Asher retweeted
so i guess this is a thing now
universities running ads to resell students' data
for training llms
💰💰💰
Asher retweeted
beware the tyranny of the microcosm you’re in — school, club, company, industry, neighborhood, city, country — make it a point to zoom out and see the bigger picture without having to abandon your small world all together
Asher retweeted
I love buying secondhand books that take three weeks to ship so by the time they come I’ve completely forgotten and it’s like I’m my own fairy godmother
Asher retweeted
UI design is underrated for AI. Sometimes they go hand in hand with new model capabilities in delightful ways.
Here's a very clever demo: you can use rich text (bold, font style, color) to assign different types of weights to each token, making image generation controllable in a highly intuitive & flexible fashion.
Paper: Expressive Text-to-Image Generation with Rich Text
rich-text-to-image.github.io…
HuggingFace demo: huggingface.co/spaces/songwe…
Authors: Songwei Ge, Taesung Park, Jun-Yan Zhu, Jia-Bin Huang
Asher retweeted
following on from my piece on friday about new interfaces for exploring photo metadata — here's me querying a 75,000 image photo library using hand motion to write geospatial SQL queries in real-time 🏞️🖐️
Asher retweeted
For example, GPT-4's success rate on "is this number prime? think step by step" fell from 97.6% to 2.4% from March to June, while GPT-3.5 improved. Behavior on sensitive inputs also changed. Other tasks changed less, but there are definitely singificant changes in LLM behavior.
Asher retweeted
the internet needs radically simpler ways of building software. otherwise we’ll forever be stuck in the prison of apps that only bigtech and VC hyperfuel can afford to build.
who’s working on this? who else is interested this?
30-50 eng working on top of existing infra/code 😵💫
Asher retweeted
Thrilled to announce Med-PaLM, our medical large language model is published in @Nature, today.
Extremely excited for the possibilities this unlocks!
rdcu.be/dgGfJ
Asher retweeted
5+ years ago I built a prototype for a search engine called Alexandria. It would translate your search query in all major languages & re-rank the results. Maybe there’s a better answer to “why is the sky blue”, it’s just in German?
Translation wasn’t fully there yet, but I feel like it is now. Someone should re-explore this idea – at San Francisco, CA
Asher retweeted
I do as well and am also cautious. I've hated both the pro-OOP and the anti-OOP trends, as no tool is just universally the right or wrong solution to every problem in the vastness of computing. It's simply a tool. A hammer can either hurt or help you depending on how you use it.
Asher retweeted
do you have any hobbies?
yeah making computers out of things that shouldn't be computers. watch me be the first to bring turing completeness to figma
(edit going to build this tonight so scroll for my live tweeting of a computer)
#Config2023 Launch 3: Advanced prototyping
We’ve added new features to help you create more realistic prototypes with fewer steps/frames/connections.
That’s more iteration with less work for everyone.
👇 Jump into the prototyping goodness
Asher retweeted
Another mad science experiment🧑🏻🔬
What if you could drag a screenshot of a UI component to VSCode and have it automatically translated to the closest matching element in your component library?



























