ML/AI Research Engineer @GoogleAI. Working on Expressive ML and approaches in Human-AI Interactions. Opinions are my own. Formerly: @GoogleBrain, @GoogleMagenta
Google
Joined March 2009
- Tweets11.5K
- Following590
- Followers1.5K
- Likes1K
All my #neurips2024 posts were on Bluesky this year. Here's where I am (ffpaladin): bsky.app/profile/ffpaladin.bโฆ
3
136
๐ฟ Sherol Chen, PhD ๐ท๐ ๐ฎ ๐ retweeted
Super excited for our awesome line-up of speakers tomorrow at the Behavioral ML workshop! If you're at @NeurIPSConf , come stop by East Meeting Rooms 19&20!
We're excited to hear from 6 invited speakers:
- Michael Bernstein (@msbernst)
- Andreea Bobu (@andreea7b)
- Tom Griffiths (@cocosci_lab)
- Hoda Heidari (@HodaHeidari)
- Hannah Rose Kirk (@hannahrosekirk)
- Sendhil Mullainathan (@m_sendhil)
๐ฟ Sherol Chen, PhD ๐ท๐ ๐ฎ ๐ retweeted
Join us at the Behavioral ML workshop at NeurIPS tomorrow!
Our focus is on incorporating insights from the behavioral sciences into AI models.
When: Saturday Dec 14 8:45am-5pm
Where: East Meeting Rooms 19&20
๐ฟ Sherol Chen, PhD ๐ท๐ ๐ฎ ๐ retweeted
In case anyone missed Dj's poster from our work at AIES last month -- you can find it here! drive.google.com/file/d/1FPHโฆ
arxiv.org/pdf/2406.16807
The dominant paradigm in AI alignment is to learn from human feedback. But what form should this feedback take? A simple thumbs up/down suffice? Finer-grained attributes ? Our paper ojs.aaai.org/index.php/AIES/โฆ led by the amazing @katie_m_collins at #AIES studies these questions
Checked out @_buildspace the past couple of weeks. Really cool to see so many projects organized and executed in coordination. My project was a YouTube channel for squirrel videos. I got 3000+ subscribers..lol. @FarzaTV sage.buildspace.so/projects/โฆ
4
146
Omg! It's Kai! She's amazing :)
Kai, a software engineer recruiter at Google, explained her top interview & application tips for SWEs! Explore our open roles โก๏ธ goo.gle/3QCYtjz
147
๐ฟ Sherol Chen, PhD ๐ท๐ ๐ฎ ๐ retweeted
__Building SimCity__ launches today w @mitpress!
โจAlan Kay: "Highly recommended!"โจ
๐๐ป30% off today at ๐งtinyurl.com/24asb3wc
(with code PUBDAY30; US address required)
๐๐ปAccidental ASMR teaser video๐๏ธ
๐ฟ Sherol Chen, PhD ๐ท๐ ๐ฎ ๐ retweeted
This ad shows exactly what *not* to do with technology and music.
Never Shake Your Baby.
Never Crush Your Piano.
-- A PSA from g.co/magenta, since 2016 working with musicians and coders to create new and beautiful things at the intersection of AI and audio.
๐ฟ Sherol Chen, PhD ๐ท๐ ๐ฎ ๐ retweeted
Generative AI is changing the game!
We are continuously updating the People + AI Guidebook so that there is more guidance for gen AI applications. At #GoogleIO we showcase 5 principles for building responsible AI products that put users first. Discover how you can optimize the impact of AI while ensuring it remains helpful to users. โ goo.gle/3wqsexj
๐ฟ Sherol Chen, PhD ๐ท๐ ๐ฎ ๐ retweeted
Gemini 1.5 Flash has really great qualities. A really good capable, natively multimodal, 1M token context window (with signup available to get access to a 2M token variant), and super lower latencies and fast response generation.
Today, weโre excited to introduce a new Gemini model: 1.5 Flash. โก
Itโs a lighter weight model compared to 1.5 Pro and optimized for tasks where low latency and cost matter - like chat applications, extracting data from long documents and more. #GoogleIO
๐ฟ Sherol Chen, PhD ๐ท๐ ๐ฎ ๐ retweeted
Get ready to #GoogleIO! ๐ช๐
Tune in to the livestream at 10am PT to hear about the latest launches, news, and AI updates from Google. โ goo.gle/io24-li
So excited for this paper! @MinaLee__ is not just an leading expert but a thoughtful and effective leader!
A Design Space for Intelligent and Interactive Writing Assistants #CHI2024 ๐ฉ๐ปโ๏ธ๐ค
What writing assistants do you use? What else are out there and how do they differ? What do we need to consider when designing new writing assistants?
๐ arxiv.org/abs/2403.14117 (1/6)
3
252
๐ฟ Sherol Chen, PhD ๐ท๐ ๐ฎ ๐ retweeted
I am really excited to reveal what @GoogleDeepMind's Open Endedness Team has been up to ๐. We introduce Genie ๐ง, a foundation world model trained exclusively from Internet videos that can generate an endless variety of action-controllable 2D worlds given image prompts.
๐ฟ Sherol Chen, PhD ๐ท๐ ๐ฎ ๐ retweeted
January 21 is Squirrel Appreciation Day!
Have fun but donโt go nuts.
ALT An image divided into a grid of four smaller images, each of a different type of squirrel from World of Warcraft standing on its hind legs in its native zone. They are as follows: Plague Squirrel: With an ominous green tinge to its brown and white fur, with glowing green eyes and scarring on its face. Squirrel: with luxurious reddish-brown fur along its back and bushy tail, as well as white fur on its face and belly. Mechanical Squirrel: made of gray metal plating instead of fur, with gold and red trim, as well as glowing green eyes. Ardenweald Squirrel: with large, pointed ears swooping outward from its head, as well as glowing yellow eyes and an even bushier tail.
๐ฟ Sherol Chen, PhD ๐ท๐ ๐ฎ ๐ retweeted
Really happy to see the interest around our โHands-on with Geminiโ video. In our developer blog yesterday, we broke down how Gemini was used to create it. developers.googleblog.com/20โฆ
We gave Gemini sequences of different modalities โ image and text in this case โ and had it respond by predicting what might come next. Devs can try similar things when access to Pro opens on 12/13 ๐. The knitting demo used Ultraโก
All the user prompts and outputs in the video are real, shortened for brevity. The video illustrates what the multimodal user experiences built with Gemini could look like. We made it to inspire developers.
When youโre building an app, you can get similar results (thereโs always some variability with LLMs) by prompting Gemini with an instruction that allows the user to "configure" the behavior of the model, like inputting โyou are an expert in science โฆโ before a user can engage in the same kind of back and forth dialogue. Hereโs a clip of what this looks like in AI Studio with Gemini Pro. Weโve come a long way since Flamingo ๐ฆฉ & PALI, looking forward to seeing what people build with it!
๐ฟ Sherol Chen, PhD ๐ท๐ ๐ฎ ๐ retweeted
๐งต๐ Our new preprint is up, and weโd love your feedback! We're "Getting Aligned on Representational Alignment" - the degree to which internal representations of different (biological & artificial) information processing systems agree. ๐ง ๐ค๐ฌ๐ #CognitiveScience #Neuroscience #AI
RT @DynamicWebPaige: For the many folks who are interested in AI and who are reaching out asking where they should consider working nextโsoโฆ