“‘THIS IS SUPER COOL, BUT I CAN’T GET IT TO WORK RELIABLY ENOUGH TO ROLL OUT TO OUR CUSTOMERS.'”
The initial excitement surrounding generative AI appears to have been excessive, with several organizations discovering that the technology is unreliable and unable to fulfill business expectations. Critics argue that large language models like ChatGPT frequently produce falsehoods and engage in plagiarism, while simultaneously consuming vast quantities of energy. Moreover, businesses are recognizing the limitations of generative AI, acknowledging that it cannot be consistently trusted to deliver satisfactory results.
According to Gary Marcus, a renowned cognitive scientist, numerous enterprises have encountered issues implementing generative AI. For instance, a British company disabled its chatbot following instances where it insulted customers and disparaged its own creators. Similarly, a California car dealership faced complications when its AI-powered salesperson offered vehicles for merely $1 each. In another incident, an airline paid compensation after a deceitful chatbot informed a mourning passenger that purchasing a full-priced ticket would guarantee them a future bereavement discount.Â
Rumman Chowdhury, CEO of AI consulting firm Humane Intelligence, commented that building products using models inclined to fabricate information is unacceptable. She highlighted that generative AI models primarily generate content rather than retrieve or authenticate data. Consequently, stringent safeguards must be established to prevent erratic behavior.Â
The rapid growth of investment in the generative AI sector raises concerns among experts regarding a potential AI bubble similar to those experienced in cryptocurrency and dotcom startup markets. Some predict that lofty forecasts envisioning AI becoming a trillion-dollar industry within a decade might prove premature, invoking memories of past market exuberance followed by downturns.
Technically, critics remain unsure if the technology can progress rapidly enough to match inflated expectations. Instead, a prolonged phase of stagnation may occur. With billions of dollars flowing into the industry in a brief span, investors anticipating profitable returns might grow impatient waiting for breakthroughs. As Marcus cautiously observed, while expressing belief in the eventual feasibility of AGI, “there’s still a long way to go” specifically addressing the existing generation of AI technologies facing multiple hurdles.
