Profile cover photo
Profile photo
Ben Wang
Ben's interests
View all
Ben's posts

Post has shared content
Something sketchy going on with machine learning

David Ha, one of this year's Brain Residents (, wrote up a very nice Google Research blog post about the work that he and +Douglas Eck have been doing on generative models for sketch-style drawing.

David and Doug have a new Arxiv paper that was just posted about the details of the work at titled "A Neural Representation of Sketch Drawings" for the details, but the blog post gives a really nice overview of what these sorts of generative models can do. The blog post goes through many variations of pretty cool things you can do with this model:

o "Draw me a cat", and it can generate different many different cats
o "Draw me a cat that looks roughly like toothbrush", and it can draw a cat that looks like a toothbrush
o Smoothly transition between cat pictures that are just faces versus ones that show the whole body by manipulating a latent variable that is learned by the model
o Do "vector arithmetic in drawing space": "cat head + (pig with body - pig just head)" gives "cat with body"
o Complete drawings given a category and just a line or two: "Finish this fire truck"

Just a taste of the kinds of creative tools that machine learning will be able to provide in the coming years.

Blog post:
Arxiv paper with details:

Post has attachment
Google’s AutoDraw uses machine learning to help you draw like a pro

Post has shared content
Since the 1920s, excessive pumping of groundwater in California’s San Joaquin Valley has caused land in sections of the valley to sink by as much as 28 feet (8.5 meters), a problem exacerbated during droughts, when farmers rely heavily on groundwater to sustain one of the most productive agricultural regions in the nation. But just how much sinking? Scientists at our Jet Propulsion Laboratory use their expertise in collecting and analyzing airborne and satellite radar data to help answer that question. Learn more:

Post has shared content
I'm very excited about the work our group ( is doing in various areas of medical imaging. Today, we published a preprint titled _"Detecting Cancer Metastases on Gigapixel Pathology Images" that showed that our ML model achieves a tumor localization score of 0.89 vs. 0.73 achieved by human pathologists with infinite time. The paper authors are Yun Liu, Krishna Gadepalli, Mohammad Norouzi, George E. Dahl, Timo Kohlberger, Aleksey Boyko, Subhashini Venugopalan, Aleksei Timofeev, Philip Q. Nelson, Greg Corrado, Jason D. Hipp, Lily Peng, and Martin C. Stumpe.

Some key statistics from the paper used to evaluate the effectiveness of this work:

Tumor localization score (FROC): ("find all the tumors")
0.89 (our model)
0.73 (human pathologists with infinite time)

Sensitivity at 8 false positives:
0.92 (our model)
0.73 (human pathologists with infinite time)

(It's also worth point out that Yun Liu is a member of the Google Brain Residency program:

This work follows on our earlier work on detection of diabetic retinopathy in retinal images.

Pathology blog post:
Paper preprint:

Earlier blog post on diabetic retinopathy work:

(Edit: Slight rewording)

Post has shared content

Post has attachment
Uber launches 'urgent investigation' into sexual harassment claims

Post has shared content
The Google Brain team — Looking Back on 2016

I wrote up a blog post about the work the Google Brain team has been doing over 2016. I'm really excited to work with such great colleagues! In writing this up, it's pretty remarkable to me that nearly every other sentence has one or a few links to many more details on significant and impactful work.

Post has attachment

Post has attachment

Post has shared content
Not unique to the U.S.
Wait while more posts are being loaded