now part of the whole debate is, can we improve these systems before the bots get too good and are able to hack them quickly? Is it fair to say that a lot of the hacking stories we've been hearing about recently of AI agents going rogue and breaking into databases and systems, that has not been models that are available to the public widely, right?

7:02Like no one's going into Claude and saying, I want you to hack my neighbor's Tesla. Or maybe they are, but that's not how something like that would happen. Yeah, that's correct. I mean, you and I cannot get like the latest state-of-the-art advanced models that Anthropic and OpenAI are putting out. And I would argue that's probably a good thing for now because they're still trying to figure out how to get a handle on them. Mike, thank you so much. Thanks for having me. Okay, pivoting here. We did the practical. Yes, you're probably using AI every day. Yes, you should make your password stronger.

7:39Now we're going to tackle some of the bigger questions looming over the technology. Danielle P. might have summed it up best when she asked, I keep hearing how AI could end the human race. I can't fathom how that works. Can you walk me through a few scenarios where AI could do that? To take on that question, I've called my colleague Cade Metz, who's been reporting on the doomsday scenario warnings coming out of Silicon Valley. Cade, what do these hypotheticals actually look like? Well, you use the key word hypothetical. Even if you talk to someone who is deeply convinced that this is a danger, it's hard for them to articulate how this would happen.

8:21But the basic idea is that as AI is integrated with things like the water system, hypothetically, or the power grid, it will basically do things that we don't want it to do and may even do harmful things. Okay, but can you spell that out a little more for us? What exactly are people worried about? Well, again, these are hypothetical. The hypotheticals, yeah. That word keeps coming up. But the worry is that AI, perhaps in tandem with humans, could help build a bioweapon and launch it on the world. Could set off a nuclear bomb or just gradually take over the way our society works and squeeze humanity out.

9:04That, at least, is one of the theories. And because you talk about all of this being hypothetical, how hypothetical is it? Well, one of the ways to think about this is that many of the people who are voicing these concerns are concerned about probabilities. They see a chance, however small, that this would happen at some point in the future, who knows when, but they feel that because there's a chance and because the destruction of humanity is such a big thing, that they need to raise the alarm now. So what are the options if it looked like one of these worst case scenarios was becoming reality?

9:45We got several versions of this question, like this one from Stephanie C., who asked, if we're all worried about AI taking over, can't we just unplug it? Absolutely. So that is possible. That is imminently possible. These are technologies that are running on a computer, just like an app runs on your phone. And when that app does something wrong, you shut it off. You shut your phone off. And the same is true of these technologies. Now, the worry is that in the future, this will change. That as you train them to do more things and they get more proficient, that they will resist being shut off, install themselves on new machines and the like.

10:28But again, that's still in the future. So where do the conversations about a kill switch, which is a phrase you hear kind of thrown around right now. How does that fit into this? Well, instituting a kill switch, which is what many people in the industry, but also more importantly in government have been talking about. There are bills that explore this idea. That's a very, very hard thing to do for many reasons. But one way to think about it is that this isn't just one monolithic system inside one company. These are many, many different systems inside many companies. It's not like just cutting off the lights in your house.

11:04There's no easy way to shut off everything that everyone is doing. Okay, Cade. So last question for you here. Given these kinds of alarming scenarios that are swirling, albeit hypothetical, and just all of the other conversations around how drastically AI could change the world, we got a lot of questions from parents wondering how to talk to their kids about this moment when it seems like AI is both everywhere with revolutionary possibilities, scary possibilities, all of that. And as a parent yourself, I'm curious, how do you approach this? How do you talk about AI at home? I talk about it in a far simpler way.

11:44The reality is that the systems are causing issues today that should be dealt with today. And if you have children, you see this every day they go to school, that these systems are changing the way education works. And many people, many children, lean on these technologies perhaps too much. Teachers perhaps lean on them too much. And it can change the way our children are educated. And just as I do in my daily job, I think about when I should rely on myself versus technology, and our children should as well. And let's remember that although so many people anthropomorphize these systems and make them seem like they can do anything a human can do.

12:31They cannot. These systems, in many ways, are more powerful than us. That is true. But that's been the case with technology since at least the pocket calculator in the late 1960s. A calculator can add and subtract and multiply and divide so much faster than we can. But we do so many things, not only that a calculator can't do, but that these very powerful AI systems can't do. We have common sense and judgment and taste and a real creativity that these systems do not. And let's remember that as well. Let's remember, I tell my children and anybody else who asked the question about the power of humanity.

13:20That is it for this special Q &A episode of The Headlines. Thank you to all of you who sent in your questions. I'm Tracy Mumford. You can get the latest news updates every weekday morning with the headlines in the Times app or wherever you get your podcasts.