Software, Data, Search, IR, ML & Distributed Systems. Open Source Geek, @theasf member. Author @tamingtext, Director AI @BoozAllen, opinions are my own.
MD / DC
Joined January 2009
- Tweets2K
- Following2K
- Followers1.1K
- Likes2.6K
We’re advancing how AI strengthens cyber defenses. Booz Allen is a participant in Anthropic’s Project Glasswing and is testing Claude Mythos Preview for defensive cybersecurity work on our own software and systems. Learn more: boozallen.co/4v1NmD2
More than 40% of agentic AI projects will be cancelled by 2027—not because the tech isn’t ready, but because they aren’t built to operate in real environments. At #BoozAllen, we build agentic AI for mission conditions—so it performs when it matters. boozallen.co/4tZZmUb
Meg builds #AI for missions with lasting importance—working at the edge of what’s possible. Interested in a role like Meg’s? Bring your #AI expertise to Booz Allen and apply: boozallen.co/4cbUiXj
Drew Farris retweeted
🎙️🎉 New pod episode is live!
S2Ep3: "Evaluating Real-World Adversarial ML Attack Risks and Effective Management: Robustness vs. Non-ML Mitigations" with guests, Drew Farris and Edward Raff.
👀 Watch, listen, or read the transcript ➡️bit.ly/47BUomi
#MLSecOps #ProtectAI
Drew Farris retweeted
Thanks so much to @EdwardRaffML and @drewfarris for joining us in the studio for this conversation.
Check out their paper that inspired the talk here: arxiv.org/abs/2306.09951 "You Don't Need Robust Machine Learning to Manage Adversarial Attack Risks."
Drew Farris retweeted
By 𝑅𝑒𝑐𝑎𝑠𝑡𝑖𝑛𝑔 𝑆𝑒𝑙𝑓-𝐴𝑡𝑡𝑒𝑛𝑡𝑖𝑜𝑛 𝑤𝑖𝑡𝘩 𝐻𝑜𝑙𝑜𝑔𝑟𝑎𝑝𝘩𝑖𝑐 𝑅𝑒𝑑𝑢𝑐𝑒𝑑 𝑅𝑒𝑝𝑟𝑒𝑠𝑒𝑛𝑡𝑎𝑡𝑖𝑜𝑛𝑠 you too can classify sequence with 1 𝗺𝗶𝗹𝗹𝗶𝗼𝗻 𝘁𝗼𝗸𝗲𝗻𝘀! New #neurosymbolic awesomeness led by @rea1mma w @BlancheMinerva @oatesbag @icmlconf
Recasting Self-Attention with Holographic Reduced Representations. (arXiv:2305.19534v1 [cs.LG]) ift.tt/CXHt1rG
Seconded! I am proud of the work @BlancheMinerva has done, continues to develop, and the fact @BoozAllen can support and drive value from open research.
Super proud of @BlancheMinerva 's work over the years and recognizing it @BoozAllen . Check it out to learn more about the awesome stuff she does, we support, and all the cool things we do through smart folks like Stella! boozallen.com/menu/media-cen…
Drew Farris retweeted
What is the state-of-the-art for Semantic Search over text? Actual semantic *search*, not some chatbot hallucinating content that I need to go check.
Drew Farris retweeted
(7/7) LEACE wouldn’t be possible without @TheDavidSJ, who proved the theorem that led to this paper. I'd also like to thank our other coauthors @ravfogel @ryandcotterell @EdwardRaffML @BlancheMinerva!
Drew Farris retweeted
Ever wanted to mindwipe an LLM?
Our method, LEAst-squares Concept Erasure (LEACE), provably erases all linearly-encoded information about a concept from neural net activations. It does so surgically, inflicting minimal damage to other concepts. 🧵
arxiv.org/abs/2306.03819
“We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run.” - Amara’s Law en.m.wikipedia.org/wiki/Roy_…
“When there is a smart person in the loop, unreliable advice [from an LLM] is better than no advice, and the advice comes much more explicitly than from carrying out a conventional search on a search engine” - Rodney Brooks rodneybrooks.com/what-will-t…
A nice analysis of how parameters are used in transformer models. Good Sunday morning reading. lesswrong.com/posts/3duR8Crv…
Congratulations on the publication of Inside Deep Learning @EdwardRaffML! It is great to finally hold it - it’s Chonky! @ManningBooks manning.com/books/inside-dee…
Drew Farris retweeted
Come by today for the @iclr_conf #MLEvaluationStandards workshop. Also super excited that @drewfarris and I @BoozAllen received an outstanding paper award! And another for reviewing! A thread on two papers in the workshop: 🧵
The field of ML has seen massive growth and it is becoming apparent it may be in need of self-reflection to ensure that efforts are directed towards real progress. To this end, we are organizing an @iclr_conf workshop on "ML Evaluation Standards". ml-eval.github.io [1/N]
Drew Farris retweeted
Excited to share the @icmlconf 2022 Workshop on Knowledge Retrieval and Language Models
knowledge-retrieval-workshop…
Please consider submitting!
We welcome work across topics including LM grounding, open-domain Q&A, bias in retrieval, analyses of scale, transfer and LM phenomena.
Drew Farris retweeted
List of ICML Workshops.
Please add yours to this thread.
➡️Follow this list for updates.
xitter.cf/i/lists/12611092147784…
Drew Farris retweeted
We release LAION-5B: 5,85B CLIP-filtered image-text-pairs, an intuitive search engine like web interface for exploration & one click subset creation, CLIP ViT L/14 embeddings, NSFW & watermark scores ( + the models used to compute them) , kNN indices, ...
laion.ai/laion-5b-a-new-era-…