Chatbot expert names the real danger of artificial intelligence

Focusing on AI doomsday scenarios is a distraction, downplaying immediate risks such as the large-scale spread of disinformation, according to a senior industry official at the AI ​​Security Summit.

According to The Observer, Aidan Gomez, co-author of the research paper that helped create the technology behind chatbots, said long-term risks such as existential threats to humanity from AI should be “studied and pursued”, but that they could distract policymakers from addressing immediate potential harm.

“I think from an existential risk and public policy perspective, this is not a productive conversation,” Gomez argues. “In terms of public policy and what we should be focusing on in the public sector—or trying to reduce the risk to civilians—I think it’s a distraction from the risks that are much more tangible and immediate.”

Gomez is attending the two-day summit, which starts Wednesday, as chief executive of Cohere, a North American company that makes artificial intelligence tools for business, including chatbots. In 2017, at age 20, Gomez worked on a team of Google researchers that created Transformer, the key technology behind large language models that power artificial intelligence tools such as chatbots.

Gomez said artificial intelligence, a term for computer systems that can perform tasks typically associated with intelligent beings, is already in widespread use and those applications should be the focus of the summit. Chatbots like ChatGPT and image generators like Midjourney have stunned the public with their ability to create believable text and images from simple text prompts.

“This technology is already used in a billion consumer products, such as Google and others. This creates a lot of new risks to discuss, none of which are existential, none of which are doomsday scenarios, Gomez said. “We should focus directly on those works that are about to influence people or actively influence them, as opposed to perhaps more academic and theoretical discussions about the long-term future.”

Gomez said disinformation – the spread of misleading or incorrect information online – has become his main concern. “Disinformation is what matters most to me,” he said. “These artificial intelligence models can create media that is extremely persuasive, highly compelling, virtually indistinguishable from human-generated text, images or media. And this is something we need to address urgently. We need to figure out how we are going to enable the public to differentiate between these different types of media.”

The opening day of the summit will feature discussions on a range of artificial intelligence issues, including issues related to disinformation such as election disruption and the erosion of public trust. The second day, which will bring together a small group of countries, experts and tech leaders convened by Rishi Sunak, will discuss what concrete steps can be taken to address the risks posed by artificial intelligence. US Vice President Kamala Harris is expected to be among those present.

Gomez, who called the summit “really important,” says it is already “very likely” that an army of bots—software that performs repetitive tasks such as posting on social media—could spread misinformation generated by artificial intelligence. “If you can do that, it’s a real threat to democracy and public debate,” he warns.

In a series of documents outlining the risks of AI last week, which included misinformation generated by AI and disruptions to the labor market, the British government said it could not rule out the development of AI reaching a point where the systems would threaten humanity, The Observer noted.

A risk paper published recently said: “Given the significant uncertainty in predicting the development of AI, there is insufficient evidence to rule out that high-performance edge AI systems, if misconfigured or inadequately controlled, could pose an existential threat.”

The document added that many experts consider such a risk to be very low and that it would entail a number of scenarios, including the system gaining advanced control over weapons or financial markets. Concerns about the existential threat from AI center on the prospect of so-called artificial general intelligence—a term for an AI system capable of performing multiple tasks at human or superhuman levels of intelligence—that could theoretically replicate itself, evade human control, and make decisions that go against human control. with human ones.’ interests.

Those concerns led to the publication in March of an open letter signed by more than 30,000 technologists and experts, including Elon Musk, calling for a six-month pause in the giant artificial intelligence experiments.

Two of the three modern AI godfathers, Geoffrey Hinton and Yoshua Bengio, signed another statement in May warning that preventing the risk of extinction from AI should be taken as seriously as the threat of pandemics and nuclear war. However, Yann LeCun, their fellow “godfather” and one of the winners of the ACM Turing award, which is considered the Nobel Prize in computing, called fears that artificial intelligence could destroy humanity “absurd.”