A.I. researchers urge regulators not to slam the brakes on its development

LONDON — Artificial intelligence researchers argue that there’s little point in imposing strict regulations on its development at this stage, as the technology is still in its infancy and red tape will only slow down progress in the field.

AI systems are currently capable of performing relatively “narrow” tasks — such as playing games, translating languages, and recommending content.

But they’re far from being “general” in any way and some argue that experts are no closer to the holy grail of AGI (artificial general intelligence) — the hypothetical ability of an AI to understand or learn any intellectual task that a human being can — than they were in the 1960s when the so-called “godfathers of AI” had some early breakthroughs.

Computer scientists in the field have told CNBC that AI’s abilities have been significantly overhyped by some. Neil Lawrence, a professor at the University of Cambridge, told CNBC that the term AI has been turned into something that it isn’t.

“No one has created anything that’s anything like the capabilities of human intelligence,” said Lawrence, who used to be Amazon’s director of machine learning in Cambridge. “These are simple algorithmic decision-making things.” 

Lawrence said there’s no need for regulators to impose strict new rules on AI development at this stage.

People say “what if we create a conscious AI and it’s sort of a freewill” said Lawrence. “I think we’re a long way from that even being a relevant discussion.”

The question is, how far away are we? A few years? A few decades? A few centuries? No one really knows, but some governments are keen to ensure they’re ready.

In 2014, Elon Musk warned that AI could “potentially be more dangerous than nukes” and the late physicist Stephen Hawking said in the same year that AI could end mankind. In 2017, Musk again stressed AI’s dangers, saying that it could lead to a third world war and he called for AI development to be regulated.

“AI is a fundamental existential risk for human civilization, and I don’t think people fully appreciate that,” Musk said. However, many AI researchers take issue with Musk’s views on AI.

In 2017, Demis Hassabis, the polymath founder and CEO of DeepMind, agreed with AI researchers and business leaders (including Musk) at a conference that “superintelligence” will exist one day.

Superintelligence is defined by Oxford professor Nick Bostrom as “any intellect that greatly exceeds the cognitive performance of humans in virtually all domains of interest.” He and others have speculated that superintelligent machines could one day turn against humans.

A number of research institutions around the world are focusing on AI safety including the Future of Humanity Institute in Oxford and the Centre for the Study Existential Risk in Cambridge.

Bostrom, the founding director of the Future of Humanity Institute, told CNBC last year that there’s three main ways in which AI could end up causing harm if it somehow became much more powerful. They are:

1. AI could do something bad to humans.
2. Humans could do something bad to each other using AI.
3. Humans could do bad things to AI (in this scenario, AI would have some sort of moral status.)

“Each of these categories is a plausible place where things could go wrong,” said the Swedish philosopher.

Skype co-founder Jaan Tallinn sees AI as one of the most likely existential threats to humanity’s existence. He’s spending millions of dollars to try to ensure the technology is developed safely. That includes making early investments in AI labs like DeepMind (partly so that he can keep tabs on what they’re doing) and funding AI safety research at universities.

Tallinn told CNBC last November that it’s important to look at how strongly and how significantly AI development will feed back into AI development.

“If one day humans are developing AI and the next day humans are out of the loop then I think it’s very justified to be concerned about what happens,” he said.

But Joshua Feast, an MIT graduate and the founder of Boston-based AI software firm Cogito, told CNBC: “There is nothing in the (AI) technology today that implies we will ever get to AGI with it.”

Feast added that it’s not a linear path and the world isn’t progressively getting toward AGI.

He conceded that there could be a “giant leap” at some point that puts us on the path to AGI, but he doesn’t view us as being on that path today. 

Feast said policymakers would be better off focusing on AI bias, which is a major issue with many of today’s algorithms. That’s because, in some instances, they’ve learned how to do things like identify someone in a photo off the back of human datasets that have racist or sexist views built into them.

The regulation of AI is an emerging issue worldwide and policymakers have the difficult task of finding the right balance between encouraging its development and managing the associated risks.

They also need to decide whether to try to regulate “AI as a whole” or whether to try to introduce AI legislation for specific areas, such as facial recognition and self-driving cars.  

Tesla’s self-driving driving technology is perceived as being some of the most advanced in the world. But the company’s vehicles still crash into things — earlier this month, for example, a Tesla collided with a police car in the U.S.

“For it (legislation) to be practically useful, you have to talk about it in context,” said Lawrence, adding that policymakers should identify what “new thing” AI can do that wasn’t possible before and then consider whether regulation is necessary.

Politicians in Europe are arguably doing more to try to regulate AI than anyone else.

In Feb. 2020, the EU published its draft strategy paper for promoting and regulating AI, while the European Parliament put forward recommendations in October on what AI rules should address with regards to ethics, liability and intellectual property rights.

The European Parliament said “high-risk AI technologies, such as those with self-learning capacities, should be designed to allow for human oversight at any time.” It added that ensuring AI’s self-learning capacities can be “disabled” if it turns out to be dangerous is also a top priority.

Regulation efforts in the U.S. have largely focused on how to make self-driving cars safe and whether or not AI should be used in warfare. In a 2016 report, the National Science and Technology Council set a precedent to allow researchers to continue to develop new AI software with few restrictions.

The National Security Commission on AI, led by ex-Google CEO Eric Schmidt, issued a 756-page report this month saying the U.S. is not prepared to defend or compete in the AI era. The report warns that AI systems will be used in the “pursuit of power” and that “AI will not stay in the domain of superpowers or the realm of science fiction.”

The commission urged President Joe Biden to reject calls for a global ban on autonomous weapons, saying that China and Russia are unlikely to keep to any treaty they sign. “We will not be able to defend against AI-enabled threats without ubiquitous AI capabilities and new warfighting paradigms,” wrote Schmidt.

Meanwhile, there’s also global AI regulation initiatives underway.

In 2018, Canada and France announced plans for a G-7-backed international panel to study the global effects of AI on people and economies while also directing AI development. The panel would be similar to the international panel on climate change. It was renamed the Global Partnership on AI in 2019. The U.S is yet to endorse it.