Amazon – Bee in Its Bonnet

Bee tells the story of Amazon AI.

• Amazon’s acquisition of Bee is surprising in that it indicates just how weak Amazon’s in-house AI ability really is and brings forward the time when it will have no choice but to pay up for Anthropic to get its house in order.
• Bee is a West Coast-based start-up that sells a $50 device that looks like a Fitbit, which listens to everything that is going on around the user and creates summaries, memos and reminders based on what it understands.
• The device is little more than a microphone, Bluetooth connection and a battery (which is why it costs $50), with all of the computation being relayed to the cloud via an iPhone and the results being displayed inside the Bee app.
• Reviews are fairly mixed, with the device being able to understand language fairly accurately but getting speakers, people and locations muddled up on a pretty regular basis.
• In short, Bee is a classic first-generation device with a slightly clunky device design and software that works reasonably well but is clearly nothing remarkable.
• This can be greatly improved over time, as the assets of this company are obviously its software, but it raises the question of why Amazon would want to acquire it.
• There are a few possibilities:
  • First, technology: which immediately raises the question of how troubled Amazon’s AI technology must be if it feels the need to acquire Bee.
  • The technology appears to work reasonably, but it is nothing remarkable and is the kind of thing that Amazon has been working on since it launched the Amazon Echo over 10 years ago.
  • If this is the reason why Amazon has purchased Bee, then I suspect that its in-house AI efforts are in even worse shape than I feared.
  • Second, user base: which is pretty unlikely as the device only became available earlier this year, and the company says the Bee Pioneer Edition is already out of stock.
  • This implies that the user base is probably 10s of thousands at most, all of whom probably already have an account with Amazon, meaning that the value of the user base to Amazon is pretty much zero.
  • Third, people: which has become a common event in the current AI arms race that is going on between the major players, where a company is acquired just for its people.  
  • Bee was founded by a group of executives and engineers who left Twitter after Elon Musk took it over, and so I am not convinced that Bee was a secret pool of AI talent with a high chance of suddenly creating AGI out of thin air.
  • Hence, I don’t think Bee is being acquired for its people.
  • Fourth, hardware: is another common reason for acquisition, but the device that Bee does not look like anything particularly special and has no remarkable function or feature that differentiates it from anything else.
  • Furthermore, Amazon has been making its own hardware for years and with Panos Panay now in charge (who knows how to make devices), I find it highly unlikely that Amazon felt it had to own this hardware product.
• The net result is that the most likely reason why Bee was acquired was either for its AI technology, or it was about to run out of money, and Amazon saw it could hire people with some edge and device AI experience on the cheap.
• Bee’s AI is demonstrably nothing special, and if Amazon felt that it needed Bee to make its own AI better, this speaks volumes about how weak Amazon is when it comes to creating its own AI.
• This serves to reinforce my long-held opinion that AI is a major blind spot for Amazon, and while it can do a great job of hosting the AI of others, it really struggles to have anything worthwhile of its own.
• Consequently, I think it more likely than ever that Amazon will base more and more of its AI on Anthropic, developing a dependency that almost guarantees that it will acquire Anthropic at some stage.
• Anthropic’s business model is highly questionable (as is OpenAI’s) as both lose billions of dollars despite having millions of users, meaning that when a correction comes and the money dries up, they will both be in trouble.
• This is the point at which I expect Amazon to take Anthropic over, as it cannot afford to have this kind of dependency on a 3^rd party.
• The only good news here is that in a correction, the price it pays could easily be less than the price it has already paid to get involved.