Nick Wagner

This Rumor About GPT-5 Changes Everything Let’s start the year on an exciting note

It may be wild speculation, but this explanation of why GPT-5 hasn't been released makes sense to me. open.substack.com/pub/thealgor...

www.cdc.gov/nwss/rv/wwd-...

If you thought there was a reasonable chance an H5 pandemic could occur, how would you prepare? I'm looking at the CDC's wastewater data reporting and growing concerned.

There's a new $50k prize being offered for outstanding contributions in the use of AI for the natural sciences or engineering. Share with the PIs you cherish or want to be nominators! datascience.uchicago.edu/research/ai-...

Are these technical solutions inherently dual use for offense as well as defense? My understanding is Anduril uses Lattice to glue together all their systems. Not just the ones involved with defending stationary positions. You would probably need a different dataset for fine-tuning on targets though

I should say 11B parameters for the smallest models I know. For high accuracy requirements (like minimizing false negatives on drone detections), you probably need even bigger models.

The multimodal models to my knowledge are still quite memory-intensive (~11B parameters that are not easily quantizable). While you can defend a base with a fast internet connection or on-prem cluster, a remote base or mobile platform might struggle to run these models.

Ok, my guess for why on a technical level Anduril wants OpenAI's help is that they have a vision problem for which existing convolutional neural networks are too brittle or expensive to train. Probably drones are too varied in design or imagery conditions for an existing approach.

One thing I've been missing is a work office filled with people who are also obsessively online. I want my DeepSeek gossip!

I have to say, it is nice being able to see all the comments and links from the machine learning and economics people again. I had almost forgotten what Twitter took from me.