Some metrics for judging how hyped AI is

If AI is as much of a game-changer as many now claim, its real-world effects should soon be readily apparent

Here's how [Anthropic CEO] Amodei and others fear the white-collar bloodbath is unfolding:

1. OpenAI, Google, Anthropic and other large AI companies keep vastly improving the capabilities of their large language models (LLMs) to meet and beat human performance with more and more tasks. This is happening and accelerating.

2. The U.S. government, worried about losing ground to China or spooking workers with preemptive warnings, says little. The administration and Congress neither regulate AI nor caution the American public. This is happening and showing no signs of changing.

3. Most Americans, unaware of the growing power of AI and its threat to their jobs, pay little attention. This is happening, too.

And then, almost overnight, business leaders see the savings of replacing humans with AI — and do this en masse. They stop opening up new jobs, stop backfilling existing ones, and then replace human workers with agents or related automated alternatives.

The public only realizes it when it's too late.
Jim VandeHei, Mike Allen, Axios, 28/5/25

Here is a technology that [Amodei] believes is the most transformational thing since electricity or whatever, capable of doing hundreds of millions of humans’ jobs within the next few years, and all he can suggest is that governments should “prepare” for the job loss, and maybe institute a 3% tax on AI. Altman used to talk a little bit about a universal basic income—the bare minimum for gesturing towards an interest in the lives of the losers of the AI automation era—but he doesn’t even do that anymore. Now it’s nothing, except the occasional grim suggestion that the social contract itself might have to be rewritten in the AI companies’ favor.
Brian Merchant, Blood in the Machine, 31/5/25

The shattering of optimism is a dangerous moment in the life of any democracy. Disillusionment with the promise of progress was a major factor in the crisis of Europe that began in 1914 and ended in 1945.
Jonathan Sumption, NYT, 2/6/25

In mid-2020, I wrote an article warning those still employed in the legacy media that their future didn’t look bright and that they best start working on a Plan B. For the past three years, in these very digital pages, I’ve warned my left-liberal Professional Managerial Class (PMC) peers that continuing to double down on mass migration, asymmetrical multiculturalism, identity politics and censorious illiberalism would inevitably summon up an equal and opposite reaction.

I’ve been proven completely wrong with both my earlier predictions, as many critics at the time warned would happen. Bloodied but unbowed and hoping it will be a case of third time lucky, I’ve spent the first half of this year trying to convince people of “the growing power of AI and its threat to their jobs”.

I’m hardly the lone ranger there. Institutions from the World Economic Forum to the OECD, public intellectuals such as Daron Acemoglu, Nick Bostrom, Yuval Noah Harari and Robert Reich, and tech industry heavy hitters such as Sam Altman, Bill Gates, Mo Gawdat, Tim Cook, Jack Ma, Elon Musk, Satya Nadella and Sundar Pichai have all warned, more or less gently, that AI-driven mass automation has already kicked off.

The sceptics vs the true believers
People can be separated into two broad camps, which can further be divided into subcategories.

The fundamental distinction is between those who believe that AI will change everything and those who are sceptical about that claim. Among the true believers, there are those excited about AI creating a ‘post-labour’ utopia and those worried it will midwife a dystopia that gives self-interested elites – be it the CCP or Silicon Valley tech bros – unimaginable power over the little people.

The sceptics can likewise be divided into the aggressive sceptics who suspect AI will turn out to be yet another over-promoted nothing burger and mild sceptics who are open to the idea AI will have an impact but doubt it will be as sudden or revolutionary as its boosters’ claim.

Let me declare an interest. I’m a true believer who believes the decisions made over the next couple of years will determine whether AI-driven mass automation uplifts or immiserates the masses.

With reference to Amodei’s white-collar bloodbath schemata, I’d argue we are past Stage Three (unaware of the growing power of AI and its threat to their jobs [they] pay little attention) and well into Stage Four (They stop opening up new jobs, stop backfilling existing ones, and then replace human workers with agents). I’m hoping against hope we don’t get to the final stage (The public only realizes it when it's too late).

But there’s always the possibility my fellow true believers and I are wrong. Perhaps we are grifting. Who’s to say we don’t have equity in an AI company? Or that we’re not angling for cushy sinecures as AI regulators? Or not trying to score a hefty research grant?

You can never be too careful about these things. That’s why I’m suggesting the sceptics and the true believers agree on some metrics to gauge whether AI-driven mass automation of white-collar roles is occurring.

Let me get the ball rolling by pointing to one metric that buttresses sceptics' claims. Unemployment rates remain at historic lows in many technologically advanced nations. If they stay at historic lows, that would indicate AI is not hoovering up knowledge-worker jobs. Or at least that redundant knowledge workers aren’t having trouble finding alternative employment.

While the unemployment rate is probably the most visible and vital metric, it would be risky to rely solely on it. After all, it could go up or down for reasons unrelated to AI-driven automation. So, let me nominate a range of metrics to determine whether AI-driven mass automation is happening.

The do/don’t believe the hype metrics
I’ve tried to play a straight bat here, but if anyone – especially sceptics – feels any of these criteria are unfair, let me know in the comments.

Employment-related metrics
Hiring freezes: It’s a lot easier to just not hire someone in the first place than it is to get rid of them down the track. That being the case, you would expect CEOs to institute hiring freezes as a first step. There’s plenty of anecdotal data suggesting this has been occurring. Nonetheless, it doesn’t appear to be showing up – at least not clearly – in the data. However, if many businesses have effectively stopped recruiting, this should be evident across the second half of this year.

Flat wage growth: Given the last half-century, this isn’t quite the metric it once was. Nonetheless, if the balance of power is swinging decisively back to Capital and away from Labour, you wouldn’t expect those who escape the scythe to be pushing hard for pay rises.

Fewer job ads for knowledge workers: If businesses are no longer in the market for knowledge workers, they have no reason to be advertising for them.

Hot demand for ‘digital transformers’: Along with things getting much grimmer for run-of-the-mill knowledge workers, you’d expect them to get much better for those AI-savvy workers capable of helping organisations to expunge their human workforces and replace them with AI agents. You’d expect there to be lots of job ads for those workers and for them to be able to name their price.

Increases in revenue-per-employee (RPE): If you work in marketing and advertising, your employer typically makes US$150,000 - $300,000 a year from your labours. In tech, the average range is US$600,000 – $1,500,000. True believers expect AI-driven automation to increase these figures by 15-100 per cent.

Surging productivity: For two decades, advanced economies have been bedevilled by sluggish productivity growth. (It’s averaged 0.5-1 per cent, rather than the 2-3 per cent standard from 1950-2000.) If AI-driven mass automation is happening, the productivity rate should soon return to at least its post-war average.

Growing underemployment: As Peter Turchin regularly points out, Anglosphere nations have spent decades churning out ever more university graduates without worrying overly much about generating the reasonably paying, reasonably prestigious jobs grads expect. But I’m not talking here about the existing surplus elites – those much-referenced baristas with PhDs in Gender Studies, for instance. I’m talking about software developers working for DoorDash because AI has recently automated away their high-paying, high-prestige jobs.

Straw-in-the-wind metrics
The following metrics are less quantifiable but could be seen as evidence that AI has disrupted the labour market and is starting to impact the broader economy and society.

Educational institutions pivoting: I’ve lost count of the despairing think pieces about ChatGPT I’ve now read from demoralised academics and teachers. But as far as I’m aware, no prominent educational institution or education system has yet announced they’re returning to the drawing board and devising a curriculum fit for a post-Gen AI world. I’d expect those educational institutions most subject to market pressures, such as private schools, to start announcing such pivots over the next 12 months if mass automation is occurring.

The political agenda becomes all AI, all the time: In April, we Australians held an election campaign in which AI didn’t get a mention. I’m not being hyperbolic. The only time AI received any attention was when one of the parties created the first fully AI-generated Australian political ad. (The ad itself wasn’t about AI; it was about making petrol cheaper.)

If AI is turning the labour market upside down, you’d expect the representatives of Capital and Labour to be frantically manoeuvring to ensure the mass automation process benefits their constituency. Likewise, political parties on both the Right and Left would be forced to decide whose side they are ultimately on. The wealth-creating, productivity-turbocharging, sci-fi-future-summoning captains of industry? Or the imperilled (former) employee class? (Although many of them would bristle at the thought, most Professional Managerial Class types are eminently replaceable wage slaves. I fear they are about to learn the hard way just how replaceable they are.)

I’m writing this post shortly after the release of a “landmark” Business Council of Australia report outlining “practical actions to make Australia a globally recognised AI leader by 2028”. Not, you understand, to fatten margins and reduce personnel headaches. But rather to facilitate higher wages and “freeing people to do more of the high-value work machines can’t, like creativity, collaboration and care”.

The ACTU is apparently still working on its landmark report. But it seems a little less sanguine about the “freeing” possibilities of AI-driven automation. An ACTU bigwig recently thundered:

We risk a future where the rights fought for over generations by working people are undermined by the adoption of new technologies. The risks are clear: workers are being subjected to unreasonable unblinking surveillance, being hired and fired by algorithm, having their creative output stolen by companies, and being discriminated against by bosses’ bots. We need to ensure that these risks are eliminated while encouraging the development of technology that uplifts working people and any productivity benefits from its adoption be shared with working people that enable it.

Life comes at you fast
Few of my former colleagues saw it that way, but that 2020 article I wrote about the career prospects of legacy media employees was meant to be encouraging. Though most of the article detailed how the legacy media reacted haplessly to a game-changing technology, I ended on a happy note. After delivering the most unwelcome news that their old jobs were never coming back, I reassured my Fourth Estate readers that:

The good news is that a new career awaits. Once you stop clinging to false hope, you can set about reinventing yourself as an academic, account manager, agency owner, app maker, blogger, communications director, community manager, copywriter, content provider, e-book author, entrepreneur, ghostwriter, media adviser, playwright, podcaster, political staffer, proofreader, radio presenter, researcher, SEO expert, social media marketer, technical writer, trade mag editor, website designer or videographer.

Unfortunately, I can’t end this article in a similarly upbeat manner. Several of the roles I referenced a mere five years ago have already been largely automated away. Most of the ones remaining will be largely or entirely automated before the decade is out.

If the predictions we true believers are making are accurate, that is.

So, let’s agree to reconvene in six months to examine the available data.