The most polarizing technology of our time

As a longtime fan of The New York Times podcast Hard Fork, getting to interview its co-host Casey Newton for nexxworks’ Radar podcast felt like a real privilege (you can listen to the full episode here). What stayed with me long after our conversation was the range of perspectives we explored on the growing polarization around AI. I wanted to unpack that tension between optimism, fear and even anger, so this newsletter is my attempt to do exactly that.

In the first half of the 20th century, Carl Jung wrote “The brighter the light, the darker the shadow.” He could have easily written that same thing about the current AI industry. The faster its solutions evolve into better and brighter (pun intended) versions, the more prominent its dark side seems to become, both in terms of public perception and in potential future impact.

On the one hand there is so much (sometimes overblown) enthusiasm about how AI can help augment humans to help us solve some of our biggest problems. The valuations as well as the enormous capex investments point in that direction. Google, Amazon, Microsoft, and Meta for instance recently announced they plan to spend a combined $725 billion on capex in 2026. That’s a 77% jump from last year's record $410 billion.

But there is also A LOT of resistance from different angles, showcasing very similar emotions but also strikingly different opinions.

Rage against the machine

Casey for instance talked about the pundits and observers who believe that the industry is on the verge of collapse because the technology just isn’t there yet. “They focus on how systems hallucinate too much and will not replace any jobs. And they believe that to reach ‘true AI’ we’ll need a completely different approach. I just think these people have been proven to be wrong. Take Claude Mythos. Even though the Anthropic products have been designated a supply chain risk by the Pentagon, government agencies are still finding ways to work with it because it's just too useful. Meanwhile, the Treasury Secretary is convening the heads of all the major banks to ask how they are planning to deal with this crazy new cyber weapon. That does not point in the direction of useless tech, right?”

Another type of friction comes from those who conclude that we’re in full on bubble mode. They do see the technology as promising but believe too much venture capital is pouring into AI companies that may struggle to build durable businesses. And that VCs will lose a lot of money. But, as Casey pointed out, “That's just the San Francisco way”. “VCs expect that about 90 % of their investments are going to go to zero. On the other hand, the amount of money invested in AI is pretty much unprecedented. So we’ll just have to find out what the implications for the broader economy will be. But, just like with the internet bubble, it does not mean that the technology is useless. Far from it.”

A third form of opposition comes from employees themselves. Casey noted that a significant backlash against AI is emerging in the United States. Even building new data centers is becoming increasingly difficult. “This is just people that are using one of their few points of leverage. In America, we're very good at preventing things from getting built and so that’s the lever they are pulling. Companies are now even volunteering to pay the increased electricity costs caused by the data centers, which you know they're not doing out of the goodness of their heart. They're terrified that they won’t be able to build the infrastructure that is so essential to them.”

Casey does feel empathy for the latter type of resistance. “They’re basically thinking “I don't know a lot about this, but if I can stop them from like building AI factories all around the country, maybe it'll be another two years before I lose my job. If you look at it from that point of view, it’s just a very rational action that people are taking in their own economic self-interest. Especially if you realize how US citizens have a very deplorable social safety net.”

Priceless humans

A really fascinating part of the whole debate around work and human relevance, is that AI is making humans simultaneously less and more valuable, depending on the domains. Casey offered an interesting example from the school system: “I was talking with the CEO of an educational tech company and he believed that the advances in AI are making the human teachers even more valuable. Some elements of their jobs will be replaced, but their rising value will also increase their wages.”

“Radiologists offer a similar example. This was supposedly to be one of the first jobs that would be replaced by AI systems. And yet the contrary seemed true. A lot of their tasks cannot be offloaded to machines and they tend to make a lot more now than they did three or four years ago. Of course, part of that is because fewer students are going into radiology on the assumption that pretty soon AI will be able to do the whole job. So it's also hard to get a really clean example. But at least in the meantime, we do see some jobs where the result of AI is that wages are being driven up.”

But the opposite, sadly, is also true. “At the same time, there's a lot of talk about how entry-level jobs are getting harder to come by. CEOs have a lot less appetite for that junior programmer. They want somebody who has a lot of experience, who can handle the hard cases. So I suspect this is just going to look a little different sector by sector in the near term. Over the medium term, however, I do expect that AI will replace an increasing amount of jobs. This will put the individual workers in a position where they have to constantly adapt.”

Casey’s own reaction to the possibility of AI disrupting his job is pretty level-headed: “I’ve more or less accepted that, between now and retirement, some new technological force is always going to be trying to kill my job. I try to set that aside and ask a different question: regardless of what happens to my job, what can this technology actually do? And what are the implications of what’s being built?”

Intelligence Native Companies

Then there is also the emerging idea of “intelligence native” companies. Since the end of February, it has become one of the main topics in my conversations with business leaders, sparked in part by some changes at Block. Its CEO Jack Dorsey framed a major workforce reduction of about 40% as part of building a company that is smaller, faster, and “intelligence native”.

I'm not entirely sure what intelligence native means, but the idea is capturing attention because it suggests a shift in how companies are designed and how value is created. And it's clear that leaders are concerned about missing the boat.

“Dorsey is making a bet that, over time, companies will look less like a group of humans using AI to augment their work, and more like intelligence systems supported by a much smaller number of people,” Casey concluded. “Directionally, he’s probably right. Even if he has proven himself to be, frankly, a terrible CEO. And even if it’s far from clear he’s right in the specific case of Block right now.”

Taking more responsibility

Casey also talked about the shift in responsibility felt or at least communicated by the AI companies in the past decade. “In the beginning, AI CEOs felt different from social media CEOs. They openly acknowledged that what they were building could cause real harm. And for a while, it seemed like they were going to approach that risk with genuine caution. During the GPT-3 era, Sam Altman was on Capitol Hill telling lawmakers: “you need to take this seriously and regulate it”. That made me pretty optimistic, at first. But since then, a lot of those safety guardrails have come off. We’ve seen OpenAI tune ChatGPT to be more engaging even, at times, sycophantic. We’ve seen dozens of people report what they’re calling AI psychosis. And in some cases, those interactions have been linked to self-harm and other deeply tragic outcomes.”

Against this backdrop, Casey said he sees a sense of shortfall in how AI companies are operating. “My view is that these companies haven’t done enough testing ahead of launch, and they haven’t been careful enough. What worries me is that even though they’ve often acknowledged the risks of what they’re building, we may still be headed toward harms that are even worse than what we saw in the social media era. And honestly, I’m quite concerned about that.”

He also argued that the government has largely fallen short. “AI companies are building systems that to some degree they still do not understand, that will be extremely hard to control and that will have huge effects on society. And they're doing it without almost any democratic oversight whatsoever. We’ve seen a real failure of imagination up till now. You had JD Vance giving speeches in Paris saying the AI race wasn’t going to be won by wringing our hands over safety. And the broader posture from the administration often felt like “beat China no matter what”.

“But it does seem like, now with Claude Mythos, the US government is starting to take some of the actual risks more seriously. There’s a growing recognition that American labs could build technologies capable of doing real systemic harm, even something as extreme as threatening the global banking system. And at least I’m glad they’re paying attention to that.”

He ends on an optimistic note, though. “On balance, technology has been enormously beneficial to humanity. I’m absolutely not someone who wishes we could return to an agrarian society. I like having an iPad, you know? And my hope is that if we can increase the amount of intelligence in the world, we may be able to solve problems that have long seemed intractable, like addressing major medical challenges. So I do still see a positive future. I just think the path to get there is very narrow. And that’s why I feel a responsibility to sound the alarm when I think it needs to be sounded.”