The big, impossible picture of AI threats and problems has drowned out the real AI invasion. In the background, an ocean of AI devices is heading into the mainstream largely unannounced.
The last time this happened was after the Second World War, when unprecedented new gizmos wiped out the pre-war technologies. It was the consumer-level things that totally obliterated that era’s lifestyle entirely in the 1950s.
It’s already here in embryonic form. The AI device deluge is coming on a much larger scale with added scope and depth. Just search “AI devices” and hit Shopping. From autonomous AI robot companions to robot cookers, magic mirrors for instant health assessments, robot dogs, actual robots, and even industrial robot heavy-lift arms, it’s an exploding dictionary of devices.
It looks so much like one of those old cartoons of the future, with automated kitsch doing everything and a few humans sprinkled around for novelty’s sake. You can wince and laugh at the same time while watching the YouTube reviews. Apparently, opening new tabs on a browser is now the equivalent of manual labour, and AI can do it for you. The absurdities aren’t too hard to find. Bear in mind that everything you’re seeing will be obsolete next year.
Current commercial applications for AI devices
A lot of these devices will look very familiar from pre-AI iterations. There are augmented reality glasses with built-in AI, translation, and more. There are also many AI wearables, from upgraded earbuds to go-anywhere Cloud-connected AI bots for all purposes, and the enigmatic AI health rings.
It also looks remarkably disorganized to the point of being utterly chaotic from a purely commercial perspective. Many of these devices are still at the barely out of Kickstarter level of development. Companies you’ve never heard of are aiming high and are creating markets for their products.
Things are very different further up the food chain. Other commercial entities are much more familiar, like major league AI agent-first programs like Microsoft Solara, supposedly the next stage in computing. This is a sort of diversified evolution of Copilot and Microsoft 365, moving the AI “from the application frame to outside the application”. This is how it runs multiple apps, connections, and workflows.
Google consumer AI devices all look pretty familiar, as if the AI is being slotted into existing hardware. The main Google page for AI products seems far more ambitious and has a much wider market reach and range.
New distribution, licensing, and emerging markets
Apparently lost in the all-pervasive blurb is a new market reality. Google and Microsoft can simply license out their AI to manufacturers. Their AI agents can act as points of sale for new products and add-ons. Would a Google agent recommend Gemini or affiliated products?
This very predictable level of “market incest’ has a few less obvious positives. AI has to work with AI and almost anything on the market to do its jobs. Cross-platform incompatibilities just aren’t an option. AI agents may also go for specifications based on prompts, adding an almost-wild card to the mix. Sanity may occur, whether anyone likes it or not.
AI devices are therefore going to have to be pretty cosmopolitan in their own operations. This could produce some interesting scenarios.
What if a top-tier AI agent finds itself dealing with a museum-like patchwork of tasks, including “whatever” types and classes of older tech?
Where are AI device productivity metrics derived and how do you get them on a balance sheet?
Do you use Software as a Service (SaaS) for standard-issue AI devices in organizations?
Do you need a dedicated neuralese network for your business?
How do you select, mix, and match AI agents in a business device context, or do you get a one-size-fits-all option?
This is already turning into a Parkinson’s Law equation, except in this case, the work is expanding the work required to do the work. In effect, the new markets are creating new markets.
AI device problems can be an unresolved future minefield
Of course, there are problems. The AI overview of AI device problems is anything but reassuring. It’s like the Internet of Things, cubed, and closely related to the Internet of Things in practice.
The expected problems are mainly AI-based. “Algorithmic hallucinations, data privacy leaks, algorithmic bias, and security vulnerabilities” are built-in, generic AI issues. With AI devices, however, these problems have an additional, perhaps lethal, dimension.
AI devices in direct contact with humans, and particularly human health issues, could be horrendously complex. Safety is the natural first hurdle for devices. Disclaimers of liability won’t work in this extremely touchy environment.
It’s an extremely tough call in the business context. A “potential lawsuit with every customer” is an unacceptable level of risk with few rewards. Whether you’re using an MRI, AI glasses, or a stovetop cooker, the AI device is the big issue for users.
Quality of technology is the key determining factor here. It’s highly debatable whether health sectors on life support can manage much ambiguity in device performance.
Add to this the fact that every facet of human interaction, from sensor calibrations to interpretation by AI or human agents, is in play. An AI device could be required as evidence.
There’s another gigantic irony here, and nobody seems to have noticed. In the Age of Deregulation, AI devices could be perfect for assessing compliance with any number of consumer laws.
Imagine this conversation:
“Isn’t there supposed to be at least some food in this product?”
“Um, ah, I think so.”
“According to this test, it’s 90% glue, 5% unidentifiable gunk, and 5% low-fat cement.”
“Oh.”
You can apply that motif to anything from property boundaries to product contract specifications to analysing your restaurant meal and finding your house keys, and it’s not that far away. The very first Star Trek tricorder was imagined as a sensor based analytical tool.
AI devices can do that. All they need is the right calibrations.
Just make it a point to watch the evolution of AI devices to see what happens next.