This is a response to Meta’s announcement “Open Source AI is the Path Forward”
I believe the strongest innovation comes by maximizing both competition AND cooperation, and enabling this dual-mode coordination is vital for truly efficient markets.
This is why:
Competition is the ideal mechanism to discover and accelerate excellence... when the desired outcome is clear but the best path isn't. The less we know about the best approach, the more benefit there is in testing many options. This is true in biology, in venture capital, in sport, and even in combat.
Competition produces optimization we refer to at Sodal as "differentiated value": the thing you do well that separates you from the pack. The strongest teams are adept at maximizing their differentiated value. By contrast, "not invented here" syndrome is a form of competitive failure that results from an organization's lack of clarity about their differentiated value.
As innovation progresses, however, it may become clear that everyone - even competitors in the same arena - are better off when they employ a proven approach. We describe solutions that benefit everyone as providing "universal value". Universal value takes many forms. Things generally work better when everyone can read and write (education), it is easy to transport goods (roads), and great tech foundations are accessible (open source software).
Universal value is the cooperative compliment to competition's clear goals and ambiguous approaches. With universal value, we know the best way to get something done, but we don't know all the cool ways people might use the capability. Every new form of universal value creates a foundation for new differentiated value - for new and more exciting competition.
The classic tech example is *nix; the elegance that Thompson and Ritchie baked into the Unix OS didn't just solve Bell's needs managing the PDP-11 - it was a template to simplify and scale OS management generally. It took decades for this value to become universal, however, with the ascendance of Linux and its open source licensing model.
Indeed, the *nix story has a cautionary component. Enabling universal value requires cooperation, and systems that optimized exclusively for competition can stifle this critical form of efficient value delivery - often at their peril.
When everyone has to replicate the same work to achieve a competitive baseline, we call this "undifferentiated value": all the stuff you have to do to compete that DOESN'T set you apart. Undifferentiated value is bad for everyone. Teams can't focus on maximizing their differentiated value, and duplicative effort reduces total value across the ecosystem
Meta is being smart, not altruistic; making excellent AI technology open is about shifting the battle from who can burn the most dev hours duplicating work to who can launch the most impactful, differentiated value on top of these remarkable, universal capabilities.