You could understandably forgive the government for wanting a magic wand it can use to wave away its economic woes. But thinking that AI will immediately be a net benefit shows a blinkered view which only buys into Big Tech’s preferred fictions.
As the government looks to turn around the country’s stagnant productivity growth, AI is being touted as one key pillar to address this.
Tech companies have opportunistically jumped on this conversation, with big promises around AI’s benefits, but only if we did away with things like copyright protections and proper regulation.
It was perhaps not surprising that US President Trump’s messages around AI parrots those of the Big Tech companies, given most of them are based in the US, and their ambitions align with those of President Trump’s.
But it was startling to hear that Australia’s own Productivity Commission started to mirror these messages, questioning the need for national AI regulation, and implying that copyrighted works, and therefore all the people whose jobs are tied into those creative works, should not get in the way of AI’s progress.
So, an existential question presents itself – as to whether AI’s development should be permitted unimpeded, but at the cost of whole sections of the economy and at the expense of an enormous group of workers and their valuable intellectual property.
This tension represents the uncertainty around the concept of AI’s impact and benefits on Australia’s productivity.
A new report shows that the whole picture is more complex than what AI hype mongers would have us believe.
Like previous general technologies such as steam, electricity and the internet before it, AI will likely take decades to be fully integrated into society, butting up against individual adoption, workplace configuration and the inconvenient realities of workplace inefficiencies.
Even a tech-forward company like Atlassian, who have gone all in on AI, revealed in their own research that the initial benefits of AI were compromised by existing organisational inefficiencies. In other words, AI doesn’t exist in a vacuum, and will come up against real-world workplace problems.
Further, Atlassian found that there was a huge disconnect between the high expectations of management in terms of AI’s benefits, versus the reality of worker experiences, who are the ones who ultimately have to deliver on its promise.
In multiple studies explored by the report, AI was actually found to create new tasks and increase workloads. Experienced coders found they needed extra time to review AI’s output and correct many of its mistakes. Teachers found that AI didn’t really understand their curricula, and marked students too generously. When AI wasn’t adding extra work, it hardly made a difference, as a huge longitudinal study conducted in Denmark over two years across 11 occupations has found.
On top of these sobering reality-checks, there are other issues that will be created when using AI, such as the ‘AI efficiency trap’ where the expected benefits become the new standard, and workers have to somehow find a way to increase their output whether it actually happens or not.
One startling study found that people who over-relied on AI experienced ‘cognitive debt’, impacting neural and cognitive processes related to learning, information retrieval and task management.
Not to mention, a lot of other issues like creating a lost generation when juniors and entry level workers are replaced by AI, impacting employee pipelines. There are also live debates about privacy breaches, unresolved issues around bias and misinformation, and environmental concerns around required infrastructure like data centres.
The litany of unresolved challenges is long. This isn’t to say that AI will not eventually produce economic benefits, but that there is a huge list of negative externalities that cannot be ignored in the meantime.
AI is, and will continue to be a complex general technology that will likely change the world, and that’s exactly why we can’t afford to cut corners and skip due diligence in favour of easy answers.
We need a national AI Act to account for the technology’s complexities, governed by a central body like an AI Commissioner to produce a landscape that’s responsible, ensure its harms are mitigated and its benefits are evenly distributed, serving the needs of the Australian public.
Jordan Guiao is Director of Responsible Technology at Per Capita, and author of ‘Disconnect: Why we get pushed to extremes online and how to stop it’