I don't often weigh in on AI issues, but had some thoughts based on an interesting recent post by @ilyaraz.bsky.social on LinkedIn. Ilya posted recent comments by two top algorithms researchers noting how well LLMs had done on their algorithms exams. 1/

New blog post, which covers: * My injury that took me out for more than a month * Recovery and the role strength training played in it * Bike training and my history with cycling * Upcoming fun adventures! Enjoy! purposefulrunning.org/big-adventur...

While playing with Llama on the (publicly available) IMDB review dataset, I noticed the "Safe LLM Streisand effect": the reviews the model refused to classify contain some truly disturbing shit, but now they stand out from the dataset of 50000 reviews. How is that for "safety"?

Sumary of 2024: tough but still grateful. Happy holidays!

I am working on a side project that is meaning to make reading news and staying informed easier. If there is interest, I might do some “building in public” as the cool kids say these days. Pls like or share to indicate the said interest (eg if you want to see me wrestling with web development).

This is peak LinkedIn

Hello, world! I work in tech, and love running, hiking and cycling unreasonably long distances. Sporadically, I blog about these things: purposefulrunning.org . On tap for the next year: run 37 miles in Joshua Tree Park, bike 200 miles from Seattle to Portland and hike up Mount Whitney.