The other day, a friend asked me what I thought of the “We have no moat” Google document that got leaked and the position of open source.
The argument in the document is that open source is improving at such a fast pace, that it is silly to think that big, slow companies can keep up. Therefore, keeping everything that they develop in-house is not productive.
I found myself nodding at first, but I was following their logic, in the way they presented it: Yes, the open source community is amazing and it means that many more researchers can work on AI than just the people working on it from Google or OpenAI.
However, the argument is flawed in the sense that of course Google has a moat: it’s their products! AI is not the end in and of itself. It’s to drive a product. This is also why OpenAI can be successful: They provide the models for those that don’t want to host and develop their own. They have data and money (= compute-power) to train very capable models.
These open source models are great, but don’t forget to build something cool with them. And Google: Please do something relevant with them, have a vision and build stuff that your customers want. Don’t just a/b test fifty shades of blue for a single button and call it a day.
(Also posted on my LinkedIn feed)