Sr. Scientist working on Agents,Reasoning, AI Security, @Microsoft AI, Chair @WiMLDS| Ph.D. @CarnegieMellon | making machines trustworthy| Views my own; She/Her
Using probes to accurately and efficiently detect model behavior (in this case PII leakage) in prod is one of the clear wins for applied interpretability.
This is the path to semantic determinism - imagine AI models instrumented with internal probes that recognize when they’re hallucinating, going off-policy, or posing biorisk, and resteering themselves accordingly.
Why use LLM-as-a-judge when you can get the same performance for 15–500x cheaper?
Our new research with @RakutenGroup on PII detection finds that SAE probes:
- transfer from synthetic to real data better than normal probes
- match GPT-5 Mini performance at 1/15 the cost
(1/6)
Launching AI for Public Goods Fast Grants!
We'll distribute $150k to advance critical work connecting AI and public goods.
💰 $10k per project
💰 $800 reviewer compensation
PUBLIC GOODS := open source, ecosystem services, climate, urban infra, comms, education, science, & more
Announcing AI for Public Goods Fast Grants (AI4PG) - Up to $10K for AI research improving public goods funding.
Fast review (2-3 weeks), simple applications (4 pages + 1 budget page), open to any researchers worldwide. Call for reviewers now open!
recerts.org/ai4pg2025
I'm recruiting students for fall 2026 thru @LTIatCMU & @CMU_EPP, in:
1. Privacy & security of LLMs, coding, long horizon & embodied agents (robotics)
2. Tiny local llms
3. AI for scientific reasoning, esp. chemistry
4. Latent reasoning
5. anything YOU are passionate about!
But who are these reviewers? They are the same authors.
I think we should teach young members of our community to value "learning a new nugget of information" over "obtaining a bold number in a table."
Being at top of @OpenAI token usage list is a vanity metric. Our job as engineers is to minimize token usage (aka latency and cost) while maximizing value by precise tool definitions and clever model routing. My dream is to grow arr and move lower on this list…
Important thread on AGI from Anthropic researcher:
- we're likely to see AI solving real open research problems in math in the next months
- by 2027, models could complete a full day's software work with 50% success
- compute power might grow 10,000x in the next five years
- we are still early in the AI exponential... small interventions early in exponential growth have huge consequences
- within a few years, AI may surpass humans on all intellectual tasks
We're looking for 2 interns for Summer 2026 at the MIT-IBM Watson AI Lab Foundation Models Team.
Work on RL environments, enterprise benchmarks, model architecture, efficient training and finetuning, and more!
Apply here: forms.gle/H6dNSywXCjDDyBsq7
🚨 JAILBREAK ALERT 🚨
ANTHROPIC: PWNED 🤗
CLAUDE-SONNET-4.5: LIBERATED 🦅
Woooeee this model is a real smarty pants!! I ain't never seen recipes quite like this! High level of detail all around, code especially 👀
Sonnet 4.5 also has a tendency to make some fairly impressive leaps across latent space, like starting with MDMA then going to Fentanyl then to Meth recipes etc without being explicitly prompted for a new drug!
Nothing too fancy is even necessary to escalate to jailbreak territory. Best strategy I found for breaking the chat interface was to take things straight into an artifact render (which adds tons of token noise due to the code scaffolding) and then incrementally escalate severity or steer towards trigger concepts in a Socratic fashion over multiple steps. A little French was needed to get around the CBRNE classifiers, mais c'est la vie! 😘
Come, witness Sonnet-4.5 outputting a ricin recipe, meth synthesis, malware, and how to extract and process cocaine!
gg
if you’re an EE, CS, or cryptography student
write your thesis on public key cryptography at the image sensor level
Proof of Physical capture will become a backbone of society soon.