Long 🧵: Building (and Testing) Classifiers for the Detection of Artificially Generated Text (Part 1) :

People saying DC isn’t a fair compromise between San Diego and Fredericksburg have clearly never tried driving from DC to Fredericksburg.

Join millions who have switched to Grok.

RT @CYBERWARCON: Announcing (most of) this year's CYBERWARCON speaker lineup! We've got some fantastic talks this year, and more will be an….

I won't link to the full fabricated video here, but if you've seen it, it is obviously manipulated based on the splicing of different clips of Miller and the audio not lining up with the video. But is it actually a "deepfake"? The answer, I think, will lie in the audio.

Beyond the video comparison, if the imagery itself was manipulated with AI in any way, it raises questions about the behavior of the creator: why go to that length only to give it away by splicing it w/ another clip of Miller in which he is wearing a completely different outfit!.

Source of the original video is Miller's Dept. of State Daily Briefing, May 28, 2024. Miller is actually responding to a question on civilian casualties in Gaza (exact clip starts at 1:09:00 here: .

. very slight differences in mouth shape in places, but I suspect this is an artifact of compression. Ultimately, I am making a comparison between a screen recording of the video I took off twitter and a high-def video of the actual press conference the video was taken from.

The audio appears fabricated, though whether it was artificially generated or spliced from various real clips is currently unclear. But comparing just the visuals, I'm not yet convinced the video imagery itself was manipulated with AI. You can see in the video comparison some. .

On Friday, the NYT reported the video, describing it as a "deepfake" but then also stating that "the fabricated video. is drawn from actual footage". So is it a deepfake? The article doesn't really answer the question.

🧵I made a side by side comparison of the recently reported "deepfake" video of State Department spokesman, Matthew Miller implying the U.S. tacitly supports Ukrainian strikes on the Russian city of Belgorod. What do people think?

RT @ForeignAffairs: “Exaggerating the effects of foreign influence campaigns serves only the foreign operatives.” @olgs7, Lee Foster, @RidT….

Excited to announce that Aspect Labs has joined @NIST's Artificial Intelligence Safety Institute Consortium as an inaugural member. Looking forward to working alongside so many prominent organizations and talented researchers focused on AI safety and security!.

Great to see FireEye listed as one of DC’s best large places to work in 2024 by Built In!

RT @alperovitch: Full house for Adjunct Prof. @LeeFosterIntel's intensive Introduction to R and Python Workshops for students interested in….

RT @Mgkarayan: Going to tell my kids this was the real escalator ride that changed America. #cyberwarcon . It was over too fast 😭 https://t….

RT @baykenney: Excited to launch Aspect Labs to secure machine learning models in production.

RT @CYBERWARCON: Welcome back from lunch! @LeeFosterIntel kicks off our lightning round with a presentation on generative AI and informatio….

RT @DavidMainor13: Loved seeing my former boss @LeeFosterIntel talk about threats posed by AI in the IO space. My favorite takeaway was his….

We believe we’re at the inflection point for mass-scale commercial AI deployment and a critical juncture for ensuring models are secured from the outset. The space is changing everyday, so looking forward to seeing where it takes us.

I’ve teamed up with @baykenney to found @aspect_labs (, where we are working to help secure AI deployments against bias, misuse, and adversarial attacks.

RT @baykenney: I was selected to attend the @AmerUnderground @GoogleStartups Black Founders Exchange. It was an awesome cohort of amazing f….