Elon Musk is right: AI needs to slow down for the good of society

• Elon Musk, Steve Wozniak, and over 1,000 others signed a letter calling for a pause on new AI models.
• The rapid rise of ChatGPT and tech likes it shows how it has incredible potential for good and ill.

There's precious little Apple cofounder Steve Wozniak and Elon Musk agree on these days, with one notable exception: AI needs to slow down.

Wozniak, Musk, and more than 1,000 other business leaders signed a letter seeking guardrails and a pause on training AI models as the technology grows more powerful. The letter argues powerful AI models like OpenAI's GPT-4 "should only be developed only once we are confident that their effects will be positive and their risks will be manageable."

Of course, this is not the first time people have called for guardrails for AI. As AI has gotten more sophisticated, however, the calls for caution are growing in turn.

"It's a very good idea to slow down development of new models because if AI ends up being good for us, then there's no harm waiting months or years, we'll get to the end point anyway," said James Grimmelmann, professor of digital and information law at Cornell University. "If it's harmful, then we just bought ourselves extra time to strategize the best ways to respond and understand how to combat it."

Which is to say, Wozniak and the other tech leaders are right. AI has been thrown out into the wild, left to grow unattended and growing stronger every day. But now, we have the chance to slow it down, giving the potential to better cultivate the technology for society's sake.

The rise of ChatGPT shows the potential dangers of moving too fast

When OpenAI's blockbuster ChatGPT came out for public testing in November last year, it caused a major stir. Understandably, people began pushing ChatGPT's capabilities. It soon became apparent just how disruptive it could be to society It started passing medical licensing exams. It gave directions on how to make bombs (after a little extra prompting). It created an alter ego for itself.

The more we use AI — particularly so-called generative AI tools like ChatGPT or the text-to-image tool Stable Diffusion — the more we see its faults, its potential for creating bias, and how unprepared we humans are to wield its power.

Insider's Hasan Chowdhury wrote that AI has the potential to "be a turbocharger that accelerates the spread of our faults." Like social media, it can harness the best and worst of humanity. But unlike social media, AI will be far more integrated into people's lives.

ChatGPT and other similar AI products already tend to misrepresent information and make mistakes, something Wozniak has spoken publicly about. It is prone to hallucinations, and even OpenAI CEO Sam Altman admitted that his company's model could spit out racist, sexist, and biased answers. Stable Diffusion bumped into copyright issues and has been accused of stealing art from digital artists.

As AI gets folded into more everyday technology, we risk bringing more misinformation into the world, at larger scales. Even seemingly benign tasks for AI like helping plan a vacation may not produce trustworthy results.

It's hard to develop responsibly when the free market demands moving quickly

To be clear, AI, particularly generative AI like ChatGPT, is incredibly transformative technology. There is nothing inherently wrong with developing machines that do much of the tedious work people loathe.

As much as the technology has created an existential crisis in the workforce, it's also been celebrated as an equalizing tool for the tech industry. There is also no evidence that ChatGPT is ready to lead a robot rebellion in the next few years.

And many AI companies have ethicists concerned with developing the technology responsibly. But a dedicated team focused on safely creating AI cannot do its job if the rush to put out products is placed higher than its societal impact.

Speed does seem to be the name of the game, too. OpenAI figured out that if it moves quickly enough, it can ward off competition and become the leader in the generative AI space. That's pushed Microsoft, Google, and just about everybody else to follow suit.

Releasing powerful AI models for the public to play with before it's ready isn't making the technology better. People cannot find the best use cases for AI because the developers have to put out fires their technology created, and users are distracted by the noise.

Not everyone wants an AI pause

The letter does have its detractors, who say it misses the mark.

University of Washington professor Emily M. Bender said in a Twitter thread Musk and the other tech leaders give into the hype cycle by focusing on how powerful AI can be and not the actual damage it can do.

Grimmelman adds the tech leaders signing the letter "are late to the party" and opened up a Pandora's box of potential complications for themselves.

"Now that they've signed the letter, then they can't turn around and say the same policies cannot be made for other technologies like self-driving cars," he said.

A moratorium or more regulation may or may not come to fruition. But the conversation has now turned.

AI has been around for decades; it can wait a few more years.