State agency Callaghan Innovation’s move to pilot GovGPT, an artificial intelligence chatbot to help you deal with government bureaucracy, is the first step in the automation of the public sector.
GovGPT will initially help businesses navigate government departments, pulling together information from ACC, IRD, and other agencies in response to questions. That makes a lot of sense. If banks and airlines are rolling out chatbots to answer low-level customer questions, why shouldn’t the government?
But a chatbot based on publicly available information is marginally useful, merely saving time googling your own answers and trawling websites. AI that can look at your tax obligations, benefit payments and health records, and suggest ways to access more support or reduce your admin when dealing with the state, will unlock a lot more potential.
It’s technically possible now, but would involve a level of data-sharing across government that isn’t currently permitted. Public buy-in must be gained to allow that. An incremental approach is needed, proving that chatbots won’t hallucinate, leak sensitive data or misinform citizens, before we apply them to analysing our personal data and making suggestions or decisions.
But technologists see huge scope for use of AI to improve efficiency in provision of government services.
“I think you could remodel most of Aotearoa’s government bureaucracy as a software API [application programming interface] which would do the same or better job, cost a fraction of the current overhead to run and free up the people inside it to do more productive, creative work,” writes Christchurch-based futurist and tech consultant Ben Reid in his book Fast Forward Aotearoa.
Reid’s view, echoed by influential historian Yuval Noah Harari in his new book Nexus, is that AI will trigger a turning point in human history. Reid refers to it as the AI-o-cene, and something that’s inevitable and needs to be embraced rather than resisted.
“A new humanity-plus-capitalism-plus-AI ‘superorganism’ is terraforming the planetary environment even more profoundly than 8 billion humans have to date on our own,” writes Reid.
But Harari cautions us to view AI as just another wave of new technology that, like the introduction of the printing press, and later, radio, TV and computers, is perceived as having done much more to advance humanity than damage it.
AI is different, he argues, as it raises the prospect of superintelligence emerging that we are unable to control. Even semi-intelligent AI poses significant risks. Harari asks: what would a 21st century Stalin do with AI?
“If the AI revolution leads us to similar kinds of experiments, can we really be certain we will muddle through again?” Harari writes in Nexus.
Wholesale automation of government processes would be music to the ears of Act’s David Seymour, a crusader in the coalition government’s push to cut public servant roles. Job displacement is one barrier to acceptance of AI in government.
But a bigger concern is the potential lack of transparency in decision-making processes, with AI systems acting as “black boxes” – their reasoning difficult for humans to interpret. While AI has the potential to reduce human bias, poorly designed or trained AI systems can perpetuate and even amplify existing societal biases.
So don’t expect the AI-driven government Reid advocates emerging any time soon. GovGPT, and other automation efforts already quietly underway in government agencies like ACC, are just baby steps on the path to extensive automation.
I’m with Reid, and Seymour for that matter, on the need to cut inefficiency in government. I’m impatient for change. We’ve precious few tax dollars to support what we currently do. But if you buy into Harari’s compelling view of the history of technology, there are enough hard lessons to warrant caution on AI.
