Andy Hickl

Time to try something new.

Welcome to the Bluesky account for Stand Up for Science 2025! Keep an eye on this space for updates, event information, and ways to get involved. We can't wait to see everyone #standupforscience2025 on March 7th, both in DC and locations nationwide! #scienceforall #sciencenotsilence

Seeing a lot of hype about AI models trained for “under $50.” Let’s be real—these models are distilled from much larger models trained with massive resources. Distillation is useful, but calling it “cheap training” seems misleading. What am I missing? www.theverge.com/news/607341/...

Aptly put: Tom Frieden, a former CDC director, compared the halting of outbreak updates to “finding out that your local fire department has been told not to sound any fire alarms”. www.ft.com/content/6067...

Merry Christmas (to those who celebrate) Photo credit: Grok

Inside me are two wolves: one wants to make cinnamon rolls from scratch and the other doesn't want to go to the grocery store Okay they're not wolves more like sloths

The Allen Institute and the University of Washington have released a new AI search tool to answer scientific queries of OpenAccess content Preprint released in @arxiv-cs-cl.bsky.social OpenScholar: Synthesizing Scientific Literature with Retrieval-augmented LMs arxiv.org/abs/2411.14199

Meet ACE - a new tool for comparing annotations across Allen Institute transcriptomic taxonomies! Join us for a webinar on how to use it in your #neuroscience research, incl. studies using our Alzheimer's datasets. 📆 Dec. 18, 9:30am PT 💻 Webinar 🎟️ alleninstitute.org/events/cell_... #neuroskyence

alleninstitute.org/careers/inte... Internships and Postbac opportunities open at the Allen Institute, applications close January 7th 2025. Great opportunity for anyone interested in life sciences research or science-support careers. Spread the word!

Cool Yule read: this article highlights how behavioral diversity in artificial intelligence “teams” can significantly enhance problem-solving & adaptability, something that could change how we approach challenges in (AI) research. 🧪 arxiv.org/abs/2412.16244

Whoa: 9 sea snake species have regained the genetic requirements for advanced color vision, showing that complex traits lost to evolution can be regained. phys.org/news/2024-12...

I’m so stealing this idea. Such a cool way to ensure #deib is built into a team’s mindset. Thanks for sharing, @sarahstern.bsky.social

The Washington Post nails it: AI is transforming healthcare, from diagnostics to personalized treatments. 2025 will be the year of AI in life sciences—thoughtful integration is critical. 🧪 Read more: www.washingtonpost.com/opinions/202...

🧪Abhishaike Mahajan examines Recursion’s shift from cell painting to brightfield microscopy. The piece highlights how #ml has unlocked brightfield imaging as a label-free alternative, addressing challenges of cell painting. A must-read: www.owlposting.com/p/why-recurs...

Predicting the results of experiments is challenging, especially in fields like neuroscience. This recent study shows that fine-tuned LLMs, like BrainGPT, can predict neuroscience results more accurately than human experts on benchmarks. 🧪 https://www.nature.com/articles/s41562-024-02046-9.pdf

@jshendure.bsky.social and team continually blow me away

Moravec’s Paradox is such a fascinating concept: today’s AI can tackle complex tasks like reasoning or drug discovery—things that push human limits. But it still struggles with what humans find simple, like picking up a coffee cup. Biology & AI remind us how differently intelligence can evolve. 🧪

This is a great fellowship for any postdoc who wants to work with #neuro data in a nurturing environment at @alleninstitute.bsky.social. Deadline is approaching!

This is the best take on the potential of predictive biology I’ve come across 🧪 open.substack.com/pub/creode/p...

Tom suggests replacing the logic of prereg -> confirmation with the logic of train/validation/test splits. The place this approach is really, really needed is in historical data — where we can never get another equivalent sample! Our ordinary understanding of replication is broken there.

Framing AI as a threat to human agency oversimplifies its role and potential. We need to focus on how AI can complement human decision-making — especially in fields like life sciences, where it’s accelerating discoveries 🧪 www.journals.uchicago.edu/doi/10.1086/...