- Google is putting its latest AI model into the hands of billions of people.
- The company has such a wide range of popular products that few rivals can match it.
- OpenAI and other AI startups will have a hard time matching this reach unless they partner with other tech giants like Microsoft.
Venture capitalist Erad Gill has an elegant theory about who wins each new wave of innovation. Either the start-ups do the best, or the incumbent tech giants capture most of the value.
In the first Internet wave, start-ups took 60%-70% of the loot, leaving the rest to incumbents. During this time, Google, Facebook, Netflix, Salesforce, PayPal, and others were born.
The mobile revolution left incumbents like Apple with 80% of the loot and start-ups with just 20%. Gill said that in the cryptocurrency space, most of them are startups.
After this week’s Google I/O, AI is likely to be in an 80/20 situation again, with a handful of big tech companies likely to benefit the most.
A big takeaway from Google’s annual conference is how the company is incorporating the capabilities of its flagship AI model, PaLM 2, into many of its existing and very popular products. was. Some examples:
- Google Docs lets Google’s AI models help create job descriptions. Then hit the button to port it to Gmail. Or vice versa.
- Grab a photo from Google Photos, drop it into Google Slides, and let Google’s Bard chatbot write the caption.
- You can ask Bard to create a table listing the customers and the model will do it, even adding additional columns as requested. Then press the button and it will be imported into Google Sheets.
- For Google Slides presentations with photos, you can ask the company’s AI model to create speaker notes based on the images.
- Point the camera at objects using Google Lens and have Bard write captions for them.
CEO Sundar Pichai said there are currently 25 products and features leveraging PaLM 2. What other tech company has such a wide and popular set of services that it can use to bring AI to the masses?
Microsoft is probably closest to Google in this regard. The company has Office 365, with hundreds of millions of paying users using its word processing, email, spreadsheet, collaboration and presentation software. All those Windows computers are still there with the huge Azure cloud business. The company is frantically injecting new AI capabilities across this product portfolio.
Other tech companies such as Meta, Amazon, and Apple may also have some of this breadth of products, but they lack the lead in AI. Or focus on something else.
This adds a new perspective to the OpenAI-Microsoft deal. OpenAI may have the best AI model to date, but without Microsoft, they don’t have the products to bring this technology to life in multiple useful ways. So OpenAI may need Microsoft more than Microsoft needs OpenAI.
Google looks to quickly build an AI moat
The existential threat posed by this powerful combination has forced Google from using AI internally to aggressively introducing generative AI capabilities into all its major products. The speed with which the company has achieved this was revealed at his I/O on Wednesday. At a conference, Insider ran into his manager at Google’s Senior Product, who had just watched a developer’s presentation. This person, in theory, should know what’s going on with Google’s latest AI services, but he knows there are some new features he plans to add to the products he oversees. I was. This is how fast companies move.
Google’s top position is not a foregone conclusion. As an engineer’s note recently highlighted, we certainly face threats from open-source AI models that anyone can access and adapt. But Google knows it can build another moat using products already used by billions of people.
Google says it currently has six products that are used by at least 2 billion people, with 15 products each being used by 500 million people. At IO, Google demonstrated its response to the OpenAI threat by demonstrating generative AI capabilities across many of these core products, from Android to Workspace.
We also previewed how we will integrate generative experiences into our most important product: search.
Tech incumbents just need to be good
“An incumbent can be 50% better than something else, but as long as they bundle it with an existing core product that has a lot of customers, they can still win,” said VC investor Erad Gill. wrote recently, citing the company’s success. Comparing Microsoft Teams and Slack.
During the keynote, Google implicitly emphasized another major advantage at the moment: infrastructure. The company announced that Character.ai, a startup that develops human-like chatbots, has partnered to use Google’s prized TPU processors to grow its business.
Generative AI startups’ reliance on cloud computing will enable big tech companies to aggressively pursue investments and multi-year deals in emerging businesses that are viewed as competitors.
Has Google lagged behind OpenAI? Possibly, but with billions of people operating his PaLM 2 model, and eventually Gemini, Google has gathered a ton of feedback from how people operate this technology, By continuing to feed it back to a wide range of products, we will be able to act quickly. This is the data used to train future iterations, creating a network effect that drives Google even further.
“The situation will improve over time,” Pichai said toward the end of his keynote.
Any Google tips? Hugh can be reached via encrypted email (hlangley@protonmail.com) or encrypted messaging app Signal/Telegram (+1 628-228-1836).
