Forgetting can happen slowly. In One Hundred Years of Solitude, Gabriel García Márquez describes how the town of Macondo drifts into collective amnesia. People forget the names of ordinary things – cow, banana, clock – and start hanging labels from them: “This is a cow, it gives milk.” “This is a clock, it tells the time.” The words remain, but the meaning fades. A society that records everything ends up remembering nothing. 

This year’s headlines feel uncannily similar: “Are we losing our ability to think?” “Are we forgetting the skills to choose for ourselves?” “Is there a decline in creative skills?” “How do we define intelligence?” Automation now labels the world for us: it is the infrastructure of daily life, shaping what we read, buy and decide. It’s natural to wonder whether, in the process, we’re forgetting ourselves. 

The global rush to close the ‘AI skills gap’ offers little reassurance. Governments and companies talk about reskilling millions to code or master prompts, as though fluency with machines were the same as understanding them. The harder task is remembering what makes human perspective so adaptable in the first place. 

The UK, for instance, has pledged to become an ‘AI superpower’, aiming to train millions in generative AI. The ambition is clear; the outcome less so. A workforce fluent in prompts may not be one rich in insight. The World Economic Forum expects tens of millions of jobs to vanish by 2030, even as 170 million new ones emerge. 

The vanishing on-ramp  

For generations, entry-level jobs were where people learned to work. You sat through routine meetings, fixed small mistakes and absorbed how things really got done. The training was the job.
 
That ladder now looks less stable. A new study from Stanford’s Digital Economy Lab – Canaries in the Coal Mine? – used detailed payroll data from the last few years from ADP, the largest processor in the US, to track how artificial intelligence is shaping employment. “At the level of the whole economy, you don’t see much happening,” says Stanford Professor Erik Brynjolfsson. “It’s only when you slice it more finely that the patterns appear.” 

Zoomed in, one pattern stands out. Among workers aged 22 to 25 in the most AI-exposed occupations, employment fell by double digits even as other groups grew. Bharat Chandar, another of the paper’s co-authors and former data scientist at Uber, notes that the evidence had been missing from official surveys: “The CPS [Current Population Survey] just doesn’t let you look closely at vulnerable groups like young software engineers or customer-service staff.” With the ADP data, “we could observe a lot of people within these groups and get pretty reliable estimates.” 

The authors stop short of claiming causation. “We don’t have an experiment,” Brynjolfsson cautions. They tested rival explanations – the pandemic, tech over-hiring, interest-rate shocks – and the pattern persisted. For now, AI “seems to be the factor most correlated with what we’re seeing.” 

Why are younger workers affected more? Chandar points to “tacit knowledge.” Early-career workers rely on what’s written down – “articles on the internet, books, Reddit posts” – the same material that trains AI. More experienced workers depend on judgment built from practice. That experience is harder for a model to copy. 

If AI is erasing the tasks that once taught novices, we may need new ways to teach judgment. Brynjolfsson sees promise in how companies use the tools: “When people use AI to augment rather than replace, it’s associated with more employment.” The on-ramp may be narrowing, but it need not vanish – if we design it back in. 

The business of taste  

It’s in the doing – the messy, repetitive, uncomfortable, sometimes tedious tasks – that fluency forms. As philosopher Hubert Dreyfus argued in Mind Over Machine, “knowing-how is the way we deal with things normally […] we take actions without using conscious symbolic reasoning at all.” If machines are learning our routines, what remains scarce is the skill they can’t absorb: human judgment. 

The World Economic Forum lists leadership, systems thinking and analytical reasoning as the top skills of the future. Yet beneath all three sits something harder to measure – a cultivated discernment that shapes every good decision, according to Nitin Nohria, former dean of Harvard Business School. He calls it taste: “the fusion of form and function, the ability to elevate utility with elegance […] the human fingerprint on decision making.” 

Researchers warn of ‘model collapse’ in machine learning – when systems trained on their own outputs start to lose originality. Work can suffer the same fate. If every decision is optimised for pattern, surprise disappears. Taste becomes a reminder that progress depends on difference. 

Pablo Picasso captured this with characteristic bluntness. “When art critics get together they talk about Form and Structure and Meaning. When artists get together they talk about where you can buy cheap turpentine.” Behind every refined judgment lies the grit of experience – the smell, the stain, the imperfect brushstroke. 

Panning for gold   

If taste is the art of discernment, then curiosity is its method. It’s not just about what we choose, but how we search. “If you over-rely on AI and totally outsource your thinking, you often get worse results,” says Tom Chatfield, tech philosopher and author of Critical Thinking. 

And, it seems, the opposite is also true: regularly tapping into your critical thinking will make you sharper. London cab drivers, for example, have measurably larger hippocampi than GPS users, thanks to a life of constantly searching for a better way around a routine journey. The way we navigate the world changes how we think. We can either take shortcuts, relying on tools that do the work for us, or we can build depth through deliberate effort. 

MIT researchers call the hidden cost of shortcuts ‘cognitive debt’: trading long-term capacity for short-term ease. The same tension applies to how we gather information. As M. Neil Browne and Stuart Keeley put it in Asking the Right Questions, some people “sponge” up data passively, while others actively “pan for gold,” searching for insights that matter. 

Leaders increasingly recognise the value of this mindset. As journalist and author Warren Berger notes, CEOs want managers who are more aware of what they don’t know, who ask harder questions like: Why do we work this way? and Where might new opportunities lie? Panning for gold is effortful, sometimes slower, but it strengthens the very discernment that shortcuts erode. 

The work of staying believable    

Discernment is not only cognitive – it’s emotional. Artificial intelligence can now mimic voices, write scripts and simulate empathy. That makes the subtleties of human connection as vital as the subtleties of thought. The next frontier of judgment lies in how we sound, feel and relate. 

In customer service, healthcare and sales, the ability to sound believable has become part of the job. According to Market Research Future, voice generation tools are already worth more than $5 billion worldwide and growing fast. The technology can reproduce tone, accent and even emotion. Yet it has created a reversal: human workers are now asked to prove they are real. According to Bloomberg, call-centre staff report being told to make small jokes, describe what they see around them or share something personal in exchanges to earn trust.