Context: The global regulatory response to the AI revolution ranges from a more opportunity-centric approach (October 27, 2024 ai fray article) to legislative proposals to an aggressive and expansive application of existing rules (October 24, 2024 ai fray article).
What’s new: On Friday, the name partners of the Andreesen Horowitz (a16z) venture firm (by some measure the number one in the world) and the CEO and the president of Microsoft issued a joint statement on “AI for Startups” (link). In the face of “the most consequential innovation we have seen in a generation” and recognizing that they don’t always agree, the four business leaders discuss policy objectives with respect to Generative AI that come down to avoiding overregulation, promoting open source, giving AI systems access to training material, investing in and educating the workforce on AI.
Direct impact & wider ramifications: This manifesto is food for thought for policy makers and regulators as it advocates a healthy symbiosis between Big Tech and Little Tech. In particular. it explains how to enable society and the wider economy to reap the benefits of AI, and what really leads to more competition: collaboration between startups and large players. Regulators and policy makers like to make startup-centric arguments when seeking to justify AI-related measures targeting Big Tech, and the a16z-Microsoft statement explains where Big Tech and Little Tech actually have mutual interests.
The current wave of information and communications technology startups is by far and away the most heavily regulated one ever. The first dotcom wave in the second half of the 1990s happened very fast, and many internet-specific pieces of legislation were enacted only in the 2000s. Today’s startups have to deal with a thicket of digital regulations and the ever more complex and confusing rules imposed on them by the mobile gatekeepers Apple and Google, which are self-appointed and self-serving de facto regulators. Now there are AI-specific legislative initiatives around the globe, and it looks like politicians have an insatiable appetite for coming up with more and more AI laws.
One of the very first internet startups was Netscape, founded by Marc Andreessen, one of the four signatories. He built the first wildly popular web browser. Microsoft built the Internet Explorer and made it (as well as its web server) available for free, which was an option for the platform maker but not for Netscape. That practice has often been referred to as “netscaping,” though by now “sherlocking” (named after an app that Apple rendered superfluous) is comparable term that one hears far more often. Whatever Microsoft’s contribution to Netscape’s demise may have been back in the day, it is water under the bridge and now Microsoft and Netscape’s founder (as well as the other founder of his venture firm) agree on several of the parameters of an environment that enables AI startups to thrive.
If the common ground between Microsoft and Andreessen Horowitz was limited to the simple business interest that Big Tech has in acquiring startups and that venture investors have in M&A as an exit route, it wouldn’t mean much. But the statement is all about growth and competition, not consolidation.
Overregulation can adversely affect AI startups in different ways. The statement mentions “unnecessary bureaucratic burdens to startups” that result from regulation that goes beyond measures against misuse of AI by bad actors. It is also about “infrastructure investments [that] are essential to creating opportunities for new businesses to experiment and grow in the AI economy.” And the statement talks about developers’ “freedom to choose how to distribute and sell their AI models, tools, and applications for deployment to customers.”
In ai fray‘s observation, opportunities for startups are diminished when competition authorities treat every AI partnership as a “merger” (September 27, 2024 ai fray article). Regulators have greater powers over mergers than other business decisions. But merger laws were not intended to be applied to commercial partnerships where no one acquires control of another company.
The joint statement with a16z contains Microsoft’s clearest pro-open-source statement in the AI context do date:
“Regulators and decision-makers should embrace a regulatory framework that protects open source and secures the ability of entrepreneurs, startups, and companies to create, build, transform and win the future.”
The statement also calls for open data (governments making certain data available for AI training) and for knowledge to remain available to AI systems despite copyright law protecting creative expression. Realistically, it is too early to predict how the dozens of AI copyright lawsuits already pending in the U.S. (with more to come) are going to be decided. Some of those claims appear overreaching, such as the New York Times’s damages claim in the billions of dollars. But the fact-expression dichotomy and, even more so, the multifactorial fair use analysis are parts of copyright law where various surprises have happened over the years. The judicial trend over the last ten or more years has favored a rather broad understanding of what constitutes transformative use of copyrighted material. There is no guarantee, however, that the trend continues. The only thing certain is that the losing side (whether it will be copyright owners or AI providers) will lobby for legislation favoring its interests.
The a16z-Microsoft statement will serve as an instructive point of reference when overregulation is described as a means of fostering startups.