Ben Rogers

Disrupting the first reported AI-orchestrated cyber espionage campaign A report describing an a highly sophisticated AI-led cyberattack

In 2023, we'd often use the phrase "It doesn't matter if we use AI, because everyone else will" when discussing LLMs internally, and this epitomizes that sentiment.

This feels like a major milestone in agentic AI. Thanks to Anthropic for the transparency.

www.anthropic.com/news/disrupt...

Nice roundup on CRISPR genome editing @economist.com
www.economist.com/leaders/2025...

Seen the dialogue focused on AI costs with the DeepSeek $5.6 mil and comparing that to the total investment of US companies on developing AI, and it's not an apples to apples comparison given salaries, CapEx, data procurement, and other costs are excluded

Still quite impressive tho

The HHS AI Strategic Plan is now available on the ASTP website.

Very grateful to the over 200 contributors across all HHS divisions who provided inputs to the plan!

www.healthit.gov/topic/hhs-ai...

Happy 45^2 (2025) everyone! 🥰🤩

The GPT-4 barrier was comprehensively broken Some of those GPT-4 models run on my laptop LLM prices crashed, thanks to competition and increased efficiency Multimodal vision is common, audio and video are starting to emerge Voice and live camera mode are science fiction come to life Prompt driven app generation is a commodity already Universal access to the best models lasted for just a few short months “Agents” still haven’t really happened yet Evals really matter Apple Intelligence is bad, Apple’s MLX library is excellent The rise of inference-scaling “reasoning” models Was the best currently available LLM trained in China for less than $6m? The environmental impact got better The environmental impact got much, much worse The year of slop Synthetic training data works great LLMs somehow got even harder to use Knowledge is incredibly unevenly distributed LLMs need better criticism Everything tagged “llms” on my blog in 2024

Here's my end-of-year review of things we learned out about LLMs in 2024 - we learned a LOT of things simonwillison.net/2024/Dec/31/...

Table of contents:

Intestinal mucosal barrier repair and immune regulation with an AI-developed gut-restricted PHD inhibitor - Nature Biotechnology Generative artificial intelligence is used to design an effective treatment for inflammatory bowel disease in preclinical models.

www.nature.com/articles/s41...

From initiation to preclinical candidate nomination, the program spanned only 12 months, during which approximately 115 drug candidates were developed and screened.

I've been saying 1 year in AI is 7 years in the real world, but after the past two weeks I need to revise that statement. It feels like each month is now a year in any non-AI space.

2024 HHS AI Use Case Inventory is now live

271 in total, compared with 163 in 2023

www.healthit.gov/hhs-ai-useca...

Now that we have full o1 from OpenAI it's a good time to reflect on the winding process that got us here. All of Q*, Strawberry, and other smaller "leaks" were about this behavior. There was A LOT of wrong information the public discourse. I count my writing even as being wrong in moderate ways. 🧵

My Bluesky feed right now is AI research, Cats, and Space photos and it's such a joy. Good job by my algorithm

AI’s computing gap: academics lack access to powerful chips needed for research Survey highlights disparity between academic and industry scientists’ access to computing power needed to train machine-learning models.

🧪 A global survey of academics reveals limited access to high-end AI tools like Nvidia H100 GPUs (10% access). With 66% dissatisfied, the gap with industry slows AI research and sparks innovation in resource efficiency. 🩺🖥️ #MLSky

In principle I’m in favor of teaching “AI literacy,” but watching the seven contradictory kinds of received wisdom about AI circulating among college profs rn makes me worry a little about the uninformed leading the uninformed. Model capacity is changing more rapidly than ppl update their opinions.