John Schulman, a notable name in the realm of artificial intelligence (AI), has stirred up quite the controversy with his recent remarks. During a podcast interview, this co-creator of OpenAI— an AI research firm leading the charge in cutting-edge AI development— hinted at an upcoming breakthrough that many deemed implausible within such a short span. The breakthrough he’s alluding to is Artificial General Intelligence (AGI): an AI that possesses intelligence on par with humans and all their capabilities. Schulman seems confident that we could achieve AGI in a mere “two or three years.”
Such bold words have rippled shockwaves through the technological fraternity; AGI has always been an entity shrouded in both awe-inspiring wonder and terrifying dread. Advocates paint pictures of a world where AI can surmount any challenge thrown its way— thereby reshaping industries with revolutionary force. But even while marveling at these possibilities, experts voice valid apprehensions regarding AGI: it could outstrip human oversight or make redundant manpower dependent on humans.
Schulman acknowledges these concerns and emphasizes the need for caution. He argues that tech companies must work together to ensure the safe development of AGI. He warns against a “race dynamic” where companies prioritize staying ahead of the curve at the expense of safety. “Everyone needs to agree on some reasonable limits to deployment or to further training,” Schulman stressed, advocating for collaboration among major players in the field.
OpenAI itself is a prime example of the race for AGI. The company has made significant strides in developing powerful AI models, with GPT-4 being a recent example. However, the release of GPT-4 also sparked concerns. Last year, prominent figures like Elon Musk signed a letter calling for a six-month pause on training AI systems exceeding GPT-4’s capabilities, citing potential safety risks.
Schulman’s comments echo these concerns. He suggests that if AGI arrives sooner than expected, companies should be prepared to hit the brakes. He proposes a pause on training and deployment until safety measures are in place. This aligns with the growing movement advocating for a pause on AI development, spearheaded by groups like Pause AI. These groups fear that superintelligent AI could pose existential threats to humanity.
OpenAI recently underwent internal changes that further highlight the tension between progress and safety. Schulman himself took over the company’s safety research efforts following the departure of key figures from the Superalignment team, a group specifically focused on ensuring safe AGI development. These departures were reportedly fueled by concerns that OpenAI prioritized “shiny products” over safety.
The coming years will be crucial in determining the path of AI development. Schulman’s prediction serves as a stark reminder of the potential both exciting and terrifying that lies ahead. Whether we enter a golden age of intelligent machines or face unforeseen dangers depends on our ability to prioritize safety and navigate this new frontier with caution and cooperation.
