- The speaker discusses the work of the Google Brain Team in using deep learning to improve machine intelligence and enhance people's lives.
- They conduct long-term research and build open-source systems like TensorFlow to support their work.
- The team collaborates with Google and Alphabet to integrate machine learning systems into various Google products.
- They focus on areas such as perception and have made significant progress in 2016, as highlighted in a blog post.
- The speaker explains that the field of deep learning and neural networks has revolutionized how problems are approached in machine learning.
- With advancements in compute power, neural nets have become the best solution for many problems.
- The speaker discusses the growth of deep learning at Google and the importance of reducing experimental turnaround time.
- They introduce TensorFlow, an open-source platform for expressing machine learning ideas and deploying them in various settings.
- TensorFlow has gained popularity and is known for its flexibility, scalability, and ability to run on any platform.
- The team is also focused on improving the speed and performance of TensorFlow models through benchmarking and optimization.

- TensorFlow demonstrates nearly linear speed up for image models on up to 8 GPU cards and close to linear speed up for 64 GPU cards.
- TensorFlow supports various platforms, including iOS, Android, Raspberry Pis, CPUs, and custom machine learning accelerators.
- TensorFlow aims to be language agnostic, with Python being the most fully developed system and C++ frontend working well for production use.
- TensorFlow has a broad usage base, with many companies and universities utilizing the platform.
- Google Photos is an example of a successful application of computer vision using TensorFlow.
- TensorFlow's model structure can be reused for different data sets to create unique product features.
- TensorFlow has been applied in various domains, including identifying text in Street View images, estimating solar energy potential on rooftops, and diagnosing diseases in medical images.
- TensorFlow has achieved performance on par with ophthalmologists in diagnosing diabetic retinopathy.
- TensorFlow has been used in robotics for perception and imitation learning, with the ability to train models using data from multiple robots for better performance.
- TensorFlow has released a large public dataset of grasp attempts for robotic grasping.

- The speaker discusses the use of deep learning in various domains, including robotics and scientific simulations.
- They explain how simulators can be used as training data for neural networks, resulting in significant speed improvements.
- The application of deep learning in predicting depth from images is mentioned, along with its potential for creating virtual stains in microscope images.
- The speaker also talks about language understanding models, specifically in the context of translation and email replies.
- They mention the success of the Smart Reply feature in Gmail, where a neural network predicts plausible replies to incoming emails.
- The speaker concludes by discussing the challenges of scaling up deep learning models and the use of data parallelism.

- The technique of using parallel data parallel copies to update shared parameters allows for quick scaling of training.
- The quality of translations generated by a neural machine translation system is significantly better than the old phrase-based system.
- The neural machine translation system is much simpler, consisting of only 500 lines of TensorFlow code.
- Efforts are being made to automate the solution of machine learning problems, such as designing neural architectures and learning optimizers automatically.
- Neural architecture search has been successful in generating model architectures that perform close to the state of the art without human knowledge.
- Learning optimizer rules has also shown promising results, with the automatically generated expressions outperforming human-designed ones.
- The concept of "learning to learn" has the potential to significantly increase the number of experiments that can be run, leading to better problem-solving in machine learning.
- Deep learning has transformed the design of computers by leveraging large amounts of data and scalable compute power.

- Deep learning is tolerant of reduced precision arithmetic and can be optimized for specific operations
- Google has developed custom machine learning hardware called TPUs for both training and inference
- TPUs are designed to be hooked together in a system called a pod, allowing for increased compute power
- Google is making TPUs available through Google Cloud and offering 1,000 devices for free to researchers
- Google Cloud also offers higher level APIs and pre-trained models for easy use of machine learning
- Google is experimenting with using machine learning to optimize machine learning models
- Deep neural nets are making significant advancements and have potential for various applications such as image captioning and robotics
- Google encourages people to visit their website for more information on their work.

- The architecture of neural networks can be complex, but skip connections are known to be effective.
- The ability to run thousands of experiments quickly makes automated systems more efficient than humans.
- Large amounts of labeled data are needed for training models, but there are ways to improve data efficiency.
- The goal is to build models that can perform multiple tasks and learn from previous knowledge.
- Faster iteration times and reducing the reliance on trial and error are important in neural net development.
- Empirical research and experimentation are crucial in machine learning.

- The speaker discusses the importance of interpretability in machine learning models for delivering production models to non-experts.
- Interpretability is particularly important in domains like healthcare, where users want to understand the reasoning behind predictions.
- The speaker explains that their interest in machine learning began when they saw the success of neural nets in certain domains and believed they were the right abstraction.
- The speaker mentions the need for larger-scale multitask models and a persistent knowledge store to develop reasoning systems in AI.
- Memory networks and differential neural computing are mentioned as interesting areas for further exploration.
- The role of data scientists and ML experts is discussed in the context of automated architecture search.
- The speaker expresses excitement about the application of neural nets in healthcare for making better predictions and improving healthcare decisions.

- The Google Brain Team uses deep learning to enhance machine intelligence and improve people's lives.
- They collaborate with Google and Alphabet to integrate machine learning systems into various products.
- TensorFlow, an open-source platform, is used for expressing and deploying machine learning ideas.
- TensorFlow has gained popularity for its flexibility, scalability, and ability to run on different platforms.
- It has been successfully applied in various domains such as computer vision, robotics, and language understanding.
- Deep learning has transformed computer design and Google has developed custom hardware called TPUs.
- Google Cloud offers TPUs, APIs, and pre-trained models for easy use of machine learning.
- The goal is to build models that can perform multiple tasks and learn from previous knowledge.
- Interpretability in machine learning models is important, especially in domains like healthcare.
- The speaker is excited about the application of neural nets in healthcare for better predictions and decision-making.

Jeff Dean is a Google Senior Fellow in the Research Group, where he leads the Google Brain project.

Jeff's slides are available here: http://blog.ycombinator.com/jeff-deans-lecture-for-yc-ai/

So I'm going to tell you a very not super deep into any 1 topic, but very broad brush sense of the kinds of things we've been using deep learning for, the kinds of systems we've built around making deep learning faster.

And this is joint work with many, many, many people at Google.

But most of it is from the Google Brain Team, which I lead.

And so the Brain Team's mission is basically make machines intelligent, and then use that new capability to improve people's lives in a number of different ways.

And the way we do this is we conduct long-term research, kind of independent of any particular application in, oh, Sorry, I'm probably supposed to stand in 1 place.

Independent of any particular application, we build an open source systems that help us with our research and deploying of machine learning models like TensorFlow.

We collaborate across Google and all of Alphabet in getting machine learning systems and research that we've done into real Google products.

So we've done a lot of work in Google Search, Gmail, photos, speech recognition, translate, and many other places.

We also bring in a lot of people into our group through internships and a new residency program that we started last year for people who want to learn how to do deep learning research and that's been a pretty successful program as well.

So the main research areas that our group is working in are these.

So in January I put out a blog post that kind of just highlighted some of the work our group has done over 2016.

And in putting that together I kind of realized we were doing a lot of different things.

So the nice thing about this is each 1 of these blue links is a link to something kind of interesting and substantial, like a research paper or a product launch using machine learning or some new TensorFlow features we've added.

So I won't go through that all now, but you can go find that blog post and learn more about some of the stuff we've been up to.

You probably already all know this, given that you're working on AI-related companies, as I understand it.

But the field of deep learning and neural networks in particular are really causing a shift in how we think about approaching a lot of problems.

And I think it's really changing the kinds of machine learning approaches that we use.

In the 80s and 90s, it was the case that neural nets seemed interesting and appealing, but they weren't the best solution at the time for a lot of problems that we cared about because they just didn't quite have enough training data, enough computational capabilities.

And so people used other methods or developed kind of shallower machine learning methods with much more hand engineering features.

And if you fast forward to now, what's happened is we've got much, much more compute.

I actually did an undergrad thesis in 1990 on parallel training of neural nets because I liked the appealing attraction of the neural net model.

And I thought if we could just get a bunch more compute by parallelizing over a 64 processor hypercube machine that it would all be even better.

It turned out what we needed was like 100,000 times as much compute, not 60 times.

But if you fast forward to today, we actually have that.

And so what's happened is we now actually have the case where neural nets are the best solution for an awful lot of problems, and a growing set of problems where we either previously didn't really know how to solve the problem, or where we could solve it, but now we can solve it better with neural nets.

And so the talk is really meant to orient you across a whole bunch of different problems where this is the case.

So really, our group started in order to investigate the hypothesis that large amounts of compute could actually solve interesting problems using neural nets.

And so when we first started, we were sort of the vanguard of people using neural nets at Google.

We did a bunch of work on unsupervised learning at a really large scale using, at that time, we didn't even have GPUs in our data centers, so we just used 16,000 CPU cores.

And we did kind of interesting things with unsupervised learning there.

But gradually, we kind of built tools that enable people to apply machine learning and deep learning in particular to a lot of problems.

And you can see the growth rate of this is directories containing model description files, either from our first generation system and starting in about 2015 our second generation system TensorFlow.

And we've deployed machine learning and in collaboration with lots of teams.

Other teams have also been independently just picking up this idea of deep learning and using it in lots and lots of places in Google products and that's why you see that growth rate and it's continuing to go up.

1 of the things we focus on a lot is how can we reduce experimental turnaround time for our machine learning experiments And because there's a very different qualitative feel to doing science and research in a domain where an experiment takes a month versus doing it in a domain where, you know, minutes or hours you get, you know, an answer and then you can figure out what the next set of experiments are that you want to run.

So a lot of our focus is on scaling machine learning models and scaling the underlying infrastructure and systems so that we can actually for some problems approach minutes or hours rather than weeks or months.

So part of that has been building the right tools.

So TensorFlow is kind of our second generation system that we built for tackling deep learning problems and machine learning problems.

The second 1 we said we should really fix some of the design problems we saw in our first system, keep the good features about it, and then design this from the start to be an open source platform so that people all over the world, not just at Google, can benefit from this and can help build a community that can all contribute to and improve the system.

Zach Stone here is our TensorFlow product manager extraordinaire and has been doing a great job of building the community both inside Google and outside.

So the goals of TensorFlow that we want to establish this common platform for expressing all kinds of machine learning ideas.

So something that can be used for deep learning, can be used for other kinds of machine learning, that can be used for tackling perception problems and language understanding problems.

And if you have a crazy new machine learning research idea that doesn't really fit into what people have done before.

We want it to be at least expressible relatively easily in TensorFlow.

And then we want to make that platform really great for research, but we also want to be able to take something you've developed in TensorFlow maybe experimentally.

And then if you now want to deploy that in a production setting, run it at a data center, run it at scale, run it on a phone, all these kinds of things, we want that to be another something you can do in the TensorFlow framework.

And by open sourcing it, we make it available to everyone.

Well, so this is a comparison of GitHub stars, which is 1 metric of popularity or interest in different source code repositories on GitHub.

And I show you a comparison of TensorFlow with a bunch of other open source machine learning packages, many of which have been around for many more years than TensorFlow.

TensorFlow is this brown line going up fairly steeply.

I think the reception for what TensorFlow does, which is enable flexible research, but also this kind of production readiness and being able to run in lots of places is pretty appealing.

And if you look at the other open source packages, which we did when we were starting to work on TensorFlow, many of them have 2 of the 3 attributes that we care about, being able to be really flexible, scalable, and run on any platform.

And they all have different emphases, but we wanted something that satisfied all 3 of those.

We've been focusing a fair amount on speed.

I think when we first released TensorFlow, we released a bunch of really nice tutorials that showed how to do different things with TensorFlow.

But 1 of the mistakes we made was we released code that was meant to be expository and clear, and not necessarily the highest performance way you would write that.

But often then people would take that as the way you should write a high-performance TensorFlow model and that wasn't necessarily a case.

So we're now adapting and trying to put out things that are both the best way from a clarity standpoint, but also are high performance.

That's been a bit of a TensorFlow got, I think, a bit of a bad rap.

But actually, our performance is quite good.

So we've been doing a bunch of benchmarking and producing reproducible benchmark results to show that our scaling is quite good.

So this is single machine scaling, nearly linear speed up for a bunch of different image models on up to 8 GPU cards, pretty close to linear speed up for 64 GPU cards for a bunch of different kinds of problems.

So if you hear TensorFlow is slow, don't believe it.

We also support lots of different platforms, and I think this is important because often you want to train a model on a large data set in a data center, but then deploy that on a phone.

And so we run on iOS and Android and Raspberry Pis, but also CPUs and if you have a GPU card or 8 GPU cards, we're happy to use that.

We also run on our custom machine learning accelerators that I'll talk about in a minute.

But really, we want to run on everything.

So there's a bunch of other device manufacturers that are developing kind of funky mobile ML accelerators, or Qualcomm has a DSP, and they're all working to make sure that TensorFlow runs well on those devices.

We also kind of want to be agnostic of language that people want, because you want to be able to run machine learning where it makes sense and different people have different sort of language environments.

The most fully developed system is obviously Python, but the C++ frontend works pretty well for production use, And then a bunch of other people, external community members, have added a variety of other kind of not fully fleshed out but reasonable support for some of these other languages.

So like a year ago, we had a, almost a year ago, we had a meeting at Google of people using TensorFlow.

We had people from most of these companies in the room, which I think normally they don't all get in a room together.

Places like Apple was actually there as well, NVIDIA, Qualcomm, Uber, Google, Snapchat, many, many Intel, many, many other places.

So In terms of stars, I showed you the graph related to machine learning platforms, this is the top repositories on GitHub overall and we're up to number 6, which is pretty good.

And all the other ones are either JavaScript or a list of programming books.

This is a visualization of where people are interested in different GitHub repositories, which is kind of cool.

Machine learning is done all over the world.

So 1 of the things that's happened as that growth in interest has happened is there's been a pretty broad set of external to Google contributors.

And so there's really, you know, I think we're up to almost 1,000 non-Google contributors across the world doing all kinds of different things for adding features or fixing bugs or improving the system in various ways, which has been really nice.

Oh, and I think It's kind of nice that there's growing use in machine learning classes of TensorFlow as a way of illustrating machine learning concepts.

So at really good machine learning universities like Toronto, Berkeley, Stanford, and other places, they're starting to use that as the core of their curriculum.

OK, so now I'm going to switch gears a bit and talk about some sort of more product-oriented applications of deep learning at Google.

Obviously computer vision now works and 1 thing you could do is make a photos product around the idea that you can actually understand what's in people's photos and that's been going really well.

As a lesson for you who are starting companies often in applied domains, I think it's really important to be able to look at the machine learning work that is happening in the world and realize that often you can reuse many of the same ideas from 1 domain.

And just by pointing it at different data sets, get completely different, interesting product features.

So if you, for example, use the same basic model structure, train it on different data, and you get something different.

1 general model trend is give it an image, predict interesting pixels.

So there's a bunch of ways you could do that.

But if you have a model structure that does that, my summer intern from a few years ago, Matt Zehler, who actually went off to found Clarify, which is a computer vision company, we were working in collaboration with the Street View team on identifying text in Street View images.

And so to do that, you can have training data where people have circled or drawn boxes around text.

And you just try to predict the heat map of which pixels contain text in a Street View image.

And then you can run an OCR model on those pixels and actually read the text.

And it works across lots of different font sizes and colors and whether it's close or far from the camera.

And so then some people in the Maps team decided they would build this thing that would help you identify whether your rooftop has solar energy potential and how much energy you could generate by installing solar panels.

And so obviously 1 of the first things you have to do is find rooftops.

And that's exactly the same model, but just with different training data where you now have circles around rooftops.

And then there's a bunch of other work to estimate the angle of the rooftop from the imagery or multiple views of the same house, and then some stuff to predict what is the solar energy potential for that.

And another area where we've applied this is in the medical domain.

So the same basic model, we want to be able to, say, take a medical imaging problem.

1 of the first ones we've been tackling is ophthalmology problems.

And in particular, taking a retinal image like this and deciding whether or not this has symptoms of a degenerative disease called diabetic retinopathy.

You want to identify parts of the eye that seem to be diseased in some way.

Then you also have a whole image classification problem of does this eye show symptoms at the level of 1, 2, 3, 4, or 5.

Some people in our group have done a really nice medical study showing that if you collect 150,000 ophthalmology images and then you get each 1 labeled by 7 ophthalmologists because if you ask 2 ophthalmologists to grade the same image 1234, or 5, they agree 60% of the time, which is slightly terrifying.

If you ask the same ophthalmologist to grade the same image a few hours later, they agree with themselves 65% of the time.

So we had to get every image labeled by 7 ophthalmologists to reduce the variance in the score and say, oh, 5 people think it's a 2, so it's probably more like a 2 than a 3.

But in any case, the punchline of this paper is We now have a model that performs on par, slightly better than the median of 8 US board certified ophthalmologists, which is cool because there's a bunch of places in the world, especially in India and other countries where there are many people at risk and there just aren't enough ophthalmologists.

So we're actually doing clinical trials in India.

We've licensed this to our Verily subsidiary who's licensed it to a camera, an ophthalmology camera manufacturer who's going to be integrating this into the actual ophthalmology camera.

Another area where being able to see is pretty useful is robotics.

If you're trying to build robots just being able to perceive the world around you clearly makes things a lot better.

So we've been doing a bunch of experiments both with real robots and also with simulated robotic environments and also with trying to do imitation learning from people performing actions and then trying to get robots to do it.

Well, anyway, it's not that exciting, except that we have a bunch of robots trying to grasp things, and they can essentially learn on their own whether they're grasping something successfully by just having a bin of things in front of them and they just try to pick something up.

And if they fail, their gripper closes all the way.

If they succeed, then they don't close their gripper all the way and they can actually see from the camera that they've managed to pick something up.

And so they can practice picking things up and we can pool all the sensor data from all the robots that are doing this to retrain a model every night for grasping so that the next day's grasping attempts are better and better and by having lots of robots do this you actually get a lot of parallel experience, much more than you can get on a single robotic arm.

And so we have a data set that we've actually released publicly of about 800,000 grasp attempts versus about 30,000 grasp attempts, which was kind of the big public data set in the past.

And surprisingly, 800,000 grasp attempts gives you a much better grasping mechanism and model than 30,000.

We've also been trying to, this is me awkwardly looking at a robot on a screen that you can't see, doing some actions.

I'm trying to like mimic the robotic nature of it.

And then we have a video of me doing that.

And then we're just trying to learn from the videos to transfer that action to the real robots.

And we're doing that first via simulator, and then we're taking that simulator and then trying to transfer those activities to a real robot.

Another place that I'm pretty excited about deep learning is in lots and lots of scientific domains you often have the case where you have a simulator of some really complex phenomena and that's often a sort of HPC style application and very computationally expensive but it kind of gives you insight into whatever scientific processes are going on and that allows you to kind of iterate in a computational science methodology.

But often those computations are pretty expensive and so 1 of the things we've been working on and this is just 1 example we have a lot of different fields of science where we've seen this to be true is you can use those simulators as training data for a null net.

So quantum chemists have a problem where they take in a configuration of molecules, and they run a bunch of time steps.

And then at the end, they get some information about how the ultimate configuration of those molecules turns out, and from that, they get a few properties about those molecules.

So it turns out that's the data that, if you use that as input, you run this really expensive simulator for an hour, and then you get these 13 numbers out.

That turns out to be great training data for a neural net.

And so you can train a neural net to do exactly that same task or to approximate that task, approximate the entire simulator.

And you can essentially, the punch line is the bottom there, you essentially get indistinguishable accuracy from using the real simulator but it's 300,000 times faster.

And that has a lot of implications for how you might do quantum chemistry.

If you suddenly have something that's 300,000 times faster, you might like run a hundred million things through your neural net-based simulator and figure out what's going on.

To identify a bunch of candidates that you might want to look into in more detail.

Another place where these kinds of pixel-to-pixel models come in is some people in Google have done a model that tries to predict depth from an input image and we have some training data where we have the true depth given the camera viewpoint and where things are in the world.

And then we try to train a model to do predicted depth from just the raw pixels of that image.

So that's a pixel-to-pixel learning problem.

And you can imagine a lot of pixel-to-pixel learning problems.

And indeed, 1 application in cameras is you want to predict depth in a portrait and then you can do kind of funky cool effects like identify the person in the foreground and turn the background black and white or like make it all fuzzy and artsy in the background, which is kind of cool.

But it turns out you can also take microscope images as the raw microscope image as input and then the chemically stained microscope image as the target for your model.

And so for example that's often how people see cell bodies and cell boundaries is you apply different kinds of stains to the cells and then you can make them show up on a microscope better and you can see what's going on.

Well it turns out so this animation that's the input, that's the ground truth, and that's the predicted output of a neural net that's trained to sort of virtually stain something without actually staining it.

And this is important because it turns out when you actually stain something that kills the cells.

So you don't get any temporal information about what's going on in the cells.

They essentially die when you apply the stain.

But here, you can virtually stain something, but then follow them longitudinally in time and see how cell processes kind of continue to happen without actually staining them.

You can also stain for things that you can't actually really necessarily develop a true chemical stain.

So if you have someone label which things are axons and which things are dendrites and neural tissue, you can have a microscope viewer that highlights axons and dendrites in different colors and cell bodies, even if that's kind of not something that you can chemically do with a real stain.

the areas we've been doing a lot of work is in language understanding models.

And so this started out as research in our group to do essentially sequence-to-sequence learning.

So you have some input sequence and conditioned on that input sequence you want to predict an output sequence.

So this turns out to be useful for actually a whole bunch of different problems, but 1 of them is translation.

So if you have a bunch of sentence pairs, 1 in French and the corresponding meaning sentence in English, then you can use a sequence to sequence model to take the input sentence 1 word at a time, or even 1 character at a time.

And then when you hit a special end of French token, then you essentially start spitting out the corresponding English translation of that French sentence.

And you have training data that is like that.

And you just try to predict the next word from that training data using recurrent neural net.

And that turns out to work reasonably well.

And then you're actually trying to find the most probable sequence, not the sequence with the most probable individual terms, and so you do a little beam search where you kind of keep a window of candidates and you sort of search over possible vocabulary items until you are happy and found a likely output sequence, and that's how you do translation.

We added a feature called Smart Reply, where essentially we get an incoming email.

So this is 1 sent to my colleague Greg Corrado from his brother.

Hi, we wanted to invite you to join us for Thanksgiving dinner.

So to reduce the computational cost, we have a small feed-forward neural net that says, is this the kind of thing where a small, short reply would make sense?

And if yes, then we are going to activate a sequence-to-sequence model.

And we're going to do a much more computationally expensive thing with that in messages input and then we're going to try to predict plausible replies and So this system produces 3 it says count us in we'll be there or sorry.

And so this is a nice application of sequence-to-sequence models.

And if you squint in the world, you'll find lots of applications of these.

And so it turns out Smart Reply, In April 2009, there was an April Fool's joke that Google put out saying, ha-ha, we're going to reply to your email automatically.

But then in November 2015, we launched this as a real product.

And in just 3 months, 10% of mobile inbox replies are generated by these smart replies.

But obviously 1 of the real potential applications of this was Translate, which is what we were doing, demonstrating that this research was effective on a large by academic standards but smallish public data set of translation data called WMT.

So when we looked to work on applying this to the real Google Translate product, we actually had 100x to 1,000x as much training data.

And so scaling this up was actually pretty challenging.

And we wanted to make the model a lot higher quality, but we did a fairly detailed write-up of the engineering behind that in this many many author paper.

So this is kind of the structure of the model that we came up with.

It has a very deep LSTM stack, each of which runs on a different GPU.

There's an attention module so that rather than just having a single state that's updated by the recurrent model, you keep track of all the states.

And then you learn to pay attention to different parts of the input data when you're generating different parts of the output sequence.

So you're about to generate the next word, and you look back at the word hello in the input sentence, and so on.

And so this thing runs on 1 replica of this model, runs on a machine with 8 GPU cards with different pieces of it in different places.

And then we run a lot of copies of this model to do data parallelism across the large training data.

So This is a technique we've been using for quite a while that we originally published in 2012 about what we call at that time a parameter server and then using many parallel data, data parallel copies to process different input data, all trying to update those shared parameters by applying gradients to those parameters.

And this allows you to scale training quite quickly.

So you can have 50 replicas of this kind of setup or 20.

I think in this case we were using about 16.

So we're using 100 GPU cards to train a model.

And the really good news is the blue line here is the old phrase-based machine translation system that didn't really have much machine learning in it, any machine learning in it, it had large statistical models for lots of different sub-pieces of the problem, so it had a target language model that told you how often every 5 word sequence in English occurred, it had an alignment model that says how words in English and French sentences align, it had a phrase table and a dictionary of plausible English and French phrases and sentences, and it was like 500,000 lines of code to glue this whole thing together, and that's the blue line.

And what we're showing is the quality of translations generated by that system as judged by humans.

And the green line has a substantial jump in quality for basically Nearly every language pair jumps up very substantially.

It doesn't look like much, but those are really big jumps in quality.

And the other nice thing is that system is 500 lines of TensorFlow code instead of 500,000 lines of GUI code with lots of handwritten logic.

And the yellow line on top is human, bilingual human, not professional translator, but someone who speaks both those languages, translations as judged by other humans.

And so you can see that for some language pairs, we're actually getting quite close to that human level quality for translation, which is pretty exciting.

See all Y Combinator transcripts on Youtube

Jeff Dean’s Lecture for YC AI