Everything about Google’s AI strategy has been confusing so far

Earlier today, Google made a series of announcements regarding Google Assistant, Bard, and Gemini — the large-language model (LLM) Bard is based on. If these names are already confusing to you or if anything about Google Assistant has ever been confusing to you, prepare to be even more confused. A thousand times more confused.

Google Bard is now Gemini, so the Gemini chat service will run on the Gemini LLM, while Duet AI in Google Workspace is also becoming Gemini, but it might not have all the features of Gemini because it’s limited to Gmail and Drive, mostly. Plus there are three tiers of Gemini models, which require hardware as small as your phone or as large as Google’s servers — and one of those tiers will require you to purchase a monthly $19.99 Google One storage plan.

Plus, there’s a new Gemini app that is replacing Assistant on your phone, but it won’t do everything Assistant does (we don’t know the limitations yet), and it’s not coming to smart speakers or TVs or your watch. This app will have access to the more powerful Gemini tier because it talks to Google’s servers. So yes, your Pixel or other Android phone might have one Gemini model built-in for local processing, and a more powerful Gemini model accessible through the Gemini app. Oh, and nobody knows what happened to Google Assistant with Bard — we presume it’s dead because it’s basically the new Gemini app?!

Gemini, like Google Assistant before it, will suffer from a set of drawbacks we’ve all already experienced. Some features are rolling out widely, while some are only coming to US English today but will get a few more languages next week. (Why not wait a week to roll it out to more people in one go?) Everyone else will get them at some point.

Ah, also, some will work worldwide, while others don’t have EU approval yet. So everything I said? If you live in the EU, you might as well forget most of it.

Now, I need you to sit down and draw the Venn diagram of your language, your IP address, your phone, your other devices, your Google One payment plan, and whether or not you have a personal Google account or a Google Workspace account. Only then will you properly understand what Gemini features you’re getting and which ones you’re not.

Look, if it’s not clear from everything I’ve written so far, I’m bitter. I really am. And I’m annoyed. I already had trouble explaining to the most techie people in my life why on earth Google Assistant supports one feature on their phone but not on their watch or Chromecast or car. Or why they can ask for something in English but not in French. Or why they could connect certain music and video streaming services to their American account but not their British account.

I realize Google has had to move in fast on AI. I realize the shareholder pressure must have been immense when ChatGPT and generative AI took the world by storm, while Google was still having fun with LaMDA and PaLM2 thinking it had all the time to polish them. I realize the company had to shuffle teams and resources around, in extremis, to fight back and show the world that it was still the king of this land we call the Internet. And I realize that, because of all of this, it had to break things fast and fix them faster.

But users, like you and me, we can’t handle this mess. No one has enough time to plow through the madness and understand it. Not even me, and this is my job. No one can follow through with these naming changes, multiple tiers, multiple features, and confusing availabilities across devices, languages, and locales.

Clearly, this is not how I thought things would be going, eight years after Google Assistant made the exact same mistakes. This is not how you roll out a product or get people excited about new technologies.

Artificial intelligence, the term on its own, is scary for most people. They don’t understand it, they don’t know how it works. And instead of easing people in and demystifying it, Google’s steps over the last year have just made everything more confusing for everyone.

From a consumer perspective, you and I should not care about how the sauce is made — even if takes a thousand cooks in the kitchen. We should just get the sauce. Period. What Google has done over the last year is invite us into an open-kitchen restaurant, while a thousand chefs argued and fought over how the sauce should be done, and the maître d’ kept asking us to be patient. Every now and then, he’d bring us a bit of sauce on the back of a spoon to let us taste it before he hastily hurried back to see if the chefs had figured out the recipe or not. Then the restaurant owner would fire half the chefs every 10 minutes and hire new ones.

Look, metaphors aside, I don’t think anyone can disagree that Google’s strategy — or lack thereof — around AI has been a scattered, incomprehensible mess. Today’s announcements don’t make it any better or easier.