Profile cover photo
Profile photo
Julian Ibarz
Staff Tech Lead Manager in the Brain Robotics team at Google
Staff Tech Lead Manager in the Brain Robotics team at Google

Julian's posts

Post has shared content
The conservative case for climate action: introduce a carbon tax, pay dividends back to families, and remove regulations rendered unnecessary by the carbon tax. Makes sense to me!

Post has attachment
California just installed last month 15% of the electric storage in the world using lithium batteries. This is yet another sign that peak power plant days are counted and will probably be replaced by battery based solutions instead that will smooth the electricity load on the grid and allow for more renewable energy sources to be integrated in the grid efficiently.

Post has shared content
New neural theorem proving paper by our team is on arxiv:

Sarah has integrated TensorFlow model evaluation into theorem prover E to guide the search proof process internally.

Although neural network evaluations are quite slow compared to hand-engineered clause-selection heuristics, we could get decent improvements using tricky evaluation tactics.

Post has shared content
448 million distracting social media posts per year

No, I'm not talking about Facebook!  I'm talking about posts put out by the Chinese government. 

They're often called 50c posts, since rumors say people are paid 50 cents for each post.  In fact there's a huge army of people writing these posts!   I learned about them from this new paper, which did a lot of experiments to study them:

How the Chinese Government Fabricates Social Media Posts for Strategic Distraction, not Engaged Argument

The Chinese government has long been suspected of hiring as many as 2,000,000 people to surreptitiously insert huge numbers of pseudonymous and other deceptive writings into the stream of real social media posts, as if they were the genuine opinions of ordinary people. Many academics, and most journalists and activists, claim that these so-called “50c party” posts vociferously argue for the government’s side in political and policy debates. As we show, this is also true of the vast majority of posts openly accused on social media of being 50c. Yet, almost no systematic empirical evidence exists for this claim, or, more importantly, for the Chinese regime’s strategic objective in pursuing this activity.

In the first large scale empirical analysis of this operation, we show how to identify the secretive authors of these posts, the posts written by them, and their content. We estimate that the government fabricates and posts about 448 million social media comments a year. In contrast to prior claims, we show that the Chinese regime’s strategy is to avoid arguing with skeptics of the party and the government, and to not even discuss controversial issues. We infer that the goal of this massive secretive operation is instead to distract the public and change the subject, as most of the these posts involve cheerleading for China, the revolutionary history of the Communist Party, or other symbols of the regime.  We discuss how these results fit with what is known about the Chinese censorship program, and suggest how they may change our broader theoretical understanding of “common knowledge” and information control in authoritarian regimes.

The conclusion is spelled out in more detail near the end:

Distraction is a clever and useful strategy in information control in that an argument in almost any human discussion is rarely an effective way to put an end to an opposing argument. Letting an argument die, or changing the subject, usually works much better than picking an argument and getting someone’s back up (as new parents recognize fast).

It may even be the case that the function of reasoning in human beings is fundamentally about winning arguments rather than resolving them by seeking truth. Distraction even has the advantage of reducing anger compared to ruminating on the same issue. Finally, since censorship alone seems to anger people, the 50c astroturfing program has the additional advantage of enabling the government to actively control opinion without having to censor as much as they might otherwise.

The paper is here:

• Gary King, Jennifer Pan, and Margaret E. Roberts, How the Chinese government fabricates social media posts for strategic distraction, not engaged argument, American Political Science Review, 2017. Copy at

The people who write these social media posts are often called the 50c army - but I doubt most of them wear uniforms as in this picture!

Thanks to +Lauren Weinstein for pointing this out!

Post has shared content
The Google Brain team's ( long-term goal is to create more intelligent software and systems that improve people's lives, which we pursue through both pure and applied research in a variety of different domains. And while this is obviously a long-term goal, we would like to take a step back and look at some of the progress our team has made over the past year.

Post has shared content
Call for Papers: 1st Conference on Robot Learning

Submission deadline: June 28, 2017
Paper acceptance notification: September 1, 2017
Conference dates: November 13-15, 2017

Post has shared content
An inside look at the Google Brain team

For the past few months, we've had NY Times Magazine reporter +Gideon Lewis-Kraus visit the Google Brain team several times and hang out with us for a few days at a time, with an eye towards writing an article about how our research team operates and what we're working on. We gave him pretty open access to our building, the people in the team, many of our meetings, etc., and over the course of several visits, he decided to focus his story on the origins of the Brain team, and on our in-progress collaboration with the Google Translate team to replace the old phrase-based translation system with a neural machine translation system (essentially part of the article is a behind-the-scenes look at how the scientific work in and came about).

This long article is the result of his visits and synthesis of what he learned. Gideon, I think the article turned out really well!

Post has shared content
You heard it here first: pre-announcing the first conference on robotics and machine learning (CoRL), taking place on November 13th to 15th, 2017!

Research at the intersection of learning and robotics is quickly developing into a large, vibrant community, and this will be a great opportunity to bring everyone together in a very focussed setting.

Paper submission deadline to be announced, but likely end of June. We're currently seeking additional corporate sponsors.

Post has attachment
I am extremely happy to finally see this paper published. This is a very encouraging result showing that a deep neural network trained on many retina images can perform as well as a human ophthalmologist.

In India, millions of people are already severely affected by this disease and a large portion become permanently blind because there is not enough doctors to diagnose it early or because the diagnostic cost is too high. Automating such diagnostic could greatly increase the number of people that can be treated before they become blind.

I had the chance to be involved as a technical advisor early on in the project when I was still in the Street View team because it turns out that some of the techniques we used then to update Google Maps using Street View images and neural networks could be applied for this medical problem.

Being able to reuse deep learning techniques for very different domains such as mapping and medical imaging is one of the main reasons why Deep Learning is growing & improving rapidly. Any improvements done on core vision models positively impact all these applications. And the more applications there are, the more incentives there are to improve such techniques, and the more people work on them.

Post has shared content
If Google says it, it must be true ;)
GOOGLE KNOWS! - I just asked Google Home if Donald Trump is insane. This is the unmodified response, I swear.
Wait while more posts are being loaded