AI needs adult supervision

Whether OpenAI did or didn’t copy Scarlett Johansson’s voice to front ChatGPT’s “Sky” personality, it’s somewhat telling that people don’t seem to find it hard to imagine OpenAI CEO Sam Altman capable of the misappropriation. Despite the promise of AI, we keep seeing examples of AI run amok, as with Google’s attempts to set AI loose on search results (to very bad effect in some instances). Speaking of the Johansson incident, Charlie Warzel writes in The Atlantic that this seems like “yet another example of a tech company blowing past ethical concerns and operating with impunity.”

However, as Warzel continues, this isn’t just an OpenAI thing, but rather “a reminder of AI’s Manifest Destiny philosophy: This is happening, whether you like it or not.” And let’s be clear: Sometimes we really, really don’t like it.

Not very comforting, is it? That’s why I’ve argued for the need to have more open source in AI to keep vendors honest. It’s also why we probably need more Microsoft, and less OpenAI, to help AI operate in a more grown-up fashion and build trust with corporations and individuals.

Trust pays

I really can’t believe I just said that. After all, I spent more than a decade raging against the Microsoft machine, even as it waged war on open source. But that was then, and this is now. Despite Microsoft’s early fight against Linux and open source, the company has been an active, trusted participant in a variety of open source communities for years.

That trust is important because trust is precisely what’s missing in AI.

Whatever Microsoft’s stumbles with open source and the U.S. Justice Department in the past, it remains perhaps the most trusted company in tech. Enterprises seem prepared to pay for that trust. According to a 2023 Piper Sandler survey of CIOs, Microsoft stands to gain the most in the spending spree on generative AI. A whopping 49% of those surveyed said Microsoft is their most strategic vendor for AI, vastly outpacing OpenAI (17%). This trust in Microsoft has remained consistent over the years. Back in 2013, for example, Microsoft was seen as CIOs’ go-to vendor for cloud. Of course, today AWS is a $100 billion cloud vendor and Microsoft is not, which could mean that OpenAI, like AWS before it, could end up taking the bulk of AI software spending just as AWS did for years with cloud.

But cloud is a long game, just like AI, and Microsoft is doing just fine in cloud. According to the company’s latest earnings, its cloud business topped $35 billion in the quarter. That isn’t an apples-to-apples comparison with AWS’ $100 billion run rate, but analysts currently peg AWS market share at 31% and Microsoft Azure at 25%. This is impressive, given that AWS had roughly 100% of the market during the early years, and held a commanding lead for years. No more. Much of that comes down to CIO trust in Microsoft, as well as Microsoft’s aggressive maneuvering in AI.

Play at your own risk

Worried that it was way behind Google in AI technology, Microsoft invested $1 billion in OpenAI back in 2019. In 2023, the company doubled down on OpenAI, bringing its cumulative investment in the AI leader to $13 billion. This close collaboration with OpenAI has made Microsoft’s Azure cloud service that much more interesting to companies anxious to put AI to use. During this time, Microsoft has mostly kept its distance from OpenAI the company, even as it has benefited from OpenAI’s technology, as with its release of OpenAI’s GPT-4o on Azure earlier in May 2024. 

This is great for Microsoft, but for the sake of the AI industry, we need more Microsoft and less OpenAI.

OpenAI has been a roiling mess of conflicts and concerns over the years, with board upheaval due to worries that Altman had dumped the company’s founding principle of AI safety over profit, high-profile employees quitting over concerns that the company puts revenue before safety, and more. To Warzel’s point, this can be seen as standard Silicon Valley: Utopian dreams crush near-term concerns about societal impact.

But AI, despite the machines-rule-over-us veneer, is ultimately about people, as I’ve recently written. People have the potential to stop or at least seriously slow down the infiltration of AI into our lives if the driver of AI change is seen as heedless of human concerns. Getting Microsoft into the driver’s seat would allay many of these concerns. Microsoft, after all, is the same company that largely tamed the unruly mess that was GitHub’s corporate culture, without exerting such a heavy hand that developers dropped the code collaboration tool. GitHub was popular before Microsoft took over, but Microsoft’s leadership has helped cement GitHub’s place in the enterprise.

The same thing could happen in AI. As Forrester Vice President Brandon Purcell stresses, “The opacity of AI has created a significant trust gap that only transparency can fully bridge.” Part of this is about opening up code (including model weights, etc.,) so that enterprises can more fully trust large language model outputs, but it’s also about trusting the companies behind those LLMs. Microsoft is good at open source and good at fostering corporate trust. AI needs adult supervision, and Microsoft may be perfect to play that part.