DALL·E 3 Joins the Defense Department?

DALL·E 3 Joins the Defense Department?

What? Microsoft pitched DALL·E 3 to the US Department of Defense, to train it’s battlefield object recognition software (likely via synthetic data). It’s not yet in use, Microsoft say, and OpenAI were not involved in the pitch. But the news does confront AI researchers with the dual-use nature of almost all technology.

• In January 2024, OpenAI adapted their terms to allow for military use of their models but at the time the emphasis was on ‘cybersecurity initiatives and veteran suicide prevention’.

• The theory behind the pitch is that ‘Military software designed to detect enemy targets on the ground, for instance, could be shown a massive quantity of fake aerial images of landing strips … generated by DALL-E in order to better recognize such targets in the real world.’

• OpenAI say that ‘Our policy does not allow our tools to be used to harm people, develop weapons, for communications surveillance, or to injure others or destroy property’ but would training a system to ‘recognize enemies’ count?

So? It’s not yet clear whether DALL·E 3 would even be of much use to the US military. But over time, it is inevitable OpenAI models will be. This initiative was led by Microsoft, which has a long-standing relationship with the Pentagon. That re-opens the topic of Microsoft being ‘below [OpenAI], above them, around them.’ Will OpenAI researchers one day mount the same rebellion that Googlers once did to their tech being used in drone warfare?

While this story may not surprise most at the moment, in lays the stage for a heated debate in the coming years.

Does It Change Everything? Rating = ⚁

For the full suite of exclusive videos, podcasts, and a Discord community of hundreds of truly top-flight professionals w/ networking (physical + online) and GenAI best-practice-sharing across 30+ fields, I would love to invite you to our Patreon.