Welcome to Wet Hot AI Chatbot Summer!
At the end of last year, I attended event hosted by Google to celebrate advances in artificial intelligence. The company’s holdings in the Chelsea area of New York now literally stretch to the Hudson River, and about a hundred of us gathered in the exhibition space at the pier to watch prepared presentations from executives and showcase the latest achievements. Speaking remotely from the West Coast, the company’s computing high priest, Jeff Dean, promised “an encouraging vision for the future.”
The theme of the day was “exploring the (im)possible”. We learned how Google’s AI has been used to fight wildfires, predict floods, and evaluate retinal diseases. But the stars of the show were what Google called “generative AI models.” These are content machines trained on massive training datasets to stamp out lettering, images, and even computer code that only humans could once create.
Something strange is happening in the world of AI. At the beginning of this century, the field snapped out of a state of lethargy known as AI winter thanks to “deep learning” innovations led by three scientists. This approach to AI has changed this area and made many of our apps more useful, providing language translations, search, Uber routing, and just about anything with the word “smart” in its name. We’ve spent a dozen years in this AI spring. But in the last year or so, there has been a sharp reprise of that earthquake, with a sudden abundance of mind-boggling generative patterns.
Most of the toys Google showed off at the New York Pier showed off the fruits of generative models, like the flagship big language model called LaMDA. He can answer questions and working with creative writers make stories. Other projects may produce 3D images from text tips or even help produce video by creating storyboard-like sentences from scene to scene. But most of the program was devoted to some of the ethical issues and potential dangers of launching content generator robots into the world. The company has taken pains to emphasize how careful it is when using its powerful creations. The most eloquent statement was made by Douglas Eck, chief scientist at Google Research. “Generative AI models are effective—there is no doubt about that,” he said. “But we also need to recognize the real risks this technology can pose if we don’t take care of it, which is why we’re in no hurry to release it. And I’m proud that we weren’t in a rush to release them.”
But Google’s competitors don’t seem to have the word “slow” in their dictionaries. While Google has provided limited access to LaMDA in the secure Test Kitchen app, other companies offer a smorgasbord of their own chatbots and image generators. Just a few weeks after the Google event, the most important release came out: the latest version of OpenAI’s powerful text generation technology, ChatGPT, a lightning-fast, logorheic gadfly that spits coherent essays, poems, plays, songs, and even obituaries at the slightest hint of a clue. Taking advantage of the chatbot’s wide availability, millions of people have experimented with it and shared its amazing reviews, to the point where it has become an international obsession as well as a source of wonder and awe. Will ChatGPT kill college essay? Destroy traditional web search? Place millions of copywriters, journalists, artists, songwriters and legal assistants out of employment?
The answers to these questions are not yet clear. But there is one. Making these models publicly ushered in a rainy, hot AI summer that is energizing the tech sector, even as the current giants are laying off some of their workforce. Contrary to Mark Zuckerberg’s belief, the next big paradigm is not the metaverse, but a new wave of AI content engines, and it’s already here. In the 1980s, we saw a gold rush of products that moved tasks from paper to PC applications. In the 1990s, you could get rich quick by bringing these desktop products online. Ten years later, the movement was towards mobile devices. In the 2020s, there will be a big shift towards the creation of generative AI. Thousands of startups will appear this year with business plans based on using the APIs of these systems. The cost of issuing universal copies will be reduced to zero. By the end of the decade, AI video creation systems may well dominate TikTok and other apps. They may not be as good as the innovative creations of talented people, but quantitatively, robots will dominate.