The AI Hype Cycle

This morning’s tech headline was “Texas man uses Apple AirTag to track and kill truck thief,” an interesting take on frontier justice with a technological twist. An “iPosse” if you will.

But the real headlines in tech these days tout GPT-4, the Large Language Model (LLM) developed by OpenAi which underpins ChatGPT, everybody’s favorite chatbot. New York Times columnist Brian X. Chen did a column comparing ChatGPT and Bard (Google’s competitive chatbot) on their performance of tasks typically assigned to executive assistants: meeting preparation, summarizing meetings, booking travel and calendar management. His conclusion? “Google’s Bard chatbot fared far worse than OpenAI’s ChatGPT, but human assistants might soon be out of their jobs.” The primary issue is these programs lack interfaces to websites and applications that allow them to
actually book tickets and hotels, or alter calendars. Once that happens, it’s only a matter of time before they start firing nukes at Russia.

None other than Bill Gates believes that these LLM-based programs will revolutionize the world: “The development of AI is as fundamental as the creation of the microprocessor, the personal computer, the Internet and the mobile phone. It will change the way people work, learn, travel, get health care and communicate with each other. Entire industries will reorient around it. Businesses will distinguish themselves by how well they use it.” Gates is a smart guy—you can read his full seven-page take on the revolution at gatesnotes.com (“The Age of AI has begun”).

And so we are full into the hype and exploitation cycle that surrounds new technologies. You may remember when the blockchain was going to revolutionize the world. Every company touted its relationship with blockchain technology in order to gain publicity, pique the interest of investors and raise the price of their stock.

In 2017, Long Island Iced Tea Corporation, a maker of iced tea and other juice beverages, announced that it was “shifting its primary corporate focus toward the exploration of and investment in opportunities that leverage the benefits of blockchain technology” (while continuing to make beverages). The company saw its stock price rise 200% following the announcement, which included a corporate name change to “Long Blockchain Corporation.” In 2021, the SEC delisted Long Blockchain Corporation shares, and filed insider trading charges against several large investors in Long Blockchain investors. It’s unclear whether they are still in business: I dialed the number on their website and got a “no longer in service” message. The NY Times noted earlier this year that, “Now, the opposite is happening—companies squarely in the crypto sphere are shedding the ‘b’ word. In early January, a Bitcoin mining company, Riot Blockchain, became Riot Platforms. In October, Applied Blockchain became Applied Digital.”

Disregarding that lesson, everybody seems to be eagerly adding “AI” to whatever it is their company does, seeking to hitch their wagons to the appeal of a technology that appears to act with human intelligence.

Speaking of human intelligence, much was made recently of GPT-4’s ability to take tests. It scored in the 88th percentile on the LSATs (Law School Admission Test) and surpassed that mark with a 90th percentile on the Uniform Bar Exam. But really, a lot of what passes for human “intelligence” is the ability to rearrange and reduce textual information into established patterns. The AI trick of writing X in the style of Y becomes simpler if you’ve seen, broken apart, and remembered thousands of examples of both X and Y in your training data. To quote Reid Hoffman, who co-founded LinkedIn and has been a successful angel investor: “I’d argue that repackaging available information actually describes an enormous share of human innovation, artistic or otherwise.” Like me, Hoffman is skeptical that these generative AIs display human intelligence simply because they can mimic human writing. But he does believe they are useful tools. Hoffman has written “Impromptu” (available as a free PDF at impromptubook.com), which he describes as a “travelog of the future.” Because of the rapid change in this technology, Hoffman was wary of writing a book that would almost immediately be obsolete. Instead, this is more of a snapshot of the state of the art, describing things which have yet to percolate into most people’s futures.

This early paragraph from the book sums up Hoffman’s thesis in “Impromptu”: “When human users treat GPT-4 as a co-pilot or a collaborative partner, it becomes far more powerful. You compound GPT-4’s computational generativity, efficiency, synthetic powers and capacity to scale with human creativity, human judgment and human guidance.” He then proceeds to provide specific examples in the areas of education, creative work, justice, journalism, social media and the workplace.

“Impromptu” is quite readable, and GPT-4 was both his collaborator and an example in the book. There’s no deep “how does this work” to understand. Instead, I recommend you travel along with Mr. Hoffman to see how generative AI can be used today. To repeat Bill Gates: “Businesses will distinguish themselves by how well they use it.” That holds for people, as well.