In a report published today and that will have interesting consequences for the software industry, the US Commerce Department agency advises in favour of monitoring but not restricting the availability of Large Open Source Models. This may represent a blow to Google, Microsoft and Open AI all of which have argued that closed, proprietary models provide the best security and protection against bad actors. Presumably “bad actors” for them doesn’t include predatory monopolies using their size and unique position in the market and technology to set terms for everyone else.
This is definitely is good news to start ups and companies like IBM and Meta that advocate for Open Models in the name of transparency, competition, and ultimately for innovation.
The poles of this “debate” may feel familiar to people who were involved in software development in the early days of the Web especially. Back then, when (actual). Mastodons and Giant Sloths roamed the Great Plains, Microsoft and its litigious, patent trolling, proprietary source allies fought with all of their prodigious available resources to keep Open Source Software from widespread adoption. FUDD being its weapon of choice. Today GNU/Linux and a host of tools are available freely (as in Free Speech, not Free Beer) and have proven to be as secure, reliable, and safe if not more than their proprietary counterparts. Indeed the Open Source stack of Linux, Apache, MySQL,1 and PERL or PHP were practically the default Web Stack for over a decade. Microsoft even contributes to and sponsors software projects and tools that fall within the broadly defined category of Free and Open Source Software.
It seems that some instincts are bred in the bone out in Redmond. Microsoft’s and Google’s framing of closed source providing greater security protections sure sounds like we’ve heard them before. Trusting technical giants with documented histories of opaque, and often anti competitive practices, shouldn’t match anyone’s idea of greater security. Indeed after the Crowdstrike outage that paralyzed operations for companies that “Trusted” Microsoft around the world the claims that opaqueness and monopolistic practices equate to better security (or stability, reliability etc) seems especially dubious. If past corporate behavior is anything to go by, OpenAI/Microsoft abd Google would likely all love to turn their fortresses protecting customers from malicious AI into prisons of vendor lock-in, opaque data and privacy protections and a “too big to regulate” posture over all.
You can read the full report from the NTIA here:
https://www.ntia.gov/other-publication/2024/fact-sheet-ntia-ai-report-calls-monitoring-not-mandating-restrictions-open-ai-models
- Now as often as not M is for MariaDB the fork of MySQL created in response to Oracle’s purchase of Sun MIcrosystems the previous Owner/Sponsor of MySQL ↩︎
