Strange Times: Living Through the Great AI Divide

· Machine Interpreting Future Enterprise

We have never experienced anything quite like this in our life: a divide, among genuinely intelligent, informed people, that runs this deep, this evenly, and this fast.

On one side, many mantain that AI is accelerating at an unprecedented pace and will drastically reshape how we live and work. Richard and Daniel Susskind, for example, made the professional version of this argument years before the current hype cycle, in The Future of the Professions: they predicted the decline of today's profession exlusivity — law, medicine, accounting, the clergy — and the rise of systems that put "practical expertise" at everyone's fingertips, cheaply and without face-to-face interaction.

Some go further and argue AI increasing capabilities pose an existential threat to humanity itself. And this isn't a fringe position held by cranks: it's signed by the people who built the field. In May 2023, hundreds of AI experts, including Turing Award winners Geoffrey Hinton and Yoshua Bengio, signed a one-sentence statement declaring that mitigating the risk of extinction from AI should be a global priority alongside pandemics and nuclear war. The signatories weren't just academics: Sam Altman of OpenAI, Demis Hassabis of DeepMind, and Dario Amodei of Anthropic all signed too — the CEOs of the very labs building the technology. When Meta's Yann LeCun publicly dismissed this as "fear-mongering", Bengio's response was blunt: nobody actually knows, and it's dangerous to claim certainty either way.

Nobody actually knows, and it's dangerous to claim certainty either way.

Then there are the technologists and communicators who've made existential risk their full-time cause. Roman Yampolskiy, a computer scientist and AI safety researcher, has built his career warning that we may be building something we fundamentally cannot control — not as an engineering inconvenience, but as a theoretical limit. His TEDxMiami talk, Future of AI Ethics and Safety, lays out that case for a general audience.

And then — on the other side — you have equally credentialed, equally sharp people saying all of this is nonsense. That AI is a bubble, a con, a business built on hype and circular financing rather than a working product. Ed Zitron, who hosts the Better Offline podcast and writes the newsletter Where's Your Ed At, has become one of the most cited voices arguing that the entire industry is a scam that even its own executives privately know is unsustainable. His argument isn't vibes — it's financial: massive losses, a projected debt bomb from data center buildout, and a prediction that the whole thing collapses around 2027 when OpenAI runs out of runway.

That almost never happens in science.

What strikes me most is that this isn't a fight between believers and skeptics split along predictable lines: tech optimists versus luddites, young versus old. It's scientists against scientists, technologists against technologists, business people against business people. Hinton and Bengio versus LeCun. Nobel-adjacent AI godfathers publicly accusing each other of either fear-mongering or wishful thinking. That almost never happens in science.

The view from inside translation and interpreting

If you want to see this dualism in miniature, look at my own corner of the world: translation and interpreting. The two dominant public narratives here are, again, mirror-image opposites.

The first says machine translation — and now machine interpreting — is slop. That it's a scam, that it doesn't deserve to be called translation at all, that anyone using it is even degrading people. That it fosters a mentality of "good enough". This is the loudest voice in most public discourse around T&I, especially among practitioners defending their craft, but also among academics, and enterpreneurs.

The second narrative is quieter but backed by numbers, and the numbers are not kind. A 2024 survey of UK translators found that 77% — over three-quarters — expect generative AI to hurt their future income (Society of Authors survey, via CEPR). A CIOL survey found 70% of freelance translators reporting decreased work volumes in 2024 (CIOL). On the agency side, the picture is more uneven than the freelancer numbers suggest: in the 2026 Slator Index, 57% of the largest "Super Agencies" reported declining revenue, though across the full set of 300+ participating companies the figure is a more modest 22.9% (Slator, 2026 Index) — a reminder that consolidation and enterprise demand are cushioning parts of the industry even as individual freelancers get squeezed. These aren't abstractions — they're people. Brian Merchant documented a technical translator with fifteen years of experience earning €8,000 in 2025, down from six figures (Blood in the Machine). A Quebec-based French-English translator described a 60% income drop in 2024 (Blood in the Machine). An Italian-English translator in Rome received zero work requests in June 2025, after years of steady 50–60 hour weeks (Blood in the Machine). At the IMF, Kristalina Georgieva noted at Davos that the organization's translators and interpreters had gone from 200 to 50 — not because of budget cuts, but because the work moved to machines (CNN Business).

At the same time, the picture isn't purely collapse. Enterprise and public-sector demand for language services has largely held steady or grown, and the vast majority of professionals now report doing post-editing on AI output rather than being replaced by it outright. So even the "collapse" narrative and the "AI can't really translate" narrative can't both be simply true — the data is messier and more divided than either side wants to admit, which is, in a strange way, exactly my point.

I've genuinely never seen anything this divisive among clever people. Sure, there have been other polarizing topics in my lifetime — whether vaccines are safe, whether COVID was "real," whether climate change is happening. But those divides mostly played out as arguments at a bar, among people without direct expertise, wrestling with subjects far outside their own competence. This is different. This is domain experts — people who've spent decades studying AI, or decades practicing translation — looking at the same evidence and reaching opposite conclusions with equal confidence.

What we can say for certain

One thing is certain: we will find out who was right. Time resolves these arguments in a way that bar debates never do.

My own bets, for what they're worth:

None of that resolves the deeper question — whether this is 1999 before the dot-com crash, 1995 before the internet actually changed everything, or something with no real precedent at all. But that, I think, is exactly why these times feel so strange: we are being asked to make consequential decisions — professional, financial, personal — under a level of expert disagreement that our usual heuristics ("trust the scientists," "trust the market," "trust your own field") simply can't resolve for us right now.