This column was written by Robert Skidelsky for Project Syndicate.
LONDON – In Mary Shelley’s novel Frankenstein, or The Modern Prometheus, scientist Victor Frankenstein famously uses dead body parts to create a hyperintelligent “superhuman” monster that – driven mad by human cruelty and isolation – ultimately turns on its creator. Since its publication in 1818, Shelley’s story of scientific research gone wrong has come to be seen as a metaphor for the danger (and folly) of trying to endow machines with human-like intelligence.
Shelley’s tale has taken on new resonance with the rapid emergence of generative artificial intelligence. On March 22, the Future of Life Institute issued an open letter signed by hundreds of tech leaders, including Tesla CEO Elon Musk and Apple co-founder Steve Wozniak, calling for a six-month pause (or a government-imposed moratorium) in developing AI systems more powerful than OpenAI’s newly released CHatGPT-4. “AI systems with human-competitive intelligence can pose profound risks to society and humanity,” says the letter, which currently has more than 25,000 signatories. The authors go on to warn of the “out-of-control” race “to develop and deploy ever more powerful digital minds that no one – not even their creators – can understand, predict, or reliably control.”
Musk, currently the world’s second-richest person, is in many respects the Victor Frankenstein of our time. The famously boastful South Africa-born billionaire has already tried to automate the entire process of driving (albeit with mixed results), claimed to invent a new mode of transportation with the Boring Company’s (still hypothetical) hyperloop project, and declared his intention to “preserve the light of consciousness” by using his rocket company SpaceX to establish a colony on Mars. Musk also happens to be a co-founder of OpenAI (he resigned from the company’s board in 2018 following a failed takeover attempt).
One of Musk’s pet projects is to combine AI and human consciousness. In August 2020, Musk showcased a pig with a computer chip implanted in its brain to demonstrate the so-called “brain-machine interface” developed by his tech startup Neuralink. When Gertrude the pig ate or sniffed straw, a graph tracked its neural activity. This technology, Musk said, could be used to treat memory loss, anxiety, addiction, and even blindness. Months later, Neuralink released a video of a monkey playing a video game with its mind thanks to an implanted device.
These stunts were accompanied by Musk’s usual braggadocio. Neuralink’s brain augmentation technology, he hoped, could usher in an era of “superhuman cognition” in which computer chips that optimize mental functions would be widely (and cheaply) available. The procedure to implant them, he has claimed, would be fully automated and minimally invasive. Every few years, as the technology improves, the chips could be taken out and replaced with a new model. This is all hypothetical, however; Neuralink is still struggling to keep its test monkeys alive.
While Musk tries to create cyborgs, humans could soon find themselves replaced by machines. In his 2005 book The Singularity is Near, futurist Ray Kurzweil predicted that technological singularity – the point at which AI exceeds human intelligence – will occur by 2045. From then on, technological progress would be overtaken by “conscious robots” and increase exponentially, ushering in a better, post-human future. Following the singularity, according to Kurzweil, artificial intelligence in the form of self-replicating nanorobots could spread across the universe until it becomes “saturated” with intelligent (albeit synthetic) life. Echoing Immanuel Kant, Kurzweil referred to this process as the universe “waking up.”
But now that the singularity is almost upon us, Musk and company appear to be having second thoughts. The release of ChatGPT last year has seemingly caused panic among these former AI evangelists, causing them to shift from extolling the benefits of super-intelligent machines to figuring out how to stop them from going rogue.
Unlike Google’s search engine, which presents users with a list of links, ChatGPT can answer questions fluently and coherently. Recently, a philosopher friend of mine asked ChatGPT, “Is there a distinctively female style in moral philosophy?” and sent the answers to colleagues. One found it “uncannily human.” To be sure, she wrote, “it is a pretty trite essay, but at least it is clear, grammatical, and addresses the question, which makes it better than many of our students’ essays.”
In other words, ChatGPT passes the Turing test, exhibiting intelligent behavior that is indistinguishable from that of a human being. Already, the technology is turning out to be a nightmare for academic instructors, and its rapid evolution suggests that its widespread adoption could have disastrous consequences.
So, what is to be done? A recent policy brief by the Future of Life Institute (which is partly funded by Musk) suggests several possible ways to manage AI risks. Its proposals include mandating third-party auditing and certification, regulating access to computational power, creating “capable” regulatory agencies at the national level, establishing liability for harms caused by AI, increasing funding for safety research, and developing standards for identifying and managing AI-generated content.
But at a time of escalating geopolitical conflict and ideological polarization, preventing new AI technologies from being weaponized, much less reaching an agreement on global standards, seems highly unlikely. Moreover, while the proposed moratorium is ostensibly meant to give industry leaders, researchers, and policymakers time to comprehend the existential risks associated with this technology and to develop proper safety protocols, there is little reason to believe that today’s tech leaders can grasp the ethical implications of their creations.
In any case, it is unclear what a pause would mean in practice. Musk, for example, is reportedly already working on an AI startup that would compete with OpenAI. Are our contemporary Victor Frankensteins sincere about pausing generative AI, or are they merely jockeying for position?
Register here to attend CEU's Presidential Lecture Series "Humans and Machines" featuring Robert Skidelsky.
Watch the first lecture "The Future of Work" from April 19 here: