It’s 2025, and companies still don’t know if AI is good. That’s the impression I got from this year’s CES, which featured AI-powered kitchen appliances, baby cribs, and other products that really didn’t ask for AI. Behold: Spicerr, the touchscreen-equipped “smart” spice dispenser that learns your tastes while cooking to recommend unique recipes. Spicerr’s utility is somewhat questionable for starters. Spicerr does not grind, and takes $15-$20 sealed capsules that are not refillable. Setting all that aside, are people really itching for a meal-suggesting salt and pepper shaker to start with? Elsewhere at the show, there was Dreo’s ChefMaker 2, an AI-powered air fryer. Yes, you read that right – an AI-powered air fryer. This concept is not as strange as Spicerr. ChefMaker 2 can extract recipes from cookbooks through its page scanning feature, and even handle complex math to calculate cooking times and times. But is cookbook scanning really a feature the public is demanding? Speaking as a member myself, I can’t say I’ve ever been – and that seems to be true for most people. Image Credit: Dreo Well, CES has more weird AI products in store. Razer’s Project Ava, named after the killer robot in the 2014 film “Ex Machina,” is an “AI game copilot,” as the company describes it. Ava basically plays the game for you without actually playing the game for you. With permission, Ava takes a picture of the computer screen, then gives pointers (eg, “Dodge while the blade is spinning”). As The Verge’s Sean Hollister writes, Ava is controversial because it is clearly trained on the game’s instructions, but does not give credit to the author. It’s also distracting. At least in its current form, Ava is on a delay of a few seconds, and interrupts the game’s audio to give instructions. I have to ask again: Who asked for this, exactly? Who will use it regularly, much less pay? As far as I can tell, the AI products coming out at CES are a symptom of the industry’s out-of-control hype. AI companies raised $97 billion last year in the US, enough to buy 42 Spheres. Vendors are throwing AI spaghetti at the wall to see what sticks, because there are no downsides – and huge potential. In many cases, they also push against the limits of AI as we know it. Finding technically feasible AI use cases has become a formidable challenge for the industry. Often, it leads to over-promising – under-delivering. ChatGPT is still wrong. Historically inaccurate image generators. And the characters in the AI video merge into each other’s bodies. So we’re stuck with IRL AI slop: an air fryer, a spice dispenser, and an “AI game copilot.” They’re not what you’d expect, but they can be today with a relatively cheap R&D lift. Here’s to a better year ahead. TechCrunch has an AI-focused newsletter! Sign up here to get it in your inbox every Wednesday.
