A Robot Just Debated Humans on the Benefits and Risks of AI, Tells Audience 'AI Can Cause a Lot of Harm'

An artificially intelligent machine took to the podium at the Cambridge Union in the U.K. to debate the pros and cons of AI with humans—at one point telling the audience "AI can cause a lot of harm."

However, it was the side in favor of AI that came out top, achieving a narrow victory by scoring votes from just over half the audience.

According to Fortune, the AI's arguments were extracted from over 1,100 submissions sent in to IBM the week before the debate, which were categorized as being in favor or opposed to AI, or discarded entirely as being irrelevant to the debate. From these, IBM's Project Debater searched and condensed the strongest and most diverse arguments to repeat when given a short opening sentence as a prompt.

The case for the benefits of AI included the argument that it could generate new jobs and bring greater efficiency to the workplace, while those against included the problem of human biases and AI being incapable of making moral decisions.

"AI can cause a lot of harm," it told the audience, New Scientist reports. "AI will not be able to make a decision that is the morally correct one, because morality is unique to humans."

But the algorithm's performance did not run entirely smoothly—it had a tendency to repeat itself at times—including when talking about AI's ability to complete repetitive tasks—and it used an argument on bias as a benefit rather than a criticism of AI. It did, however, manage to throw in a couple of laughs, plucked from its "bank of jokes," CNN reports.

It is not the first time the AI has performed in front of humans. It did so in February, when it lost in a head-to-head battle against Harish Natarajan, who is the world record holder for most debate competition wins. Natarajan also performed at the Cambridge Union, for the pro-AI side.

Dan Lahav, an IBM computer scientist who works on Project Debater, told reporters at Fortune some people sent in obscenities and racist language seemingly so that the AI would repeat them during the debate. However, unlike algorithms like Microsoft's Tay, it was able to snuff out discriminatory and inflammatory language.

The team behind the machine hope it can be used as a tool to collect feedback from large groups—for example, by governments seeking public feedback or by companies wanting to understand their reputation among potential customers.

When it comes to how the U.S. public perceive AI, the majority seem to be largely in favor—in a 2018 Gallup poll, 79 percent of respondents agreed AI has had a "mostly positive" or "very positive" impact on their lives. However, 73 percent expected AI to eradicate more jobs than it generates. The majority (63 percent) also expected AI to exacerbate economic inequality.

AI experts are also conflicted when it comes to the harms and benefits of AI, with many saying its future will be determined by how we humans program the algorithms to operate.

"I think it is more likely than not that we will use this power to make the world a better place. For instance, we can virtually eliminate global poverty, massively reduce disease and provide better education to almost everyone on the planet," Erik Brynjolfsson, director of the MIT Initiative on the Digital Economy and author of "Machine, Platform, Crowd" told Pew Research.

"That said, AI and ML [machine learning] can also be used to increasingly concentrate wealth and power, leaving many people behind, and to create even more horrifying weapons. Neither outcome is inevitable, so the right question is not 'What will happen?' but 'What will we choose to do?'"