Google Next '24 - Strategy and AI

We have a clear picture of where Google is investing in with their AI for enterprise customers and their future feels underwhelming and lacks the "Googleness" we knew and loved.

I was invited to attend Google Next, Google’s cloud-focused conference, over the last week. The conference was notably focused on AI and, as a user-experience-focused AI developer, I have a few thoughts about the current positioning of AI at Google.

Google has always brought a unique attitude to their product development and the promotion around their releases. They often take a promising existing product idea or capability and “Googleify” it, making meaningful end-user improvements to deliver a better version that they can then scale easily to millions of people using their massive infrastructure footprint. Think Gmail, Google Docs, Google Maps,… and the list goes on. This strategy requires less novel or risky product development strategy and still creates compelling, new-feeling products that sell quickly.

This makes Google a bit of a bully in the market because it means they often squash or absorb products instead of creating new products. Their AI research has been very revolutionary but they struggle to bring that innovation across as a core net new product. Google Stadia has a special place in my heart in this space but the failure of that line of business further emphasizes the difficulty Google has with creating and sticking to something novel. This sets the stage for AI to be a fresh opportunity for Google to redefine itself.

Keynote Stage at Google Next 2024

The problem is AI in its current form offers unknown future value. It’s like touchscreen technology before the iPhone popularized, enhanced, and honed the experience of multi-touch gestures. It doesn’t matter if you build a better touchscreen at a better scale or ratio if the end solution isn’t bringing something unique to the experience. Designers and engineers took something potentially helpful and showed a compelling future direction with clear demonstrable examples that shaped the industry. (I still doubt we’re done improving touchscreen applications despite the general focus shifting to other digital mediums.)

Google is looking at this AI space and seeing some aggregate activities that they can bundle up as products but, because of the uncertainty, it’s hard to establish a clear bet. Google still hasn’t found that product space and hasn’t created a compelling product offering that differentiates itself in the market enough for any meaningful sticking power. So they’re sort of stuck trying to front some offerings that don’t differentiate and only marginally offer a better experience (if at all).

Visiting this conference, I see Google Next as a promotional activity to get enterprise-level customers to buy into a future where AI has a larger role in their businesses. Google wants additional buy-in and funding to pursue AI and to build themselves a stable space that can offset the dramatic changes coming to their search and advertising businesses. They want and need to sell this vision. But do they have anything to offer?

It’s obvious to me that there’s a disconnect between what large-scale companies think AI is and how creative engineers are looking to use it. Perhaps I shouldn’t be so surprised by this but I am because this is Google. The Google that helped evolve the web from websites to web apps.

To cover a bit of the details discussed in the conference, a majority of the focus was directed toward developer productivity tooling. I think it’s likely that this is due to the currently popular and wildly successful products (ChatGPT & GitHub CoPilot). This productivity focus is probably justified for Google since neither of those products came from them. Still, I’m not sure what Google can really do to “Googleify” those tools. It still feels like the teams at Google are simply trying to capture some of the market before their competitors take over the space entirely.

Other AI topics included the existing well-known use cases that don’t seem to bring anything new to the table: semantic search, summarization, sentiment values on data fields, chatbots for docs, video generation, AI agents, video captioning, translation, and other solutions that will bore anyone who’s been in the current LLM AI development space for over a year. I’m aghast that Google is targeting these use cases at all because of the difficulty of building something generic and useful. AI integration with Google Drive is interesting, but it misses the domains of their clients and will likely lead to an overly simplistic RAG solution that has comparatively little meaningful value.

When applying AI, the domain of the business matters a lot. As an example, one of the keynotes included a real-time sentiment analysis of comments on different products within a store including what product attributes were being discussed. It’s a cool possibility: get what people think of all of your products as concrete data based on what they said. However, if you have product names that themselves compute as a negative sentiment (ex: angry dolphin headphones), the whole thing has the potential to be thrown off. It doesn’t just work out of the box despite the ease of integration in Google demos. The reality is complex and evolving.

Google isn’t really innovating. Sure the ability to integrate AI into real-time pipelines is cool but that’s also not really AI. That’s just a data thing and something likely to come from AWS just as easily as Google Cloud. Without a meaningful lift to the AI aspect, I struggle to see where Google is “Googleifying” AI. Instead, AI at Google feels like another Angular, a web technology that Google heavily pushed toward enterprises and effectively alienated any startups along the way.

Gemini is released and has some pretty interesting capabilities due to the long context window but, as mentioned in a recent post on our channel, this doesn’t necessarily solve all of the problems in the space. Not only this, but other competitors are hot on Google’s tail here too, which limits the impact of their lift. While there can be an interesting platform integration with Google Workspaces, I struggle to see where many startups care about what Google is doing on their platform.

Now, admittedly this is a pretty negative take on the conference in general, which is focused around enterprise. But I’m someone who has an interest in advancing how AI is used beyond simply chatbots and the simple use cases at play today. Multi-modal is interesting but it seems like it’s largely being applied in the same way as non-multi-modal applications within Google (via Search, sentiment, etc.). I expected more from Google and suggest that if you’re looking for the latest engineering approaches involving AI, you look elsewhere too.