China Morning Missive
With the announcement of the “Genisis Mission”, the Trump administration has made dominating the AI ecosystem a national priority. While there have certainly been a host of differing opinions on possible ulterior motives, from a geopolitical standpoint the decision makes perfect sense. So too, however, could this move end up directly impacting – inadvertently – a diaspora of Silicon Valley companies.
Over the past month there’s been increased media coverage of how American startups are fully leveraging the Chinese open-sourced AI models. After all, that is sort of the point of having access to open-source software. Take what’s already available and then iterate and improve upon that original source code.
The issue is the point of origination of that that source code; China.
Any time a competing Chinese software solution, or hardware offering for that matter, gains ground on all available American alternatives, Washington policy wonks step in on the grounds of “national security”. Not to say that such concerns aren’t warranted, rather that the issue here is the wide spread in usage (and, now, dependency) on Chinese AI models.
Rhetorically speaking, is there anyone who thinks the American government will permit home grown platforms build on Chinese source code?
Much as the NBC article highlights, “while models from American companies continue to set the pace of progress at the frontier of AI capabilities, many Chinese systems are cheaper to access, more customizable and have become sufficiently capable for many uses over the past year.”
It is the use of the word “customizable” which should stand out to all readers. It goes to the primary – I would say existential – issue facing the entire American AI theater. The dogmatic pursuit of closed source architecture, the building of silos (competitive moats), remains the greatest threat to America’s position.
Basically, you have five large dogs all chasing after a single bone. The likelihood is that the dogs will attack each other viciously long before catching the bone. You see this in the perversion of all incentives, mainly in the rapid acceleration of CAPX spending.
What this then leaves is the Chinese AI scalers to just continue moving forward building better software and directing all efforts to addressing the issue of compute.
https://www.nbcnews.com/news/amp/rcna242430