JohnPDickerson
Joined 5 years ago
Joined 5 years ago
Mozilla is nowhere near large enough to solve this problem on its own, but: (i) Mozilla AI, which I'm leading, is pushing to solve this at the infra and maybe platform layer by providing developers the ability to choose to use their own AI tooling and (ii) Mozilla writ large is a big enough machine that it can create a community and alliance across dozens or hundreds of players to create a viable alternative to a bleaker, non-open Internet. That's my hope!
We do run into this branding question frequently, and will add some clarity to the website.
Chiming in here as a Mozillian focused on AI not specifically related to Firefox - I agree! Just a heads up that a separate public benefit corporation, Mozilla.ai, exists and is supporting a suite of commercially-licensed, open source, general AI dev and enablement tools. That includes mcpd, what we're calling "requirements.txt for MCP", meant to enable more trusted automated interfacing between machines.
A goal here is to support developers looking to build out AI-enabled systems that interact with each other and with the Internet, be that through a traditional browser or some other way.
You may enjoy some of our projects: https://www.mozilla.ai/open-tools/choice-first-stack
We're very small compared to the Mozilla mothership, but moving quickly to support open source AI in any way we can.
The posted numbers for my home institution, the University of Maryland, are a bit off; current offers are $25k for 9-month and up to $36k for 12-month, if funded on an RA over the summer. That RA summer funding is not guaranteed. Again, clearly, this needs to go up.
Similar pitch -- use a small adversarial perturbation to trick a classifier -- but LowKey is targeted at industry-grade black-box facial recognition systems, and also takes into account the "human perceptibility" of the perturbation used. Manages to fool both Amazon Rekognition and the Azure face recognition systems almost always.