Google I/O 2023 – Fightback

Google mounts a steady fightback.

• Google is clearly still on the back foot, but the sense of panic and disarray has gone to be replaced with a sense of purpose and its hardware division has finally come up with a compelling product.
• Unsurprisingly, the whole show was all about generative AI and how Google is not the laggard that the market seems to think it is and how it will use it to make its products better.
  • First, models: Google announced its next foundation model which is imaginatively named PaLM 2 (pathways language model) upon which it is going to base most of its services.
  • PaLM 2 is a foundation model which means it serves as a starting point from which to train other models to do specific tasks like a medical knowledge base or be a chatbot like Bard.
  • While Google declined to give details on PaLM 2 it did say that it would be available in a range of sizes, the smallest of which it says will be able to run a mobile device.
  • Besides augmenting productivity, it is also being included in consumer services which are now open for anyone to experiment with.
  • Top of the list here is Bard which is already running on PaLM 2 and can link to other systems such as Adobe Firefly to also generate images.
  • This is an interesting development and fires the starting gun on the next ecosystem battle where chat providers fight with each other to be the best LLM on which to base one’s service or plugin.
  • Second, productivity: Here Google is doing precisely the same thing as Microsoft and including generative AI in its office-like apps that make it easier and faster to create content.
  • Bard will now be able to create spreadsheets, tables and slides but also search through all of the user’s data and easily find what is being searched for.
  • This functionality runs on the device or within the user’s private space on Google Cloud so that there is no data leakage or insecurity.
  • Microsoft has announced exactly the same thing and its dominance in productivity apps is likely to ensure that for as long as Google does not come up with something groundbreaking, its share will remain intact.
  • Third, the consumer ecosystem: Gmail, photos, maps etc will all be enhanced with generative AI to improve their usability.
  • From creating an effective complaint to moving objects around in photos and immersive route previews in maps, Google is trying to enhance its ecosystem services to keep both Android and iOS users engaged with its apps.
  • Fourth, Search: which is the current elephant in the room as the theme doing the rounds at the moment is that chatbots like ChatGPT make search obsolete.
  • The fact that pure generative AIs have to be frozen in time to prevent them from going crazy, makes them wholly unsuited for search which is why Google is going for a hybrid approach.
  • This approach will of course enable the current advertising model to continue, but it does make sense in its own right.
  • The generative AI gives a conversational answer to an enquiry, but this is also merged with the traditional search algorithm to give better results.
  • A good example given was an inquiry that starts with the user looking for a bike that is good for hills and a daily 3-mile commute.
  • Bard comes back with a paragraph about these sorts of bikes and the features or characteristics that such a bike would need to have but then follows it up with a list of potential products to consider.
  • Google has effectively merged generative AI and search and uses the conversational front end to provide a better user experience.
  • In my testing, I have found Bard to be better than Bing because Bing is relying on Bing Search to affect the search and then GPT-4 to formulate the answer while Bard uses the superior Google Search.
  • Hence, for the moment, there is no immediate threat to search, and Google has done a pretty good job of creating a product that both uses generative AI to give good answers but preserves its ability to monetise the data it collects.
  • Fifth, hardware: where Google typically launches a series of devices which amazing AI features that no one cares enough about to buy.
  • Furthermore, flaws in hardware routinely ensure that they receive mediocre reviews which further limits their sales.
  • However, Google has finally come up with a product that may appeal to many people.
  • The idea is simple (as all the best ideas are) which is the Pixel Tablet which comes with a magnetic wireless charging dock which massively enhances its functionality.
  • The dock provides a place to store the tablet (looks like Google Nest Hub) and ensures that it is always charged but when docked it automatically uses the speakers in the dock for sound.
  • It has Chromecast built-in and so it can be quickly used as a viewing, listening or video calling device and at the same time can serve as a hub from which to control all the smart devices in the home.
  • This is the first device that I think gives Google a shot at a hardware hit, but it remains to be seen whether it has made the usual errors in hardware that will limit its appeal.
  • I think that its $1800 folding Pixel is deeply uninteresting, offers very little that others do not and will not sell in any meaningful volume.
• The net result is that there was not a lot here that others are not already doing meaning that Google is playing catch up in terms of what it is making available for regular users to tinker with.
• That does not mean that Google is not a leader in AI behind the scenes, but it has been far more timid than its competitors in terms of getting its wares to market.
• This is because it has the most to lose in this space given its dominant position in search compared to Microsoft which has nothing to lose and can come in guns blazing (like Apple did in smartphones).
• This was a pretty good show after the disastrous Paris event where Google made itself look like a deer in the headlights and should go some way to allowing people to realise that Google still has a lot to offer in this space.
• That does not mean that I want to buy the stock but if it underperforms on the basis of losing its edge, it is one I would look at.