Low- and middle-income countries (LMICs) must build expertise in AI and share their knowledge with the wider public, to avoid being overrun by the technology.
That’s the advice of French mathematician and philosopher Daniel Andler, emeritus professor and honorary member of the Institut Universitaire de France.
AI is already transforming science in LMICs but these countries are also most at risk from the “steamroller of uncontrolled AI”, says Andler in an online interview with SciDev.Net on the promise and pitfalls of AI around the world.
“To counteract the threat, they need to develop local expertise in building AI systems, getting them in running order, deploying them and educating the general population to their use,” he says.
If AI is used without professional training, there is a danger of “duplicate or unreliable results”, he says.
Andler is optimistic about the role of AI in international development “as a tool to organise the collective intelligence of scientists … fight misinformation … and facilitate constructive dialogue”.
He says it provides an “enormous boost” for scientists trying to solve global challenges.
Deskilling
Howewer, speaking in January at the International Science Council’s Muscat Global Knowledge Dialogue, in Oman, Andler had warned that AI was beginning to have a significant impact on scientific research, acting as a collaborator in the scientific process, and that it posed a risk of scientists becoming deskilled as a result.
“What if scientists rely on technology and then it fails?” Andler tells SciDev.Net during the online interview.
“Part of the scientific research process depends mainly on critical thinking.
“So, we need to find a method to ensure that technology use incorporates this type of thinking.”
As a solution, Andler proposes a holistic approach to using AI in research, with scientists conducting research both with and without the technology.
“Scientists can improve their skills when they engage in real research,” he says, adding that “researchers must avoid using AI when moving from a blank page to writing the very first page of their papers”.
“They should also avoid it when finalising their papers. In between these two stages, AI can offer multiple advantages.”
He also warns against granting too much autonomy to AI, which is already beginning to influence our way of seeing the world.
“In today’s world, our reality is largely shaped by market forces. Tomorrow, our reality will increasingly be shaped by AI, because AI will play a role within our thinking processes.
“AI inevitably brings a set of assumptions regarding how the world works and how it should work. Together, they form a worldview which stems from the culture in which AI systems are manufactured.”
Short leash
Andler explains that we are already experiencing how businesses, personal relationships, and even political ideas increasingly follow uniform patterns shaped by social networks and platforms—pushing aside local cultures.
“The more autonomy we grant to AI, the more pressure it will put on our way of making sense of the world we live in, and on our values and preferences… This is why it is so essential that people, individually and collectively, keep AI on a short leash.”
The rapid development of AI concerns Andler, who says that if it’s used by the wrong people, it may cause more harm than good.
Andler believes that AI developers respond to what is called the ‘Maslow’s hammer’ rule—a cognitive bias that involves an over-reliance on a familiar tool.
“For them, everything looks like a nail that has to be hammered, and this makes them somewhat shortsighted,” he says.
He adds that AI development should comply with some sort of ethical code, to help govern its development.
“People think that AI poses existential risks to humanity, but I don’t believe so,” he says.
“However, if AI becomes powerful and exclusive to authoritarian leaders, they may impose a reality that we may not like.”
Will AI ever substitute for real scientists? Andler doesn’t believe this either.
“AI makes better use of data than humans. However, scientific research doesn’t revolve solely around data.”
This piece was produced by SciDev.Net’s Global desk.