Google I/O Thoughts?

22 points by tsycho 1 year ago | 20 comments
Interesting announcements (so far):

* Gemini 1.5 Pro context window increased to 2 million tokens * Gemini Flash launched (lower latency model) * Increased image and video generation capabilities * Cool new interface that can analyze live video * 6th generation of TPUs (Trillium) that deliver faster performance and lower TCO

  • verdverm 1 year ago
    Always enjoy the possibilities that I/O shows

    Other impressive things:

    - "where did I leave my glasses?" (live video feed with memory)

    - asking for a summary across emails / meetings

    - agent demos (like handling all the steps to return shoes, building physics learning examples based on user provided preference "basket ball")

    - "why is my record player doing this?" (ask question while recording video, get genAI response with links)

    - Google Workspace Gemini Panel (like that it sits on the side rather than taking over the UX)

    - Build a JSON object of books & authors from a picture of a bookshelf

    - Context caching for model calls (saves sending large amounts of text for every call)

    Google Astra seems like it will be the foundation for a number of next gen capabilities

  • atleastoptimal 1 year ago
    Google has "demoed" dozens of seemingly world-changing products that lead nowhere (remember Google Duplex?). It's impossible to ascertain from an event like this the extent to which demos are cherrypicked or faked entirely.
    • jfoster 1 year ago
      They used to launch products at Google IO, then over the years it shifted into previews (some of which didn't eventuate, like Duplex), and now it's just demos of stuff they might turn into a product one day.

      Is there a single thing that they showed at Google IO this year that can be used right now?

    • nunez 1 year ago
      I think Duplex is still in use; it's just in the background doing its thing.
      • jfoster 1 year ago
        They might (maybe?) be using it to check/update open hours for places on Google Maps, but you can't make reservations using any of Google's Assistants.
    • LauraMedia 1 year ago
      It feels a bit like this tech event I've always been looking forward to died to be honest. IO was always cool, you could always see general product updates, but this year seems to be AI only. Half of the talks/videos will also be AI focussed, it's a bummer because it's always a source to learn new SDKs or design ideas through them.
      • tsycho 1 year ago
        All the video understanding and agentic stuff (if it actually works at low-enough latency in real life) is very impressive.

        Put this on a robot, call it Jeeves, and it becomes a pretty cool personal assistant :)

        • tikkun 1 year ago
          Looks great, but I've become skeptical of Google demos and as such will believe it only after I've used it :)
          • rogerthis 1 year ago
            Every year they demo lots of things I never see later.
          • nunez 1 year ago
            My thoughts:

            - Is this the first I/O that released _zero_ new hardware?

            - Very very sad to see them make EVERYTHING in this conference about Gemini

            - Google video search is crazy interesting

            - Google Search being AI-first is definitely not.

            • niutech 1 year ago
              They have released Pixel 8a and you can remove AI from Google Search by clicking the Web filter.
            • Adrig 1 year ago
              I used to look forward to it. I liked Google positioning with its Pixel line up, their prosumer software and what they cooked in the lab (AI, Project Starline)

              Now I don't even bother navigating through the announcements that are never launched, the fake demos, the tiptoeing around every potential issue... What a shame.

              • duringmath 1 year ago
                Underwhelming so far, everything is disjointed and is months away from being available via labs.
                • tsycho 1 year ago
                  I thought Gemini 1.5 Pro is available immediately, and the Workspace stuff is coming in 1 month.

                  Fair point if you are talking about the video understanding stuff, I don't know when that's coming.

                  • duringmath 1 year ago
                    It's just missing the wow factor, they can still turn it around if the rumors about the new pixie assistant are true but so far yesterday's preemptive OpenAI event seems like it was unnecessary.
                    • 1 year ago
                • rvz 1 year ago
                  This one is worth watching.

                  The AI race to zero is accelerating and thousands of startups have been steam-rolled before they could even reach the market and VCs.

                  Even if they survive this, by the time they have reached VCs for funding, they will finish them off by explaining that they won’t pay for their product when the best AI tools are there for free or close to free.

                  • dzhiurgis 1 year ago
                    Google is becoming a Toyota. Useable, but utterly behind.
                    • solardev 1 year ago
                      Who is Toyota behind?
                    • 1 year ago