Google (A)I/O

Google, a small company you may never have heard of, spent most of this year’s I/O conference shotgunning AI products all over everything they make. There’s a bunch of products and services with varying names with overlapping and nested functions. I’d explain them to you, but I couldn’t be bothered to actually care enough about the details to learn what the differences are.

My takeaway was this: no one knows how to implement or market AI in a way that non-enterprise customers can understand, let alone get excited for. This isn’t limited to Google, it’s Amazon, Apple, Microsoft and the rest.

AI is useful. It is better than it was a year ago. At some point, vaguely around a year ago, it probably became good enough for consumers to pay attention to. But the purveyors of fine AI don’t know how to make that happen. AI is the end-all example of “this is a feature, not a product,” and boy are companies missing the point.

Google, for example, has decided to let AI interrogate and respond to every search query as though that’s what the users want. It shouldn’t. Typing a string of words into a search box should result in a list of web results. Typing a question should result in an answer. The list of web results should use an AI algorithm behind the scenes for better results. The question’s answer should display an AI result that tells you it’s an AI answer.

Entering “What is a lunar eclipse” should result in something like “Gemini says: A lunar eclipse occurs when the Earth passes between the Sun and the Moon, casting a shadow on the Moon and blocking sunlight.”

Entering “lunar eclipse” should result in a list of websites on the topic. It technically does, but you need to scroll past the AI’s response to interpreting that as a question, which it’s not, and a list of suggested questions you might want to ask (the search query wasn’t a question to begin with) and a list of “People also ask” questions which is somehow different than the previous section.

Gross and critically, less useful. Google is blind to how to be a search engine because they squirted too much AI into their own eyes. AI doesn’t require Google zero. It requires better product design.

There is much more consumer value in AI enabling something that was unattainable before (which is why consumer vibe coding is exploding) than it is in making something already easy a tiny bit easier (like arranging an Uber).

Enterprise customers benefit from both evenly; saving time means saving money. Enabling someone to do something they couldn’t means making more money. It’s an entirely different dynamic than the consumer market. Because of the cost of training new AI models and customers using them, companies need to target consumers and enterprise customers.

For reasons that defy logic, enterprise customers will tolerate crappy products. Despite consumers increasingly accepting crappier products, they still need them to be better than enterprise customers. Big Tech either doesn’t know how, or doesn’t care to, make good products. It’s likely both at this point.

If they paint AI on everything, you can’t avoid using it. So even when you don’t mean to or don’t care to use it, you can’t avoid it and they get to tell shareholders you’re an active AI user.