By Will Knight | 08.17.23 |
Greetings, non–artificial intelligences (and any artificial ones who have deigned to subscribe), This week we look back at an extraordinary moment in recent tech history—that occurred only five months ago. In March, thousands of artificial intelligence experts signed an open letter calling for a halt on the development of more advanced algorithms. The letter's warning that "AI systems with human-competitive intelligence can pose profound risks to society and humanity," and the backdrop of excitement and confusion about the capabilities of ChatGPT created the impression that many experts had become worried about machines capable of deceiving, displacing, or even eventually annihilating us. An obscure strand of AI research, focusing on the theoretical, long-term dangers of AI, suddenly went mainstream. But what did those who signed that letter actually believe? |
Rewriting the AI Doomsday Letter 💌 ☠️ |
This March, nearly 35,000 AI researchers, technologists, entrepreneurs, and concerned citizens signed an open letter from the nonprofit Future of Life Institute that called for a "pause" on AI development, due to the risks to humanity revealed in the capabilities of programs such as ChatGPT. "Contemporary AI systems are now becoming human-competitive at general tasks, and we must ask ourselves ... Should we develop nonhuman minds that might eventually outnumber, outsmart, obsolete and replace us?" I could still be proven wrong, but almost six months later and with AI development faster than ever, civilization hasn't crumbled. Heck, Bing Chat, Microsoft's "revolutionary," ChatGPT-infused search oracle, hasn't even displaced Google as the leader in search. So what should we make of the letter and similar sci-fi warnings backed by worthy names about the risks posed by AI?
|
Two enterprising students at MIT, Isabella Struckman and Sofie Kupiec, reached out to the first hundred signatories of the letter calling for a pause on AI development to learn more about their motivations and concerns. The duo's write-up of their findings reveals a broad array of perspectives among those who put their name to the document. Despite the letter's public reception, relatively few were actually worried about AI posing a looming threat to humanity itself. Many of the people Struckman and Kupiec spoke to did not believe a six-month pause would happen or would have much effect. Most did not envision the "apocalyptic scenario" that one anonymous respondent acknowledged some parts of the letter evoked. A significant number of those who signed were, it seems, primarily concerned with the pace of competition between Google, OpenAI, Microsoft, and others, as hype around the potential of AI tools like ChatGPT reached giddy heights. Google was the original developer of several algorithms key to the chatbot's creation, but it moved relatively slowly until ChatGPT-mania took hold. To these people, the prospect of companies rushing to release experimental algorithms without exploring the risks was a cause for concern—not because they might wipe out humanity but because they might spread disinformation, produce harmful or biased advice, or increase the influence and wealth of already very powerful tech companies. |
Some signatories also worried about the more distant possibility of AI displacing workers at hitherto unseen speed. And a number also felt that the statement would help draw the public's attention to significant and surprising leaps in the performance of AI models, perhaps pushing regulators into taking some sort of action to address the near-term risks posed by advances in AI. Back in May, I spoke to a few of those who signed the letter, and it was clear that they did not all agree entirely with everything it said. They signed out of a feeling that the momentum building behind the letter would draw attention to the various risks that worried them, and was therefore worth backing. But perhaps it was a mistake to try to cover so many issues potentially raised by existing and recently developed AI in a letter that would inevitably be defined by its most outlandish and scary claim. Some AI researchers have spent the past few years warning presciently about the more immediate societal problems that large language models could cause, including exacerbating ingrained biases. Their concerns were barely audible amid the furor the letter prompted around doomsday scenarios about AI. The prominence of that apocalyptic strand of thinking was reinforced by a follow-up statement in May, also signed by many high-profile AI researchers, that compared the extinction threat of AI to that of nuclear weapons and pandemics. |
Nirit Weiss-Blatt, author of The Techlash and Tech Crisis Communication, who reviewed the MIT paper before its publication, says the letter and statement ended up serving the interests of the tech firms building cutting-edge AI, because the focus on far-off worst-case scenarios makes regulators believe the technology is both incredibly valuable and hard to handle. "All those professors, they're not thinking about existential risk at all," Weiss-Blatt says. "But they lent their name to the extreme AI doomers. That's the real misinformation here."
In the end, the letter asking for a pause on AI development may have done the opposite of what many of those who signed wanted. By making discussion of doomsday scenarios more prominent, the letter made it harder for concerns about less-than-superintelligent machines to win notice or inspire action. |
|
| Ex-Google CEO Eric Schmidt is reportedly backing a company modeled on OpenAI that will tackle scientific challenges with the help of AI. (Semafor) With some political campaigns already dabbling in deepfakery, the FEC signals that it could limit use of AI in political ads. (The Verge) Dave Limp, Amazon's devices chief, is retiring from the company after 13 years, the latest high-profile departure under CEO Andy Jassy. (The Wall Street Journal) Google-DeepMind, the division of Alphabet tasked with competing against OpenAI, is said to be testing generative AI tools capable of offering a range of advice and assistance on everyday tasks. (The New York Times) |
That's all for this week. I hope the newsletter helped alleviate a little existential angst, or at least drew your attention to more pressing problems with AI. Feel free to send me a message on X (formerly Twitter) or Mastodon, or just reply to this email with feedback. Thanks for reading, and see you next week! |
|
|
|
No comments:
Post a Comment