With ChatGPT, AI takes aim at the professions

Hey there, time traveller!
This article was published 11/02/2023 (986 days ago), so information in it may no longer be current.

The World Economic Forum held its latest gilded gathering in Davos, Switzerland, beginning Jan. 16, amid a panoply of global crises for politicians and business leaders to discuss.

Reportedly top of mind for many attendees, however, was ChatGPT — a free new tool powered by artificial intelligence (AI).

Released to the public by San Francisco-based tech company OpenAI in December, ChatGPT has elicited equal parts awe, curiousity and horror from countless people who may have considered themselves immune to the accelerating disruption being caused by AI.

ChatGPT represents an exponential leap in the progression of natural language processing models. These machine-learning programs can, on their own, produce strikingly insightful responses to human interaction, based on untold training hours spent scanning billions of pages of text harvested from the internet.

For example, among many other things, ChatGPT can fix computer code, write undergraduate-level essays, dissect public policy, compose poetry, act as a therapist, craft press releases or explain abstract philosophy or the function of mRNA vaccines in seconds, based on simple user inputs.

The program joins others that comprise generative AI — machine-learning models that go beyond detecting patterns and making predictions based on data, to generating bespoke content including art, videos and music.

The prevalence of AI in the industrialized world is already inescapable. A former chief business officer within Google estimates the average person interacts with AI software at least 15 times a day without even knowing it.

But with generative AI, computers are now encroaching into the realms of human creativity and expression once perceived as untouchable by machines. In doing so, programs such as ChatGPT have stoked fear among knowledge workers and artists, who may not have heeded the recent rapid advances in automation disproportionately affecting low-skilled workers in manufacturing, hospitality, retail and logistics.

And yet, for all of ChatGPT’s wondrous output capability, users have also uncovered its profound flaws, highlighting the general limits of machine-learning systems — limits that appear exceedingly difficult, if not impossible, to overcome.

AI is remarkably adept at executing narrowly defined tasks, even if those tasks are incredibly complicated, so long as objectives are codified and programmed accurately and reinforced properly during the design and training phases of a model’s development.

What AI remains unable to do is negotiate ambiguity by using intuition or common sense to adapt its processes on the fly when encountering a novel situation not factored into its training data.

This includes understanding simple cause and effect, or applying an ethical filter to its outputs. In contrast, these cognitive capacities are developed in humans from an early age, through social and cultural experience and exposure to the infinite number of variable dynamics and interconnections that make up daily life.

For example, one technologist notes how ChatGPT “implied that it would be possible to bike from San Francisco to Maui, and struggled with simple logical problems, the kind a second-grader could solve.” Other users got results from ChatGPT that were deeply racist, such as suggesting citizens from Iran, Syria and North Korea were prime candidates for torture.

Another tech critic points out, “Nobody really knows how or why these AIs produce their results — and the outputs can change from moment to moment in inexplicable ways. Once that first blush fades, it becomes clear that ChatGPT doesn’t actually know anything — instead, it outputs compositions that simulate knowledge through persuasive structure.”

None of this is meant to dismiss the way AI and related technologies such as big data and quantum computing will be fundamental to solving humanity’s biggest challenges, including mitigating the climate crisis.

Optimists argue these new tools will ultimately create more jobs than they threaten — and they could be right.

But amid the breakneck speed of contemporary technological progress, combined with the new-age religiosity emanating from Silicon Valley, it’s often forgotten that not every problem requires AI or an automated solution, or how the rush to adopt them may actually aggravate pre-existing inequities.

In many ways, the unfettered proliferation of nascent technologies is already upending how societies operate in ways that risk severe political consequences.

In 2018, the co-director of the Stanford Institute for Human-Centred AI wrote: “A human-centred approach to AI means these machines don’t have to be our competitors, but partners in securing our well-being. However autonomous our technology becomes, its impact on the world — for better or worse — will always be our responsibility.”

This demands emphasizing the ethical design and deployment of machines that, at least for now, cannot think for themselves — or on behalf of humans.

Kyle Hiebert is a Winnipeg-based researcher and analyst, and former deputy editor of the Africa Conflict Monitor.