The headline on an interesting ZDNet story Monday reads like a classic case of Silicon Valley sour grapes: “ChatGPT is ‘not particularly innovative,’ and ‘nothing revolutionary,’ says Meta’s chief AI scientist.”
The story isn’t quite as provocative, but in light of Monday’s news that Microsoft will reportedly invest another $10 billion in ChatGPT and DALL-E 2 creator OpenAI, it raises worthwhile points about the value of technology, application, and user experience.
The article highlights comments made last week by Yann LeCun, Meta’s lead A.I. researcher and a Turing Award winner, during a Zoom meeting with reporters and executives. On the call, LeCun repeatedly stresses that ChatGPT, the OpenAI-developed chatbot taking the internet by storm, is built on well-worn technology that’s been in development for years.
“It’s nothing revolutionary, although that’s the way it’s perceived in the public,” LeCun said. “It’s just that, you know, it’s well put together, it’s nicely done.”
LeCun adds that he’s “not going to criticize” OpenAI for its consumer-facing engineering—though he fans the competitive flames a bit by invoking IBM’s Watson, which famously competed on Jeopardy! but never fulfilled the public’s fantasy of an all-knowing, ubiquitous application. He further contends that OpenAI will soon face competition from other companies with access to this relatively common tech.
“To be clear: I’m not criticizing OpenAI’s work nor their claims,” LeCun tweeted Tuesday. “I’m trying to correct a *perception* by the public & the media who see chatGPT as this incredibly new, innovative, & unique technological breakthrough that is far ahead of everyone else. It’s just not.”
While LeCun took some flak for seemingly disparaging his employer’s business rivals, A.I. researchers generally agreed with his fundamental ideas.
Still, LeCun’s comments put an even bigger spotlight on the importance of packaging tech products and bringing them to the masses—practices that will define success and failure of generative A.I. in the short term.
Even if the technology available to Microsoft, Google, Meta, and smaller startups is more or less the same, the results of early generative A.I. products show that consumers are acutely attuned to user experience.
ChatGPT and DALL-E 2 resonated with the public because they piqued our imagination and, despite a raft of still-troubling shortcomings, attained a certain threshold of reliability. The response allowed Microsoft to begin the process of integrating the tech into enterprise products, starting with last week’s announcement that Azure cloud customers were getting broad access to the generative A.I. tech developed by OpenAI. Media reports suggest the company also wants to bake it into products like Office and Outlook soon.
Meta’s two attempts at unveiling similar large language models, meanwhile, have been comparatively inauspicious.
The company faced a barrage of bad press last summer following the release of BlenderBot 3, a chatbot that occasionally alternated between incoherent, offensive, and conspiracy-minded. Then, in the fall, Meta’s science-focused large language model, Galactica, proved woefully biased and uninformed, prompting the company to take down the demo after three days. (Meta, in its defense, noted that the release of both products helped generate feedback that will improve outputs moving forward.)
It’s no coincidence, then, that Meta has made minimal waves in recent months about incorporating generative A.I. into Facebook, Instagram, or WhatsApp. (LeCun hinted at Meta developing technology that allows users to create art, such as video ads, though no product releases are imminent.)
Google is expected to nip at Microsoft’s heels sooner rather than later, heralding (in a layoff notice, no less) that it’s “getting ready to share some entirely new experiences for users, developers, and businesses” based on its A.I. tech. Many observers, including LeCun, have noted that Google is moving a bit slower to market than OpenAI because it faces more reputational risks in the event of a setback.
Microsoft, Meta, and other aspiring A.I. champions still face plenty of questions about their ability to convert the novelty of generative A.I. into valuable, dependable revenue generators. For now, products that are “well put together” and “nicely done,” to steal LeCun’s comments, will have an upper hand.
Want to send thoughts or suggestions to Data Sheet? Drop me a line here.
Jacob Carpenter
