Human/AI interaction. Visualization as design, science, art. Professor at Harvard, and part-time at Google's People+AI Research initiative.
Joined July 2008
- Tweets733
- Following484
- Followers17.1K
- Likes2K
Visualize transformer attention!
AttentionViz, created by Catherine Yeh and expanded by Yida Chen, helps you explore transformer self-attention by visualizing query and key vectors in a joint embedding.
Paper: arxiv.org/abs/2305.03210
Website: attentionviz.com/
Language models have some beautiful spiral plots reflecting positional patterns. And a vision model has heads that arrange images according to brightness and hues. But there’s a lot more to find! What else can you see?
Here's a close-up of hue and brightness heads in a vision transformer. Work done with tremendous help from Aoyu Wu and @chenxcynthia!
Toasters have blinking lights, cars have speedometers. Should chatbots have dashboards too?
A speculative essay: The System Model and the User Model: Exploring AI Dashboard Design
arxiv.org/abs/2305.02469
Martin Wattenberg retweeted
I want to show the NSF there would be broad support+utility for a "National Deep Inference" service for >100b LLMs.
If your research would be enabled by an inference service on open LLMs w API access+overrides to internal activations, params, gradients:
Please Like this thread!
Martin Wattenberg retweeted
Students explore the aesthetics of computing in new computer science course at SEAS. "CS73: Code, Data, and Art" is co-taught by @wattenberg and @viegasf, and teaches students how to create abstract art and communicate data sets through visualizations. buff.ly/3GXpkTE
Thinking about grad school next year? Interested in visualization, machine learning interpretability, or human/AI interaction? Consider Harvard. @viegasf and I are continuing to build our lab!
Martin Wattenberg retweeted
The labyrinthine patterns traced by birds on the wing – in pictures theguardian.com/artanddesign…
I'm teaching with @OpenProcessing for the first time, and am completely impressed with how polished and friendly the system is. Every detail is on point. Last class a student spontaneously said, "OpenProcessing is just so great." I agree!
Tiny neural networks have a surprisingly rich inner life, and may hold clues to how their larger cousins work. This image: evolution of feature vectors during learning. Full story: transformer-circuits.pub/202… (in collaboration with the great interpretability team at @AnthopicAI)
The project is a systematic study of how a neural network represents more features than it has neurons, under a variety of conditions. (Oh, and I can't tweet without a typo, apparently. That's @AnthropicAI of course!)
Martin Wattenberg retweeted
SMASHOMANCY is a "divination system for cracked phone screens" and @djbaskin retains pole position as one of my favorite living creatives.
We've had this dream for a while and now it's becoming real! @aleffert and I made a divination system for cracked phone screens called ⚡️SMASHOMANCY⚡️
We're doing screen readings at a repair shop TODAY from 4:30-6:30 PM at 1099 Irving St. in SF! Drop your phone & join us.
What are good chart colors to represent past, present, and future? Not looking to "gray out" a prediction, but just represent time.
All excellent images, but the "cyclops" picture is just startlingly good. Kudos for a brilliant prompt.
You’re unable to view this Post because this account owner limits who can view their Posts. Learn more
The new baby name data is out! Check out @NamerologyTalk to find out the latest trends. What alcoholic drink appeared as a name for the first time? Which famous robot is erasing a name? Which steamy series had the biggest effect?
Martin Wattenberg retweeted
The year's biggest baby name style influencer: Raya and the Last Dragon. In a landmark for Asian characters, Raya was America's #1 fastest-rising popular name, and Namaari the #1 rising uncommon name
#AAPHM #AsianAmericanPacificIslanderHeritageMonth
