From our ‘Thinking outside the box’ series, Professor Matthew Beck takes a closer look at generative AI and the assumptions driving its rapid adoption. He raises important questions about who really benefits and what we might be overlooking.

This opinion piece grew out of an all-staff email that focused largely on the upside of generative AI, followed by a lunchtime conversation that was a little more balanced. The email listed the commonly cited benefits of the technology (efficiency, productivity, faster outputs, lower costs) but, as always, the potential costs were abstracted away.

This is a pattern that is not particularly new. History is full of technologies introduced with enormous confidence, engineers building things just because they can, and corporations proceeding with them because profits are all that count. Ethical questions, public consequences and social responsibility usually arrived later, often after the damage cause becomes someone else’s problem.

While artificial intelligence promises a rosy future, we must first understand it is often only as good as our collective past. Much of the data used to train it is historical, which means it carries all the race, gender, age and human prejudices of our past. Several high-profile AI systems have drifted quickly into racist, abusive or conspiratorial behaviour after exposure to poor data or weak safeguards. Machines, it turns out, can absorb bad habits remarkably quickly. More problematically, AI decisions are often harder to challenge because the machine is assumed to be objective and emotionless. Australia has already had a warning about what happens when automated systems are treated as an authority. Robodebt showed how damaging that can become when institutions place too much confidence in automated systems.

Just as problematic for humans is the voracious appetite of the data centres that support this technology. They are currently intensifying competition for urban land in Sydney, diverting space from housing and doing little to improve affordability (and there are currently no studies on what living near an industrial-sized data centre may do to humans)¹. Perhaps more worryingly, this technology is even hungrier for electricity and water. At what point will these data centres be “fed” in favour of humans, particularly in a country that has historically faced water scarcity and limited supply?²

Even leading figures in this industry are suggesting the public should grow comfortable with deeper dependence. Sam Altman recently proposed that artificial intelligence may eventually be treated like electricity or water, metered and paid for like a utility. It is a revealing idea. Public utilities exist because governments recognise that some services are too essential to leave entirely to private monopoly. The AI industry seems to imagine the reverse: privately owned systems becoming indispensable, with the public funding the infrastructure while companies retain control of the value, driven as much by commercial ambition as by a god complex.

Generative AI is driving major disruption in labour markets, with little evidence of a credible plan to manage workforce transition and employment impacts. In this light, the OpenAI public utility thought bubble is almost like paying the hangman for tying your noose. Atlassian has made substantial redundancies while investing heavily in automation. Commonwealth Bank, despite strong profits, has also reduced its workforce in favour of AI, and ANZ has followed similar paths. Shareholders generally approve of this sort of efficiency. Workers tend to experience it differently.

Those that think their job is safe from AI disruption may need to consider what will happen when swathes of displaced works retrain and begin looking for employment elsewhere. A surge of competition for jobs in healthcare, skilled trades and personal services will likely mean underemployment in those sectors, and almost assuredly downward pressure on wages. That is before the machines fully arrive there as well.

We also might think that a future without work seems idyllic, but for many people, work is not simply about earning a wage. Research has shown that work provides routine, purpose, and meaning, while also offering social contact, mental stimulation, and a sense of achievement that are closely tied to overall well-being.

The cost of managing this disruption is also being pushed onto institutions already under strain.  Already under strain, technical education providers and universities are expected to retrain workers, prepare younger generations and somehow absorb the transition. This huge change is occurring in a context where while Australia sits among the lower public funders of tertiary education across the OECD. It is almost as if private firms collect reap the rewards, while the public purse bears the cost.

There is a further economic perversity on the horizon. If firms reduce human labour, leaving the vast majority without meaningful employment, the obvious question is who continues purchasing the goods and services being produced. The wealth divide will only grow larger, and if you do not own any land or productive capital, it is not unreasonable to expect that you will be left further and further behind. It leads one to ask the question: can you barter with a machine? And if so, what would a machine want?

Many of the companies behind this infrastructure have perfected the art of extracting value locally while minimising what they return through taxation. They have no sense of social responsibility. GenAI simply offers another mechanism to increase profits, reduce labour costs, and avoid accountability, all while presenting disruption as innovation.

None of this should be surprising, as we have already experienced a similar phenomenon. Social media already showed how quickly technological optimism can become social harm: mental health impacts on children, corporate surveillance of daily life³, and private platforms shaping public behaviour with minimal accountability. These companies are perhaps the most adapt at privatising the gain and transferring all the pain to the public.

There is no obvious reason to expect generative AI to be exempt from the same logic. There are already examples of this technology being used in place of proper medical care. Class actions are pending over generative AI allegedly encouraging self-harm, including cases linked to death. When such harm occurs, the response is often the familiar corporate shift of responsibility back to the user, coupled with the assurance that the models are still improving. Just as disturbingly, generative AI is now being programmed to mimic voices and likenesses, meaning there is a non-zero possibility that your identity may be co-opted for corporate purposes, soulless and eternal, whether you like it or not.

Another problem with GenAI is data security. We keep being told to put more and more of our lives online, yet every few months another organisation seems to prove it cannot protect what it already has. Qantas, Optus, Medibank, and now the Canvas ransomware hack affecting university students and staff. At what point do we admit that a full picture of who we are is already floating around somewhere on the web? Your name, address, phone number, date of birth, health records, financial details, education history, passwords, voice, face and habits. All sitting there, waiting to be bought, stolen, copied, guessed, or stitched together by someone with the right tools.

And now we are adding GenAI to the mix, a technology that can write convincing scam emails, clone voices, impersonate people, find software weaknesses and automate cybercrime at a speed humans never could. Anthropic has reportedly developed AI so capable in cyber operations that it was considered too dangerous for general release. So where does this end? How long before we start moving backwards, not because we are nostalgic, but because the digital world has become too unsafe to trust? Maybe the future is not tap-and-go, cloud storage and banking apps. Maybe the future is cash under the mattress, paper records, and asking whether the “smart” society was actually very stupid.

The cynic in me (borne out by ample evidence) thinks that many of the companies behind this infrastructure have perfected the art of extracting value locally while minimising what they return through taxation. They have no sense of social responsibility. GenAI simply offers another mechanism to increase profits, reduce labour costs, and avoid accountability, all while presenting disruption as innovation. Capitalism assumes growth can continue forever, when in reality the world has finite resources. If land, water and energy are limited, we should ask whether they exist first for human need or for machine expansion.

None of this means AI should be rejected outright. While AI clearly has some benefits, this think piece is, by design, to take the position of devil’s advocate. However, a future with no jobs, no housing and no water is a future built more for machines than for people. We should not accept what is being pushed on us simply because corporations (with trillions of dollars at stake) say it is good for us. Nor should we assume the utopia they promise is certain (particularly when they’ve to us lied before). The decisions we make today shape tomorrow. We are not passive observers; we are active participants. We can decide what kind of future we accept and what kind we refuse. That is why we must ask harder questions about the future generative AI is creating, rather than simply accepting promises of profit, efficiency, and productivity, because these are not the only things that matter in life.

¹Never mind the fact that large tracts of what was once prime agricultural land, which supplied Sydney’s markets and restaurants with the fresh food we all love, will become industrial land and the site of the second Sydney airport. Once lost, it can never be recovered

²From a transport perspective, we also need to think about the role of AI. With self-driving cars, we may be embedding further inequity in terms of which countries can support the processing and energy demands of automated mobility, given the limits of their water and electricity networks.

³Smart speakers listen to you all day, every day. AI reads all your cloud-based email. Worse yet, did you know your robot vacuum cleaner can be used to see you naked?