Isn’t Microsoft also a trillion dollar corporation? If we add their 39% market share in foundation models (likely due to enterprise use of Azure OpenAI Service) to OpenAI’s 9% market share, the result is around 48% market share, compared to Google’s 15%, which is less than half of the MSFT/OAI pair…not to mention a cursory comparison of Gemini vs ChatGPT apps.
Just because OpenAI isn’t in “extraction mode” yet doesn’t mean it’s not a scary monopoly.
Yes, Microsoft is one of the world's biggest companies, and it underinvests in research and development, preferring to hoard cash. OpenAI is in effect a client state of Microsoft that Microsoft is using to make Google look flat-footed and force them to enter the chatbot market. Nothing that transpires between Microsoft and OpenAI is really at arms' length. Personally, I don't think this is a positive development for the industry or for humanity in generally. We were doing better before we had sycophantic robots confidently misleading us.
> Yes, Microsoft is one of the world's biggest companies, and it underinvests in research and development, preferring to hoard cash.
I'm not sure if you're being sarcastic, but:
1. If you're talking about basic research, Microsoft Research has been a thing since the 90's, is highly prestigious, and has published far more papers than Google, based on their respective research websites. (To be fair, Google started much later.)
2. If you're talking about product development, MSFT is vastly more diversified in terms of revenues than any of the "Magnificent 7" because of their varied product lineup.
3. The basis of their relationship with OpenAI is literally them investing double digit billions to catch up on the AI race once they recognized the opportunity.
> OpenAI is in effect a client state of Microsoft that Microsoft is using to make Google look flat-footed and force them to enter the chatbot market.
I'm not sure about Microsoft's influence in OpenAI's strategy, but it's pretty clear Google was caught flatfooted by their own strategy of locking away transformer technology behind products that didn't threaten their search monopoly. There's a reason the researchers who invented transformers had to leave and start a different company to bring its true potential to the market. Which, even if it was just a chatbot, is what has kicked off the AI boom.
Everyone I've encountered thinks of Microsoft Research as a bad pattern, that includes all the refugees from MSR Silicon Valley who joined Google after Microsoft dissolved it in 2014. Perhaps it is a bias of the people I've worked with in my career but the core early contributors at Google who came from DEC WRL also viewed separate research divisions as a bad idea.
Anyway my statement was meant to be objective. Look at how much Microsoft spends on R&D for the last 25 years, compared to the amount Google spends, in absolute terms and as a fraction of revenues.
Having a separate research division being an anti-pattern is an interesting topic! I remember getting into a related discussion almost a decade ago with a professor who left academia to join Google, his point being product-driven R&D was strictly "better" than "bluesky" R&D because (IIRC) the work is more directly related to market needs.
My contention was that this ignores the transformative potential of long-range theoretical research. For instance, somehow very few consider Xerox PARC to be an anti-pattern.
From what I hear in the last few years even MSR has changed its ways to steer its research more in line with needs of product divisions, and I actually consider that a loss. Who knows what paradigm-shifting inventions like GenAI are being steered away from?
> Look at how much Microsoft spends on R&D for the last 25 years, compared to the amount Google spends, in absolute terms and as a fraction of revenues.
Hmm, at the risk of relying on sycophantic bots, AI overviews suggest most recently Microsoft spent 13.2% of revenues vs 14.8% for Google (and 30% for Meta!) Of course even a single % point is in the millions at their scale, but there are a ton of confounding factors including differing product margins and payscales (and CEO obsessions like Metaverse!) At least at a quick glance MSFT and GOOG seem comparable.
The problem here is how "R&D" is defined. Unfortunately, even day-to-day product development is lumped in with R&D. I've done "R&D" in academia, private research firms, and big tech, and they are all poles apart. "Actual" R&D is very researchy, often based in discovering new aspects of reality, whereas "product" R&D is just regular product development. Which could be considered discovering new aspects of the market I suppose. They are both valuable but on very differnt timelines.
Just because OpenAI isn’t in “extraction mode” yet doesn’t mean it’s not a scary monopoly.
Source, figure 2 in: [1] https://iot-analytics.com/leading-generative-ai-companies/