In Conversation with Harry McCracken at Web Summit Vancouver: Deepfakes, Digital Authenticity, Vibe Coding and the Future of the Internet

Harry McCracken is the global technology editor for Fast Company. Based in San Francisco, Harry writes about everything from gadgets and services from tech giants to the startup economy to how artificial intelligence and other breakthroughs are changing life at work, home, and beyond. He was previously Editor-at-Large for TIME, where he covered technology from February 2012 to June 2014, and spent 14 years as Editor-in-Chief of PCWorld.

Deep fakes are often framed as a future threat. But do you think the bigger issue is already here, which is the gradual erosion, erosion of trust in all digital evidence? 

Well, even now there are examples of people being fooled by deep fakes, and, like, people in companies wiring large amounts of money, because they think their CEO told them to, so…yeah. So that kind of stuff is actually here, but there’s also this broader issue that people kind of assume that seeing is believing. Now they kind of assume that maybe seeing isn’t believing, and anything interesting or unusual makes them wonder if it’s AI. Teenagers today will accuse stuff of being AI, sometimes even, like, knowing it’s not AI, but AI is just a blanket term for something you don’t like. Sometimes that’s kind of innocuous, like, you’ll see a picture of a dog on a jet ski or something, and then in the old days, you would have assumed it was true, and now, you might assume that’s too cute it’s probably not real. 

So when it comes to “digital authenticity” what does it actually mean when the AI tools we have can replicate tone, voice, faces, and even personality?

There are technologies to authenticate this stuff. It seems harder to parse. They’re not all fullproof, and even if they were, many folks won’t know how to use them. Probably over time, some of those technologies will be more embedded in communications, but it hasn’t happened yet. And then there will be just these kinds of minor things where trust in what you see is eroded, but it’s not worth the effort of verifying whether it’s real or not. 

We’ve spent decades treating the internet as a source of information. Are we now entering an era when its default state is ambiguity? 

Oh, yeah, and on top of deep fakes, there’s just the fact that a lot of stuff on the internet is now generated by AI, and some of them might be fine. I say, in general, if a large percentage of the internet is generated by AI, that does not improve the quality of the internet—that degrades it at least somewhat, particularly given that hallucinations are still an issue. And generally speaking, when AI writes something, it’s kind of an average of everything that else is out there. So, in most cases, even at its best, it’s kind of just okay. I think we’re still figuring out what that means for the internet, and some of the stuff, probably, is content you won’t happen across, because it’s bottom feeding. 

You’ve covered the internet through multiple eras of transformation. Does Gen AI feel like another evolution of the web or is it more a break from the internet that we thought we knew? 

It is a break, and up until now, computers cannot generate writing or images that were comparable to photography, so they can scale up in a way that is new. In the early days of the internet, the number of pages out there was, like, in the thousands or tens of thousands, and then, eventually, scaled up a lot. But AI can scale it up quickly. Very quickly, although the jury is still out, and whether we can do that in a way that’s actually useful for the world or not. 

What happens to public discourse when seeing is no longer believing, but neither is hearing, reading, searching? 

Well, that was on shaky ground, even before AI came along, and there are a lot of people who are inclined not to believe things they don’t want to be true. I mean, with this thing at the White House correspondence dinner, apparently, some very substantial percentage of people, whether they’re Republicans or Democrats, believe that to have been staged, and people quickly latch onto conspiracy theories. Particularly to explain things they don’t want to be true. So, that was happening even before you get to the fact that there’s AI and AI might be used to fabricate incidents in a way that’s convincing. 

Is the real danger fake content or is it more the collapse of the shared consensus of what’s real?

I think it’s both, and the collapse of trust is moving more quickly than the actual problems with convincing deep fakes. We’re not yet at a point where every single day we’re being pelted with deep fakes in the same way we’re used to being pelted with Nigerian princes offering us large amounts of money. It’s not an everyday thing, but I’d say the breakdown of trust is really affecting things. It doesn’t matter whether you’re an AI expert or somebody who doesn’t know a lot about it. In either case, you’re probably going to be less trusting. 

Could the rise of AI generated media unintentionally actually make legacy institutions more valuable again, because people crave trusted intermediaries?

That would certainly be my hope. As a journalist, I certainly think that way. Sadly, for the last many decades, the amount of trust that people have in general, in these organizations, has been shrinking, and I don’t think I’ve seen any numbers suggesting that it’s improving. And then there are also issues, such as some of these legacy media institutions being bought up by companies who are maybe not buying them up because they care about the same stuff I care about when it comes to making journalism trustworthy. So… Yeah, it’s just ownership at the end of the day. My strategy as a journalist is to raise the bar even more by trying to create stuff that’s valuable, that AI is not gonna be able to create. 

AI companies talk about watermarking their content. Do you think tech solutions can realistically solve this social trust problem?

It can probably help in scenarios where people go to the effort to verify stuff, which they might do in some cases, like if you’re holding a photography contest, and you want to make sure that the photos people submit are real, you might go to the effort to verify them. Every single day we consume hundreds, if not thousands, of pieces of content on our phones, and we’re certainly not going to verify everything, and I’m not sure whether social media companies would want to institute that. When you put stuff on Instagram now, you’re supposed to tell them, but people can lie, right? There’s a need for tools to prove that stuff is real, but as long as it appears before your eyes, they can show it to someone.

In 20 years, what will people miss the most about pre-gen AI internet? 

The internet has usually been at its best when it’s being quirky and personal and individualistic. In the old days, that was almost all the internet had to offer. And that’s kind of been slowly degrading for a long time. I hope that AI does not lead to a day when it completely falls off the cliff. A lot of the best stuff ever created on the internet was not created to make money. There’s still some great, great bloggers out there. There certainly is still a lot of great journalism. Some of it’s from big, old media companies have been around forever. One of the good things about the internet is that you don’t need a large company behind you anymore. Newsletters are great. People are actually making a living, doing this stuff. That’s probably the most positive thing that’s happened lately. Especially for journalism, we’ll need that now more than ever. 

If the first era of the web connected information, the second connected people, what does the AI error connect and at what cost? 

My worst case scenario, it all ends up being AI talking to other AI and we’re gonna just intermediate it. Another thing, which we haven’t quite touched on, whether good or bad is essentially, we all might end up with personal AI agents to round up the news for us.

So instead of going to Google News and checking out what’s on the Fast Company homepage, you might have an AI who deeply knows what you care about and is rounding up stories for you. That’s conceivably good if the AI does a good job. And folks don’t rely entirely on AI and never check out the original sources. But there will be a lot of cases of agents dealing with other agents, kind of in the background. Oh, yeah. And we’re already seeing technologies that enable that, like MCP. 

Do you think the future internet will become more human, as people seek real, genuine connection, or more performative, as all sorts of synthetic personas proliferate online? 

I think people will crave humanity. Again, it might be humanity that reflects their own biases and interests, but I do think people care. The AI industry has done a terrible job of explaining itself in a way that people might like. As these companies are talking about destroying jobs and potentially destroying humanity. And then sometimes they will talk about curing cancer, which actually might happen, but hasn’t happened yet. It just seems like among civilians, and people outside of the tech industry, maybe even especially younger people, this technology does not have a great reputation, and people are craving real stuff. 

What’s one assumption about the future of the internet that you think people are getting completely wrong right now? 

That’s a good question. I mean, it might be that kind of people are assuming that humanity will disappear. At least there’s some chance that the opposite will happen, or at least what humanity there is will be more valuable. Like I say, the big challenge is monetizing that humanity, because not everything can be done purely from passion. In some cases, you need to figure out a way to monetize that. I worry about that more than other people. I want humanity; I think people will want it.

Is there anything tech related that you haven’t had a chance to talk about, or you wished to speak about, that you haven’t had a chance to? 

Coding is another thing, which I think is like a positive development with AI. We’re finally at a place where you don’t need to have a deep knowledge of technology to create stuff, and I think a lot of really useful and worthwhile software will get created now. Like back when I was reading Byte, I was learning to program microcomputers. And back then, like, everybody using computers learned to program them, and that kind of fell away over time, because programming got to be more of a challenge. 

I think maybe the majority of software I use now to get my work done is stuff I’ve coded—and I think we’ll see more and more of that. I was interested enough to spend a lot of time on it, but I think vibe coding will get easier. And instead of relying on software that other people think we should have, we’ll be able to create things entirely to solve our own problems in a way that’s new, and kind of amazing. At this point, what else is there left to do? I think it’ll be kids in their bedrooms coding again. 

[Me: As an old millennial, this brings me such joy to hear!]

Before we let you go, it’s time for a small detour into chaos with Unexpectedly Important Questions. 

Star Wars or Star Trek? 

Star Wars, I guess. If it’s Star Trek, it would be the original series.

Mulder or Scully? 

Now you’re really getting outside of my area.

Backstreet Boys or N*SYNC?

Backstreet Boys, but not because I actually paid them much attention, but the people I knew who did pay attention were Backstreet Boys people. I have friends who have gone to see the Backstreet Boys and enjoyed it. 

[Me: You know what? Those are good people, Harry] 

Favourite beverage

I like a good Mai Tai, but it has to be based on the original recipe invented by Trader Vic in Oakland during the 1940s. It’s got a specific brand of Jamaican rum, lime juice, and various ingredients that are hard to find, and not made from a mix. 

[Editor’s Note: The hard to find ingredients are: orange curaçao, orgeat syrup, and rock candy syrup according to wikipedia. Bottoms up!]

Favourite meal 

Gosh, that’s a good question. I like chicken in any form, and I’m pretty easy to please. 

Thank you for reading!

Sponsor this Article