this week, in At its annual I/O developer conference in Mountain View, Google showcased a dizzying number of AI-powered or enhanced projects and products. They included a new and improved version of their Bard chatbot, tools to help you write emails and documents or manipulate images, devices with built-in artificial intelligence, and an experimental chatbot-like version of Google search. For a full recap of the event, complete with insightful and witty commentary from my WIRED colleagues, check out our live Google I/O blog.
Google’s big pivot, of course, is largely driven not by algorithms, but by generative AI FOMO.
The appearance last November of ChatGPT, OpenAI’s remarkably smart but still rather flawed chatbot, combined with Microsoft adding the technology to its Bing search engine a few months later, sent Google into something of a panic. ChatGPT proved very popular with users, demonstrating new ways to deliver information that threatened Google’s control of the search business and its reputation as a leader in AI.
ChatGPT’s capabilities and the AI language algorithms like those that power it are so amazing that some experts, including Geoffrey Hinton, a pioneering researcher who recently left Google, have felt compelled to warn that we could be building systems that one day we will have. difficulties to control. The OpenAI chatbot is often amazingly good at generating coherent text on a given topic, summarizing information from the web, and even answering extremely complicated questions that require expert knowledge.
And yet unrestrained AI language models are also silver-tongued agents of chaos. They will gladly fabricate facts, express unpleasant prejudices, and say unpleasant or disturbing things with proper cues. Microsoft was forced to limit Bing’s chat capabilities shortly after launch to prevent such embarrassing misbehavior, in part because its bot divulged its secret codename, Sydney, and accused a New York Times columnist of not loving your spouse.
Google worked hard to temper the chaotic streak of text-generating technology as it prepared the experimental search feature announced yesterday that responds to search queries with chat-style responses that synthesize information from across the web.
The smartest version of Google search is impressively narrow-minded, refusing to use first person or talk about your thoughts or feelings. He completely avoids topics that could be considered risky, refusing to give medical advice or offer answers on potentially controversial topics, such as American politics.
Google deserves credit for reining in the wild side of generative chatbots like that. But in my tests, the new search interface felt incredibly tame compared to ChatGPT or Google’s own chatbot, Bard.
As the company moves the technology into more of its products, the generative AI revolution may turn out to be a lot less fun than you might expect from the initial shock and amazement of ChatGPT, a chatbot that has an edgy charm. Gone are the wild delusions and imaginations of powerful AI bots. Instead, there are new ways to fill out spreadsheets, compose email courtesies, and find products to buy.
Even if the “AI damners” warning about roving AI is overstated, it will be interesting to see how companies like Google and OpenAI balance developing more powerful generative language models with the need to make them behave.
Google has invested heavily and heavily in AI in recent years, with CEO Sundar Pichai often touting the company as “AI first,” and the company is desperate to prove it can advance the technology faster than OpenAI. . A high-level message from Google’s AI ad stream was that the company is not going to hold back any longer, as it did with the LaMDA chatbot that was announced long before ChatGPT came along, but never went public.
In March, some big names in AI research signed an open letter calling for a six-month pause on building machine learning systems more powerful than GPT-4, which powers ChatGPT. Pichai was not a signatory and said in his keynote address yesterday that the company is currently training a new, more powerful language model called Gemini.
A Google source tells me that this new system will incorporate a variety of recent advances from different large language models and may dwarf GPT-4. But don’t expect to experience all the power or charisma that Gemini can offer. If Google applies the same chaos control methods seen in its chat-like search experiment, it may seem like another surprisingly clever autocomplete.
