A problem shared is a problem halved, or so the saying goes. That is the philosophy of the open-source software movement, which sees volunteers coming together to code, patch and improve software tools.
The approach has been a huge success: most computing devices, including closed commercial machines like Windows PCs or Apple iPhones, use some open-source software. But while openness has benefits – having many eyes on code makes it easier to find and fix bugs – there are risks, too.
Open-source software intended to pilot consumer drones is now being used to guide a cheap Ukrainian cruise missile. Traditionally, military tech is custom-built, restricted and expensive. Ukraine has shown this needn’t be the case in the war with Russia: as 91av has reported, cheap consumer drone technology is helping Ukraine hold back the invasion.
Advertisement
But while many readers may cheer on Ukraine’s fight, the fact that anyone can now download a cruise-missile guidance system should give us a moment’s pause.
Open-source models have fewer guardrails and the ones that exist are easier to remove
Open source is also key to another war, though this time a commercial one. In the US, AIs like ChatGPT use mostly private codebases, with access sold for a monthly fee. In China, however, open-source models lead the way. These are generally a few months behind the US models in capability – but anyone with the necessary hardware can download and run them for free.
While the closed US models generally come with guardrails to prevent chatbots from spewing out harmful material, open-source models have fewer such limits and the ones that exist are easier to remove.
That creates headaches for attempts to regulate AI models, such as testing them for national security risks before release, as was recently suggested by Google DeepMind CEO Demis Hassabis. But it also offers the opportunity for more people to find ways of using AI models that US tech firms may not have thought of.
Open source means problems can be shared – it’s just that some will get easier to solve, while others get much harder.
