I spent an hour and a half with Gus Docker on the Future of Life Institute Podcast talking through what I think is the central question of the next few decades: what happens to human work, and to human meaning, once machines can do most jobs better than we can. The episode is called "From Peak Horse to Peak Human," and that title comes straight from the argument I made on the show.
Why the last few technological revolutions won't predict this one
Gus and I started with the standard objection to automation anxiety: every previous wave of technology, from the steam engine to the personal computer, destroyed some jobs and created others, and employment came out fine. I don't think that pattern holds this time, because those earlier revolutions automated physical or narrow cognitive tasks while leaving room for humans to do everything else. What's different now is the range. Cognitive automation isn't stopping with secretaries and travel agents, it's heading toward every job that can be done by manipulating information, which is most of them.
The "peak horse" problem
The analogy I used to make this concrete is peak horse. Horse labour grew for centuries until the combustion engine arrived, and then the working horse population collapsed within a couple of decades, because horses had no other economically useful skill to fall back on. I think humans are the first species that could plausibly do the same thing to itself, not because we lack value as people, but because our economic value has historically depended on being useful for tasks that machines are now learning to do too.
Infinite demand doesn't save us, and neither does the lump of labour fallacy
A common pushback is that human wants are infinite, so there will always be new jobs to invent to satisfy them. I addressed this directly on the show. Infinite demand for goods and experiences doesn't automatically translate into infinite demand for human labour if machines can produce and deliver those goods and experiences more cheaply. The "lump of labour fallacy," the idea that there's a fixed amount of work to go around, gets invoked a lot in this debate, usually to wave away automation concerns, but I think it's being misapplied to a situation that's genuinely different from past disruptions.
Fully-automated luxury capitalism and the abundance economy
We spent real time on what I'd consider the optimistic scenario: an abundance economy where automation drives the cost of goods and services down toward zero, and where something like fully-automated luxury capitalism becomes achievable. I think that outcome is genuinely possible. My caveat, which I pushed on with Gus, is the transition. Getting from today's economy, where most people need a wage to live, to an abundance economy without a brutal employment cliff in between is the hard part, and it's the part getting the least serious policy attention.
Rethinking education for a world with AI tutors
One of the more concrete parts of the conversation was education. If AI can already act as a patient, personalised tutor for almost any subject, then a lot of what school is currently structured to do, standardised instruction at a fixed pace for large groups, stops making sense. I talked about what schooling could look like if it were built around personalised AI tutors from the start rather than retrofitted onto a factory-era model.
What I actually use LLMs for
Gus asked me for concrete, personal examples rather than abstractions, so I talked about how I use large language models day to day: as an extended memory, for drafting, and for getting a second opinion on the emotional tone of something I've written before I send it. I think these mundane uses matter more to how people experience AI right now than any headline about AGI.
Where meaning comes from if jobs disappear
If work stops being necessary, the obvious worry is that life loses its structure and purpose. I don't think that follows automatically. I pointed to aristocrats, retirees and children as three groups of people who have never needed a job to find meaning, and who generally do fine, filling their time with play, learning, relationships and creative pursuits. That's not a complete answer to what a post-work society should look like, but it's evidence that meaning and employment aren't the same thing.
Four futures for superintelligence, and why consciousness matters to safety
Toward the end, we moved into territory closer to Conscium's own work. I laid out four broad futures for how superintelligence could turn out, ranging from catastrophic to genuinely good for humanity, and argued that which one we get may depend partly on whether advanced AI systems are conscious. My reasoning is that a system capable of experiencing something has a basis for empathy that a purely mechanical optimiser doesn't, and empathy could function as a safety property, not just a philosophical curiosity.
That led into two problems I think are seriously under-discussed: how we verify that an AI agent is actually doing what it's supposed to do, and how we avoid the twin failure modes of over-attributing consciousness to systems that don't have it, or under-attributing it to systems that do. Getting either wrong has real consequences, for trust in AI systems in the first case, and for how we treat them in the second.
Listen to the full conversation on the Future of Life Institute Podcast>>
Calum on the Future of Life podcast
Calum Chace