We just came back from our visit to Mountain View to attend Google IO 2024 in person. For Manfred it has been a flashback to the days of regular IO attendance as GDG leader and heavily involved Android developer and open source community member. For Lukas it was the first bigger conference as university student, and his first visit to the Googleplex, thanks to our friend, host, and guide Matthew McCullough.
Two days of attending keynotes and sessions, and numerous chats with our peers left us with lots of impressions that we discussed and talked about and want to share with you all.
Before we dive into the ideas, thoughts, and feedback, following is a list of our activities. Every attendee chooses different sessions, talks to different people, and checks out different displays and showcase pieces:
- Google keynote
- Developer keynote
- Whats new in Android
- The State of CCS and Web UI
- Whats new in Android development tools
- Gemma models: Unveiling the latest advancements
- Quantum computing: Facts, fiction, and the future
- AI as a tool for storytellers: A conversation with Turing award winner Ed Catmull
- QA session with Google product leaders for GDEs and GDG leaders
- People-centered AI: Creating responsible generative AI products
- How to build an adaptive UI with Flutter
- All displays and showcase areas
Many, many others presentations and workshops are also available online as recordings with additional material, but let’s see what our takeaways and thoughts are:
AI is everywhere
Even before we attended the event, we knew that its all going to be AI, AI, AI. The consumer keynote with the count of over 120 mentions confirmed it all and it didn’t really slow down. Given all the hype and craze that has been surrounding AI that is not surprising. Google just had to play along to stay relevant and please investors. However, despite the long history of Google as leader in the ML space public perception has shifted against them. We believe that Google really stepped up and showed that they are more than just a contender in the field of AI. Maybe with all the new AI-powered features reaching consumers, the perception will change.
AI is not a product
Google seems to have to finally arrived at the understanding that AI is not a product. So many other companies in the space, including OpenAI, still seem to think AI is product. Personally we are not even convinced AI is a feature for some specific product. In many cases, AI really is just an implementation detail.
And Google is going all in on that idea. AI-powered features are in search, workspaces, photos, the development tools and the platforms. One good example is that Google is planning to ship an LLM on Android devices. And not as an app, or a specific widget, no .. it will be part of the installed libraries. A tool available for other apps on the phone to use to provide a useful feature. And the app designers can either showcase that something uses the LLM, or completely ignore that and just show off the cool feature.
Furthermore Google is working on new feature-sets or products that will become possible thanks to AI. Things like a wedding planner, an office manager, a holiday planner, a DIY repair assistant, and others are all becoming an increasingly possible as much products than currently available.
AI integration is not simple
All of the new potential features and products planned by Google show that they have moved on from the trough of disillusionment in the hype cycle around AI. The general public and many companies are still on the peak of inflated expectations or even just on the way there. When we look at all the AI and data platform companies, we just scratch my heads in disbelief – many seem to have no idea what to do apart from slapping an AI label on whatever they do.
Google on the other hand is walking the walk. They built up their data centers, created new specific processors (TPUs and CPUs), and are now rolling out that power across their many products for new exciting features. Lots of these features will probably be duds, but some of them will be awesome, and we saw a few very promising demos. We are definitely excited to see what is to come.
AI risk mitigation is even harder
One of the aspects that also showed up in force, is that AI usage it risky. Hallucinations are common, wrong or inappropriate output shows up regularly, and the border between helpful and creepy features is a very narrow edge. Google is very well aware of it and has huge teams working on solving these problems, yet even their keynotes had demoes that failed in the eyes of some audience members. Finding my car number plate in my own photos easily might be good, but what if someone else takes lots of photos and uses the same query for more nefarious reasons.
We believe that smaller features that are tighter controlled in terms of applicable usage and heavily integrated in a specific product will help with the risk and creepiness reduction. However in the AI is just a tool, and to a large degree the human users decides upon its usage and application. AI can be good and bad, just like fire.
Communication is hardest
Last but not least we need to chat about communication as it continues to be a weak spot for Google. The keynotes, the sessions, and Google communication in general can suffer from the following issues:
- Sometimes it is not clear who the target audience for a specific session or message is? Investors, employees, consumers, developers? Often Google seems to want to address all at the same time, and then succeeds confusing everyone. Keynotes get too long and want to cover all different aspects and leave everyone bored or baffled during some segments.
- Timing and transparency of communication is off. If you are restructuring your teams and significantly reduce the employees backing a specific technology, you must have a public statement ready and launched the day the restructure happens to avoid negative impact on the related communities. For example, the Flutter community suffered significant unrest due to the recent layoffs, yet at the conference Google seemed very much committed to Flutter and leadership confirmed that commitment when asked directly in the GDE session. Sending that message out earlier and loudly would have helped.
- Professionalism is good, but Google seems to be overdoing it. Sure, the event is broadcast world-wide, has a huge audience, and is very important for the business of Google. At the same time sessions should not be full of looks at the teleprompter from stiff presenters, who don’t seem to have any joy in presenting to the audience. We need fun and enthusiasm back, maybe even a failure here and there, and a laugh!
Other random thoughts
- The crowd management in terms of accessing the site of the event was understandably tense, due to the protests.
- Android development and tooling are amazing. A good time for Lukas to play with it.
- Google’s work on web technologies continues to drive improvements – we love it.
- The possibilities to play with models, tweak them, integrate them in your products, and generally learn more about AI are overwhelming and super interesting at the same time. What do you want to do first?
- Quantum computing is years away from affecting your every day life, but also mind-blowing and super interesting.
- Suncream and a hat are very useful for an outdoor event in California.
Conclusion
Now some of our feedback sounded kinda harsh, didn’t it? Just take it as a kind hearted suggestion from us to Google. We loved attending the show and strongly believe that amazing things are in the pipeline and ahead. Google understands AI, is integrating it in many of its products, created focus for the teams, and has numerous other amazing efforts on the way.
It’s just one of the construction or DIY jobs we regularly engage in – nothing is as simple and fast as it looks at first, but in the end the result is always more rewarding when it wasn’t easy to get done.
Manfred and Lukas