Microsoft is putting OpenAI’s technology into almost everything it makes. This is creating single-source risk that I think can only be resolved by the full acquisition of OpenAI.
Last week’s Microsoft Build 2023 was all about how generative AI, ChatGPT and OpenAI’s technology would enhance Microsoft products, as well as start to make a dent in Google.
There were four key announcements, starting with Copilot, which puts generative AI technology into all of Microsoft’s core operating system and Office 365 products.
This is what Microsoft hoped to achieve with Clippy all those years ago, the dream of which now finally may be realised. In addition to the assistant in Office apps, Copilot will also exist as a button on the taskbar and be able to assist the user with other tasks that are carried out on the PC.
The demos were excellent, but as ever in these situations, one will never know how good it is until it is available in the wild. And it has to be said, Microsoft’s history in this department is pretty bad.
The quality of Copilot will become a yardstick of how users view the quality of Microsoft’s core products, which substantially increases Microsoft’s dependence on OpenAI as a technology supplier.
Bing and ChatGPT
One of the big shortcomings of ChatGPT is the fact that its knowledge is locked to everything before September 2021. To update it would require fine-tuning the whole GPT model all over again, which is time-consuming and very expensive.
To fix this, Microsoft is bringing Bing Search to ChatGPT where the search results from Bing can augment ChatGPT’s answers to keep it current and up to date. Microsoft will also ensure interoperability with ChatGPT plugins, so that plugins can call both ChatGPT and Bing and integrate the results.
This pretty much already exists as Bing Chat, but its inclusion in ChatGPT – where the free version is thought to already have 1 billion users – raises the possibility of a huge increase in the use of Bing Search.
Google has already done the same with Bard, but its usage is likely to be a tiny fraction of that commanded by ChatGPT. I expect that Google will soon include generative AI answers as part of its search algorithm.
And, given that its search is much better than Bing’s, I expect that the overall quality of the service will be better. This is what I have experienced with Bard to date although, like ChatGPT, it makes the most ridiculous and obvious errors, firmly supporting my view that, despite the illusion, these machines remain as stupid as ever.
The kind of enquires that are conducted on ChatGPT and Bard remain very different to search, so I do not see this as a reason to get worried about Google’s search business yet.
Azure AI Studio and Microsoft Fabric
Two other key announcements of note:
Azure AI Studio: a platform to build one’s own models as well as build the functionalities that sit on top of them. This will also build support for AI safety so that developers can test their apps to minimise the eventualities when they go off the rails.
Microsoft Fabric: essentially a direct competitor to Snowflake which is a data warehouse, analytics and manipulation platform. This is a natural move for Microsoft given its existing business, but it has some work to do to make it as good as Snowflake.
In summary, Microsoft is saying it’s infusing all of its products with generative AI, which I think could take their functionality to the next level. Microsoft is hoping that this will enable it to steal share back from Google in productivity and also make a dent in search where Google has basically been unopposed for more than a decade.
Microsoft has a single-source dependency problem
While all sounds good and fits the current AI craze perfectly, the catch is that it creates a big single-source dependency for Microsoft on OpenAI.
Assuming these products prove to be popular, should OpenAI cease supplying Microsoft with technology five years down the line, Microsoft’s core business and profitability would take a hit and its share price would crater.
For a company of this size, this is a risk that it cannot afford to take. So I suspect that Microsoft will have no choice but to acquire OpenAI outright.
This is why RFM research considers Microsoft and OpenAI to be one company when it comes to the quality and level of advancement in AI. Because Microsoft already owns 49% of OpenAI, it will be virtually impossible for anyone else to acquire it, leaving Microsoft as the only candidate.
The more success that Microsoft has with AI, the greater this dependency risk will become. Until it manages to acquire the company, this will be a growing problem. Microsoft still looks pretty expensive on a valuation basis and so I would still be looking elsewhere for investments in AI.