Despite AI’s widely publicised successes, there are very good reasons to be sceptical about its possibilities for replacing humans, says one observer at NTU.
The last several months have seen Singapore swept up in an enthusiasm for artificial intelligence (AI).
In April, NTU engineers succeeded in building a robot that could assemble an IKEA chair. And over the last few months we’ve heard much about “Nadine” – an AI with a human face, designed to respond to human speech and emotions.
Prime Minister Lee Hsien Loong recently flagged AI as an area Singapore is making good progress developing frontier technologies in, noting especially Alibaba’s establishment of a joint research institute at NTU.
But beyond these headline-making, attention-grabbing events, how likely is it that AI will be transforming our lives in the near future?
SOLUTION TO ALL OUR PROBLEMS?
The discussion about AI is almost always dominated by either “techno-optimists” or the “doomsayers”.
For the former, AI will solve all the world’s problems, bringing wealth, new jobs, new industries, and even strengthen social cohesion; for the latter, robots pose a threat to people’s livelihoods and ways of life.
But both these views take on faith the underlying idea that AI is coming, sooner or later, like it or not. These positions leave very little room for reasoned consideration of how transformative AI may really be.
Despite AI’s widely publicised successes, there are very good reasons to be skeptical about its possibilities for replacing humans.
The history of AI is hardly a history of triumph. Since the first electronic computers were built during World War II, their creators thought of them as giant electronic brains, promising that they would soon be painting great works of art, solving mathematical proofs, and directing armies into battle.
Very little of this materialised. The early pioneers of AI in the 1950s and 1960s soon came to understand how difficult building “thinking machines” could actually be.
Sure, AI has made progress since then – it can now beat humans at Go and chess and help us find our way to the nearest supermarket or movie theatre.
But history suggests we should be cautious about extrapolating too far from these beginnings to draw straight line conclusions about the pervasiveness of AI in our lives.
Games with very clear rules and boundaries are one thing. But, as the problems with autonomous vehicles are beginning to show, dealing with the openness of real world situations is quite a different matter. City streets – filled with pedestrians, children playing football, bicycles, and parades – are a far cry from a chess board.
Both personal and national security increasingly rely on AI-driven technologies of facial recognition. But recent reports have suggested that such technologies are plagued with difficulties.
Reports from China suggest that the iPhone X cannot distinguish between the faces of some ethnic Chinese people. Other facial recognition schemes have completely failed to identify individuals with darker skin tones, differentiate between twins or recognise someone who has changed their make-up.
Even in the realm of linguistic communication, AI has experienced some revealing difficulties. Microsoft’s chatbot “Tay” had to be shut down after developing a penchant for racist remarks; and Facebook’s “Bob” and “Alice,” despite being programmed to communicate in English, seemed to generate their own private language.
We also need to ensure that the promises of AI do not distract us from pursuing other kinds of solutions to social and economic problems.
For example, Singapore has invested in AI and smart city systems as a solution to the problem of an ageing population, including devices to monitor elderly citizens in their homes.
But this seems unlikely to be a panacea – such efforts should go hand in hand with the training of elderly care workers and investment in aged care facilities. Sensors and smart devices need to connect the elderly to real people who can provide human care in a sustainable and scalable manner.
Rather than giving in to premature triumphalism or panicking about robots taking our jobs, Singapore should be thinking about how AI research might be most productively channeled into fields where it has the greatest potential.
This approach will allow Singapore to become a useful use case study for other cities and nations that are also trying to figure out also how to unlock AI’s potential.