Summary
DeepSeek’s AI breakthrough highlights an urgent reality: as the cost of running advanced AI plummets, human labour—with its wages, benefits, and downtime—faces imminent redundancy as the economic incentive to automate gathers pace. The focus on DeepSeek’s model or training budget claims misses the real story: AI’s overall trajectory is shifting work incentives on a global scale.
AI Costs Near Zero: Breakthroughs like model distillation mean AI is rapidly outpacing human labour.
Work Ethic Becomes Irrelevant: Even a diligent workforce can’t match AI that learns continuously and incurs no overheads.
Jevons Paradox: Lower costs spur heavier AI use, quickly automating high-value cognitive tasks once considered safe.
Societal Upheaval: Rising automation threatens to leave a generation without work, potentially triggering debates on Universal Basic Income and sparking wealth inequalities.
Opportunity Amid Disruption: Successful organisations will blend AI’s capabilities with human-centred roles and web3 collaborations, offering more creative, flexible ways to work.
The core message: stakeholders—governments, businesses and individuals—must confront these shifts or risk being left behind as the future of work rapidly evolves.
Media Missing The Point
The noise across traditional, social and independent media about the impact of DeepSeek’s R1 model on AI stocks and US AI dominance has been deafening, but it has all missed the point—a point that will impact everyone. We’re witnessing the cost of access to powerful AI-driven technologies tend towards zero. Marc Andreessen called the open-source DeepSeek a “…profound gift to the world” as the cost of access to powerful AI becomes increasingly trivial.
At the same time, the UK media has also been awash with commentators explaining that 9 million people who are not in work lack a work ethic and that overseas workers are cheaper and work harder. This also misses the point that, as the cost of AI technology approaches zero, this competitive advantage will evaporate. An individual’s work ethic becomes irrelevant if there is no longer demand for their labour.
I had a discussion about this on the UK's LBC Radio.
A Crucial Tipping Point
We’ve reached a crucial tipping point that commentators, pundits, politicians and business leaders are missing—primarily because, although the impact will be profound, it is not immediate. This isn’t about whether DeepSeek spent just “a few million” versus billions on training its model, whether it was a trade by the hedge fund behind DeepSeek to short US tech stocks, or whether its reasoning model is better than OpenAI’s. It’s about the rapidly diminishing cost of training and deploying AI models, and the growing number of powerful open-source models able to compete with proprietary US offerings.
The cost of AI inference is heading one way: down. Meanwhile, the cost of employing people remains stubbornly high, and is even rising. Wages, pensions, taxation, healthcare and training all add up in ways that simply don’t apply to an AI. Nor do holidays, illness, weekends, evenings, or bad days.
Dario Amodei’s Davos Perspective
At Davos, Dario Amodei, CEO of Anthropic, emphasised that advanced AI systems are on track to surpass humans in almost every cognitive domain. He also pointed to DeepSeek’s rapid rise as just one more sign of how quickly this transformation is happening. Even if DeepSeek’s budget claims have been overstated, Amodei cautioned, the overall trajectory is clear: the cost of training, running and refining AI models is plummeting.
AI Model Distillation
DeepSeek may have used AI “model distillation,” which will accelerate the rollout of ever-smarter and cheaper models built on the work—and investment—of precursors. Traditionally, training a cutting-edge AI model took huge swathes of data (e.g. the entire internet) and astronomical compute budgets—tens to hundreds of millions of pounds. Now, with distillation, new models can effectively “read the minds” of more advanced models, absorbing their knowledge in a fraction of the time and at a fraction of the cost.
As well as using a mixture-of-experts (MoE) approach, this is believed to be how DeepSeek was able to use existing and more advanced models to train its own at a fraction of the cost. Distillation could trigger an exponential growth loop. Each new “master” model spawns an army of “student” models specialised for particular industries or tasks, each becoming cheaper yet more powerful in short order. While not autonomous self-improvement, the economic incentive to build such models may still produce a similar result.
The Impact of AI on Work Ethic
What work?
At the same time, the UK’s work ethic is under scrutiny compared to emerging economies where labour is cheaper, and employees are said to work longer hours for less pay. This is inextricably linked to the development of AI, which displaces both cognitive and, eventually, physical labour.
If entry-level or low-wage roles are disappearing, the remaining jobs will demand higher levels of skill, experience and the ability to collaborate with AI in augmented human–machine teams. Those who choose not to work, or who cannot secure work in competition with AI, may never work at all.
At Implement AI, we’ve seen that AI agents can often be cheaper than human resources in the UK and the USA. Certain tasks can still be carried out by cheaper labour in countries where wages are much lower, but if we run DeepSeek on our own infrastructure, the cost can drop by an order of magnitude—and no human worker, anywhere in the world, can compete.
If advanced AI is going to handle both routine and high-level cognitive tasks—at minimal incremental cost—what happens to the work available and the incentive to work at all? You could be the hardest-working person on Earth, but you can’t outpace an AI that never sleeps and learns at hyper speed. This pivots the conversation away from “Who works harder?” and towards “Who can create the most meaningful and adaptable work environments?” The real winners will be those who blend technological capability with human-centric, purpose-driven work models.
Jevons Paradox and the Plummeting Value of Cognitive Work
Jevons Paradox states that when technological improvements make a resource cheaper or more efficient to use, overall consumption of that resource often increases rather than decreases. This was true of steam power, and we’re seeing a replay in AI. As costs for training and running models plummet—via advances such as distillation, more powerful processors and efficient datacentres—adoption will surge. Lower costs will spark a wave of new use cases and more intensive usage, as organisations flock to automate tasks once too expensive or complex to delegate to machines. This undermines any national competitive advantage offered by low-cost cognitive labour.
While the media often focuses on physical automation (robots), the immediate revolution lies in accelerating the automation of cognitive roles—coding, legal research, financial analysis, design and other highly skilled tasks. Global GDP sits at around £85 trillion (roughly $110 trillion), half of which is knowledge work. In the long run, AI could capture a sizable portion of that, delivering it at a fraction of the current cost.
The Social Consequences: UBI and Wealth Crystallisation
The implications of wealth crystallisation extend beyond simple income inequality. We may witness a fundamental restructuring of economic mobility, where traditional paths to wealth through skilled labour become increasingly constrained. This raises urgent questions about the future of innovation and economic dynamism in a highly automated world.
If AI can handle the bulk of today’s labour—cognitive or physical—what happens to people left with nothing to do? How does one maintain a work ethic with no work available? One proposed solution is Universal Basic Income (UBI), where every citizen receives a guaranteed payment to cover essential living costs. While this could prevent widespread poverty, it raises crucial concerns: would UBI undermine personal ambition or devalue the notion of personal development? And who pays for it?
Wealth may also crystallise at the top. If a handful of large companies—and the nations supporting them—own the most sophisticated models, they will reap enormous profits. The rest might receive enough to subsist but never gain real upward mobility. Such structural imbalances can lead to social unrest and cultural stagnation—potentially on a permanent scale this time.
A Lost, or the Last, Working Generation?
A dystopian possibility is that an entire generation will never experience “work” in the traditional sense. Beyond earning a wage, work confers identity, social mobility, purpose and a sense of collective contribution. If AI does most jobs more efficiently and cheaply, how do young people build skills, confidence and self-worth during those formative years?
Younger generations entering the workforce increasingly expect such technological empowerment. Cognitive work is seen as something you do, not somewhere you go. They seek roles where they can exercise autonomy, pursue meaningful goals and leverage cutting-edge tools to amplify their impact. Organisations that fail to deliver risk losing top-tier talent to more forward-looking competitors.
We need to ask ourselves what uniquely human roles will persist. Perhaps a future renaissance will arise in emotional intelligence fields—counselling, therapy, the arts, community-building—areas where AI has yet to match the nuances of human empathy and creativity. Alternatively, we could see a spike in entrepreneurial, hyper-local or artisanal ventures. But these niches won’t absorb billions of displaced workers.
Re-imagining Work
There is no return to traditional employment models. As AI becomes cheaper and more capable the challenge isn’t merely coping with AI’s rise; it’s harnessing multiple disruptive forces to engineer more adaptable, empowering and meaningful ways of working.
DeepSeek’s headlines might steal the limelight, but every month sees new innovations pushing down development and operational costs. This combination of relentless progress and intensifying competition signals a looming crisis—or an unprecedented opportunity—for businesses and governments alike.
We must confront the reality that human labour—especially knowledge work in developed economies—will lose its competitive edge as AI outperforms and underprices it. Low cost overseas labour will follow as the cost of AI continues to fall. Governments will struggle to safeguard traditional jobs; businesses will be compelled to automate to remain competitive; and individuals must keep ahead of the encroaching technological wave. UBI may lie on the horizon, but that brings a fundamental shift in how we value personal contribution and societal worth.
For the next generation, the most troubling outcome is that they never step onto the employment ladder. If that’s where we’re heading, we need to rebuild our incentives and frameworks to ensure people develop skills, chase goals and find purpose. Work might look drastically different in five, ten or twenty years, but one thing is clear: those who embrace and master AI stand to shape the future economy, while everyone else risks being left behind.
The big question is: which side of that equation will you be on?
Thanks for reading.
