Microsoft, AI, and Ethics (Premium)

On the eve of Build 2023, it’s perhaps useful to examine the ethics of Microsoft’s sudden AI explosion and whether our perception matches reality.

First, a story for context.

Everyone in our community seems to know that Microsoft at some point fired an entire team of internal Windows testers and created the Windows Insider Program as a bigger pool of free labor. But one wonders if this story is, if not apocryphal, then at least not the entire story. That is, perhaps what Microsoft found over time was that this team was less and less effective because of the incredible range of PC configurations out in the world. Or maybe it’s just baloney, a story told by Microsoft’s detractors that picked up steam and became part of our understanding of the world.

Here’s what I do know about it: Not much. I’ve heard the story. I feel that Windows quality has been on a steady downward spiral since the story originated, so it’s easy to believe it could be true. But I don’t know if it is true. I’m not sure that anyone does outside of Microsoft’s senior leadership.

But I can Google with the best of them, and here’s what I found.

In 2014, Bloomberg reported that one of Satya Nadella’s first major decisions as CEO of Microsoft was to “shift the culture” at the software giant, with the downside being that doing so would entail “engineering and organization changes,” which are codewords for layoffs. We know now that what Nadella had in mind was laying off most of the employees that Microsoft acquired with Nokia when he axed the Windows Phone business. But citing sources, the publication speculated at the time that, “with new cloud methods of building software,” Nadella felt that developers could test and fix bugs instead of relying on separate teams of testers.

This led to reports, like this one, claiming that Microsoft would “cut its QA [quality assurance] department,” which is more than a bit of a leap: if Microsoft ever had a single “QA department,” that was back in the 1980s, as the company grew so big and complex that it required multiple QA teams. But it certainly didn’t have one such team in 2014. And what did this have to do with Windows specifically?

Days later, the ever-reliable Mary Jo Foley filled in the gaps. Citing an internal memo from Windows lead Terry Myerson that doesn’t specifically mention testers, she wrote that he was indeed laying off “a substantial number of testers” as part of an effort to “decrease the number of testers as compared to the number of developers.” But this was part of a broader plan on Myerson’s behalf to undo the siloed organizational structure of his terrible predecessor, Steven Sinofsky. Notably, the goal was to “make the OS team work more like lean startups than a more regimented and plodding one adhering two- to three-year planning, development, testing cycles.” As notably, the goal was not to get rid of all the testers.

So how did we get from a reasonable streamlining to “they were all fired“? Leaving aside disgruntled ex-employees (like this guy) who played no role in corporate strategy and can’t possibly know what really happened, we can’t really say how. But we did. Because that’s how stories like this get started, amplified, and contorted. It’s like the telephone game where the message changes or is diluted in the retelling, and by the time we hear it, it no longer resembles the original. Not helping matters, Microsoft started the Windows Insider Program in October 2014, just months later, giving millions of non-employees out in the world a chance to best test Windows 10. Coincidence?

Well, yes, probably. One might even argue that Microsoft saw the need to maintain a high level of testing in a changing world and that bolstering its internal efforts with telemetry from more real-world PCs could only improve the quality (not to mention reengage the community that had been cast aside by the Sinofsky regime, which only trusted its own data, didn’t listen to feedback, and believed that only it had the answers.) Whether that was the result is debatable, but it’s also a story for a different day.

For this day, let’s turn our attention to Microsoft and the sudden and unexpected unshackling of its OpenAI-licensed AI capabilities, starting with the Bing chatbot this past February. My reaction to this reads like the 5 stages of grief, assuming you can cycle through the stages non-concurrently with multiple stops on each. But let’s chalk that up to the fog of war. After all, no one saw this coming. I mean, just look how badly Google handled it.

Instead, what I’d like to focus on are Microsoft’s own words and the words of more credible experts in this field. Here’s the best example.

In a New York Times op-ed published just two weeks after Microsoft’s announcement, Reid Blackman, a government advisor on digital ethics, noted that the software giant had behaved recklessly in exposing over one million people across 169 countries to its AI.

“You don’t have to take my word for it,” he wrote. “Take Microsoft’s. Microsoft articulated principles committing the company to designing A.I. that is fair, reliable, safe, and secure. It had pledged to be transparent in how it develops its A.I. and to be held accountable for the impacts of what it builds. In 2018, Microsoft recommended that developers assess ‘whether the bot’s intended purpose can be performed responsibly’ … But the prompt, wide-ranging and disastrous findings by these Bing testers show, at a minimum, that Microsoft cannot control its invention. The company doesn’t seem to know what it’s dealing with, which is a violation of the company’s commitment to creating ‘reliable and safe’ A.I.”

The charge here is that Microsoft was ahead of the game in preaching ethical AI before it had a product to market, but the second that it saw a way forward, it immediately violated its own principles. “Microsoft’s ‘responsible A.I.’ program started in 2017 with six principles by which it pledged to conduct business,” Blackman continued. “Suddenly, it is on the precipice of violating all but one of those principles. (Though the company says it is still adhering to all six of them.)”

And this is where Microsoft and AI intersect with the QA tester controversy from almost a decade ago: in March, just one month after the Bing chatbot introduction and the resulting controversy when it went off the rails multiple times, Microsoft laid off its “ethical AI team.” The assumption was that the software giant was unhappy with internal criticisms of its strategy and was simply silencing a critic, a move that was, if true, ironically unethical.

But is it true? Not exactly.

“Microsoft laid off its entire ethics and society team within the artificial intelligence organization as part of recent layoffs that affected 10,000 employees across the company, [we have] learned,” The Verge reported at the time. “The move leaves Microsoft without a dedicated team to ensure its AI principles are closely tied to product design at a time when the company is leading the charge to make AI tools available to the mainstream, current and former employees said.”

That sounds pretty damning, even though the article goes on to explain that Microsoft still maintained “an active Office of Responsible AI, which is tasked with creating rules and principles to govern the company’s AI initiatives.”

I wonder how many readers read that far? Not many, it seems, given that Microsoft felt the need to issue a statement.

“Microsoft is committed to developing AI products and experiences safely and responsibly, and does so by investing in people, processes, and partnerships that prioritize this,” the statement reads. “Over the past six years we have increased the number of people across our product teams and within the Office of Responsible AI who, along with all of us at Microsoft, are accountable for ensuring we put our AI principles into practice. We appreciate the trailblazing work the Ethics & Society did to help us on our ongoing responsible AI journey.”

And that last bit is interesting: what is “Ethics & Society”? The Verge article suggests that anyone related to AI and ethics at Microsoft—e.g. “that team”—was laid off. But that team was being slowly cut down over a three-year period as Microsoft reorganized around AI. Then, in late 2020, employees were tasked with monetizing the software giant’s AI capabilities swiftly—a sentiment CMO Chris Capossela repeated internally recently only to be demonized for it—as that team prepared to be moved into other parts of the company.

What other parts? MSN? Xbox? Or something a little more relevant?

Microsoft has been slow to comment on this publicly. In April, it talked up responsible AI without addressing its internal machinations. And then in May, it cited progress in this initiative and finally addressed the fate of the Ethics & Society team.

“We have made changes over time as our needs have evolved,” Microsoft Chief Responsible AI Officer  (yes, really) Natasha Crampton wrote. “One change that drew considerable attention recently involved our former Ethics & Society team, whose early work was important to enabling us to get where we are today. Last year, we made two key changes to our responsible AI ecosystem: first, we made critical new investments in the team responsible for our Azure OpenAI Service, which includes cutting-edge technology like GPT-4; and second, we infused some of our user research and design teams with specialist expertise by moving former Ethics & Society team members into those teams. Following those changes, we made the hard decision to wind down the remainder of the Ethics & Society team, which affected seven people. No decision affecting our colleagues is easy, but it was one guided by our experience of the most effective organizational structures to ensure our responsible AI practices are adopted across the company.”

This reads very much like Microsoft’s 2014 rationale for removing the silos from Microsoft development, doesn’t it? Rather than have one overreaching AI panel looking over everyone’s shoulders, each team can have its own ethical AI members that ensure the work that team is doing adheres to the company’s broader responsible AI requirements. This seems like a reasonable way to infuse the company with not just AI, but also responsible AI.

According to The Verge article that generated that attention, the Ethics & Society team had roughly 30 employees at its peak in 2020, so we can assume that at least 23 people—“engineers, designers, and philosophers”—ended up elsewhere at Microsoft. The fate of those 7 people is unclear, but reading into Microsoft’s comment, I think we can assume they were let go.

7 people.

That sounds insignificant. But is it? A small number of people can make a big difference depending on their roles. For example, the U.S. Supreme Court consists of just 9 justices. Closer to home, the UK CMA regulatory body that somehow managed to block Microsoft’s acquisition of Activision Blizzard consists of just 11 board members, none of whom were elected by UK citizens. And even closer to home, there are just 7 people among Microsoft’s executive officers, the company’s top players.

Further muddying the waters, Microsoft isn’t alone in casting out internal AI complainers: Google fired ethical AI researcher Timnit Gebru in 2020, and Geoffrey Hinton, described as the “Godfather of AI,” left the online giant and quit Google so he could speak freely about the dangers of AI as the online giant ramped up its competitive efforts against Microsoft. Everyone is doing it!

Here again, it’s hard to know the full story, but you can see the two sides lining up on ideological grounds and casting shade on the other. From Microsoft’s perspective, it is aggressively competing in an important new battleground, but it is doing so ethically. From its critics’ perspective, the software giant is behaving irresponsibly by releasing AI before it’s ready while being fully aware of the risks.

I’ve been critical of Microsoft’s actions, but I see both sides of it. The post-Gates Microsoft got slow and soft, and its AI push feels like something from the 1990s, when the company ran roughshod over the personal computing industry. Today, Microsoft is much bigger than it was then, but it’s also much less powerful in its own industry, and it takes as many lumps as it gives out. Maybe more: arch-nemesis Apple, once an also-ran, is now the largest company on earth and just one of its products, the iPhone, earns more profits and revenues than all of Microsoft. And the rest of Big Tech—Amazon, Google, and Meta—are just more reminders of markets Microsoft could have dominated.

It’s clear that the software giant sees AI as its moonshot, a chance to leapfrog its competition and perhaps even return to its previous perch atop the industry it once dominated. We can disagree with its methods—and I find myself doing just that more often than I’d like these days—but it’s hard to criticize the ambition. This company was sleepwalking for years, but it’s awake now. You bet it is. And while the infamous Gates (re)quote from years past—“they had awakened a sleeping giant”—is perhaps a bit fanciful given that all of Big Tech will be swimming in AI for years to come, you never know. The future can change again. And Microsoft is betting on it.