AI May Be Dangerous, but Here's Why Experts Say You Shouldn't Freak Out Too Much Yet

It's hard to know who to believe when it comes to the debate over the dangers of artificial intelligence (AI). 

A growing number of experts have expressed their alarm over the increasing capabilities of AI. Meanwhile, a new paper fuels the notion that some concerns about AI abilities may be overblown. 

"Poorly designed and improperly deployed AI systems have already caused harm to many human beings, so fears about the technology, or maybe anger at some of the ways in which it's being used, is absolutely warranted," Irina Raicu, the director of the internet ethics program at Markkula Center for Applied Ethics at Santa Clara University told Lifewire in an email interview. "But there are also fears about AI-related outcomes that might be far into the future or might never come to pass."

The AI Doom Squad

Alarm about AI is becoming eschatological. Public figures have said they are worried super-smart AI could bring the end of the human race. Geoffrey Hinton, sometimes called the godfather of Ai, recently announced his resignation from Google, saying he now regretted his work and is frightened by the abilities of AI. 

In March, the Future of Life Institute released an open letter calling on all AI labs to immediately pause for at least six months the training of AI systems more potent than GPT-4.

“AI systems with human-competitive intelligence can pose profound risks to society and humanity, as shown by extensive research and acknowledged by top AI labs,” according to the letter. “...Advanced AI could represent a profound change in the history of life on Earth and should be planned for and managed with commensurate care and resources.”

AI for Good?

But many AI experts are expressing doubts about some of the most pessimistic predictions for the new technology. One of the most startling recent claims about AI is that large language models like the one that powers GPT-4 can learn new abilities independently. For example, Microsoft researchers said that the GPT-4 language model showed sparks of artificial general intelligence. The idea conjures up images of AI running amok. 

The fears of self-teaching AI, at least, may be overblown. A recent paper by Stanford University scientists shows that instances when a large model suddenly displays an ability it was not designed to possess are actually a "mirage" produced by researchers.

AI is dangerous insofar as it has the potential to be put to harmful uses, John Yawney, the Chief Analytics Officer at the firm Adastra told Lifewire in an email.

"As with any technological advancement, there's the potential for bad actors to use AI to nefarious ends by configuring an AI algorithm to solve malicious objectives such as creating fake video, audio, or generating other fraudulent transactions," he added. "There are also concerns about the potential for military applications, and these cannot be wholly dismissed, but there is always the counterbalance driven by the fears of retaliation and diplomatic controls that need to be enforced."

Even some of the more mundane threats, like AI taking away jobs, may not come to pass, Srinath Sridhar, the CEO of Regie.ai, said in an email interview. "People often think AI or automation will replace humans, but I believe this is far from true and probably stems from fear of change," he added. "If you look back at the history of automation in sales, every single time it has actually increased the number of jobs."

AI may have as much to fear from humans as vice versa. Jo Ann Oravec, a professor of Information Technology and Supply Chain Management at the University of Wisconsin-Whitewater, predicts that fearful people could lash out violently against AI-driven robots and autonomous vehicles. 

"Science fiction stories have often emphasized sensationalistic aspects of human-AI interaction, with their conclusions often resulting in death and destruction," Oravec said in an email. "However, more nuanced discourses on how our lives could be enhanced by AI are needed. We could certainly eliminate many boring aspects of our workplace and educational lives with the help of AI. Humans will indeed need to learn how to work alongside AI systems as collaborators, which will require new kinds of patience and understanding than in human-human collaborations."