The Flaw in Altman’s Thesis for AI Devices
One of the problems with this guessing game is that it may well prove irrelevant, because the arguments about our need for AI-specific devices don’t hold water. OpenAI CEO Sam Altman, talking with Ive in a slick video explaining the io deal, seemed to deliberately obfuscate how people currently use AI chatbots. He described how asking ChatGPT currently requires someone to get out their laptop, launch a browser and start typing. Come on! At the most basic level, you can already talk to ChatGPT and ask it questions. Even more convenient is the option to talk to a locked phone—such as Google’s Pixel—and ask the Gemini chatbot a question (which it answers!). You can do the same with Meta Platforms’ Ray-Bans smart glasses—no need for laptop, browser or typing. Whatever OpenAI is planning, can it really be simpler than using a device we already carry with us? (Bloomberg reported Thursday that Apple is also planning to get into the smart glasses market).
Now, Altman obviously knows all this. The real reason he is talking up the value of new devices is surely strategic: OpenAI, like Meta, doesn’t want to be dependent on smartphone makers—such as Apple—for distribution of its apps. The company has tried unsuccessfully to strike deals to preload its ChatGPT app onto phones, an executive testified at a Google antitrust trial recently, in an attempt to replicate Google’s strategy of expanding availability of its search app by getting it preloaded onto smartphones and browsers. Altman likely thinks OpenAI needs to control its destiny by developing its own hardware.
That’s fine, but consumers may have other ideas. Aside from the difficulty of persuading people they need another device, there’s the time it will take OpenAI to get to market with its new device. According to The Wall Street Journal today, Altman told OpenAI staff on Wednesday evening that he hopes to ship the first new AI device by late next year. Eighteen months is a long time in the fast-evolving AI world. By the time OpenAI’s devices finally make it to market, people may be talking to AI assistants on their phones or smart glasses as freely as they upload videos to TikTok or Instagram.
Datadog Invents New Language
If there’s a guaranteed way to produce impenetrable jargon, it’s by combining AI and open-source technology. Check out this blog post from Datadog on Wednesday about—according to our grizzled enterprise software reporter Kevin McLaughlin—its new open-source AI model and an accompanying new benchmark to measure how accurate the AI model is at forecasting server and app performance issues.
The blog post, however, takes our monthly award for incomprehensible jargon. Even Kevin couldn’t translate this sentence: “Observability time series have challenging distributional properties, including sparsity, spikes and noisiness; anomaly contamination; and contain high-cardinality multivariate series, with complex relationship among variates.”
Translated by ChatGPT, that means: “This kind of data is messy. It has missing parts, random ups and downs, weird spikes and lots of things happening all at once that are connected in confusing ways.”