Reiichiro Nakano
@reiinakano
Followers
4,575
Following
904
Media
81
Statuses
1,497
i like building awesome things with awesome people 🇵🇭 🇯🇵
San Francisco
Joined August 2009
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
#NevaPlay
• 464665 Tweets
SAROCHA X MODERN GURU
• 366497 Tweets
#MEGANxRM
• 350056 Tweets
Ravens
• 256544 Tweets
NAMJOON
• 253539 Tweets
Linkin Park
• 122306 Tweets
Lamar
• 116817 Tweets
Dybala
• 109719 Tweets
Mahomes
• 86787 Tweets
Flowers
• 77297 Tweets
Chester
• 63812 Tweets
Kelce
• 53143 Tweets
Xavier Worthy
• 48278 Tweets
#TZUYU_SOLO_RUNAWAYDAY
• 41517 Tweets
#BAZAARMENThailandxGemini
• 34680 Tweets
FOR ASIAN MEN WE PAVED THE WAY
• 33443 Tweets
サブスク解禁
• 33106 Tweets
FANZA
• 29174 Tweets
Derrick Henry
• 27076 Tweets
Mark Andrews
• 23592 Tweets
Hayırlı Cumalar
• 20546 Tweets
進次郎氏
• 17816 Tweets
Isaiah Likely
• 16807 Tweets
夫婦別姓
• 15957 Tweets
ちいかわ寿司
• 15350 Tweets
モモンガ
• 15259 Tweets
アストロボット
• 14944 Tweets
TransisiSEJUK PRABOWOJKW
• 14481 Tweets
KitaKOMPAK KitaMENANG
• 13845 Tweets
ニューヨーク
• 12687 Tweets
ハチワレ
• 11824 Tweets
#ドミノシェイクは水不使用
• 11419 Tweets
Pinned Tweet
Style transfer without a specific style image. I minimize only the content loss from neural style transfer and directly optimize brushstrokes via backpropagation. In this case, "style" is an intrinsic property defined entirely by the artistic medium.
44
617
3K
I am on an H-1B, in the process of getting my green card and relocating my family to the US.
Me and many other colleagues in a similar situation have signed this letter.
I do not know what will happen next, but I am confident we will be taken care of.
the board should resign.
This all happened after midnight, during a company wide holiday. Support was always near unanimous.
38
173
2K
84
172
3K
New blog post: Neural Style Transfer with Adversarially Robust Classifiers
I show that adversarial robustness makes neural style transfer work by default on a non-VGG architecture.
Blog:
Colab:
4
261
1K
At
#NeurIPS2019
, talking to a guy from DeepMind about how I can't reproduce their results because it's so resource intensive.
"It's not too bad actually, we only use 16 GPUs, it's not like we use a cluster of TPUs or anything."
11
96
780
Some personal news: I moved to San Francisco and joined the OpenAI Fellows Program!
Super excited to learn from all the amazing people here. It's a dream come true to be able to do machine learning full-time!
25
19
609
My work is out on arXiv!
Neural Painters: A learned differentiable constraint for generating brushstroke paintings:
There are also
@GoogleColab
notebooks so you can reproduce all the experiments yourself:
10
148
422
Sharing my winter break project. I tried combining
@hardmaru
's world models and
@yaroslav_ganin
's SPIRAL to see if an agent can learn to paint inside its own dream. It can! These strokes are generated purely inside a world model, yet transfer seamlessly to a real paint program.
3
99
399
My first ever citation is from a Schmidhuber paper! 😳😁😃😍
6
10
388
New blog post: Solving simple probability problems with intermediate steps and an external calculator
Link:
A transformer is used to map questions to intermediate steps, while an external symbolic calculator evaluates intermediate expressions.
8
83
367
I built a demo for arbitrary image stylization in the browser using
@TensorFlow
.js at . Instead of having a single network per style like previous algorithms for fast style transfer, only a single model is used for all style images. more details in replies
8
105
333
I haven't tweeted much about ML in a ~year, but happy to finally share what I've been working on with amazing collaborators - we trained GPT-3 to use a text browser!
We trained a research version of GPT-3 that can search the web, synthesize information, and cite its sources to provide more accurate answers to questions.
70
359
2K
19
32
317
Not sure why this hasn't gotten more attention. Seems like a viable alternative to Colab that addresses a lot of my pain points e.g. no random restarts, easy upgrade to more powerful GPU, actually uses Jupyter UI instead of some weird closed-source fork
2
44
146
6
74
312
"GAN Playground - Experiment with Generative Adversarial Nets in your browser": built with
@deeplearnjs
4
126
304
Directly optimizing brushstrokes using backpropagation makes it very easy to find a sequence that "paints" an image.
Ironically, the almost perfect reconstruction makes the end result feel less creative and more "robotic". 🤔
5
54
275
""ML frameworks don’t just enable machine learning research, they enable and restrict the ideas that researchers are able to easily explore.""
Perhaps the next big step forward is something completely inexpressible in the TF/PyTorch paradigm.
2
72
228
Colab is slowly becoming a full-blown browser IDE, ala
@replit
but with free GPU/TPUs
You can now edit📝, create🆕, save💾, and move➡️ files and folders📂 through file browser on the left.
22
147
696
0
29
211
💜
2
7
209
.
@SchmidhuberAI
explaining why GANs are a special case of his 1990 work on unsupervised minimax
5
28
152
What if Adam is only as popular as it is because it's easier to type than Adadelta or Adagrad or RMSProp
2
9
141
Every time I read machine learning code, I can't help but think I'd understand things 10x faster if it were statically typed..
7
7
109
A machine generated this image.
In 1995.
If a GAN generated this today it would break machine learning Twitter.
2
17
90
Today I had the opportunity to listen to Rich Sutton as he spoke to a few of us at the
#NeurIPS2019
RL Social.
He was incredibly nice and humble, patiently listening to all questions and responding thoughtfully.
Very short thread of his insights (sadly I arrived quite late)
2
0
90
The Joy of Neural Painting ()
Someone wrote an implementation of Neural Painters using PyTorch and .
They use NoGAN by
@citnaj
to speed up training of the GAN-based Neural Painter. Very cool work!
1
28
88
Not sure why this became so popular. Passing a conditional as a string scares me 😨
3
1
85
It's happening
0
22
80
Happy to announce that this model has made its way into magenta.js! Now anyone can use this model in their own apps with a few lines of code. Also wrote a blog post about my experience porting the model!
magenta-js lib:
Blog post:
2
19
68
In any case, this is another followup to my work on neural painters on which I am currently sprinting as fast as I can to write a paper for. All code will be open-sourced as
@GoogleColab
notebooks.
Sharing my winter break project. I tried combining
@hardmaru
's world models and
@yaroslav_ganin
's SPIRAL to see if an agent can learn to paint inside its own dream. It can! These strokes are generated purely inside a world model, yet transfer seamlessly to a real paint program.
3
99
399
1
4
69
@hardmaru
I'll be giving a talk on Neural Painters in this year's
#NeurIPS2019
Creativity Workshop! Looking forward to meeting everyone in Vancouver!
#neurips4creativity
1
8
64
the hardest part of this was the google doc crashing with too many people trying to edit it at the same time
actually guy posting a link to a google doc and saying “sign if you agree” is exactly what we did
29
49
2K
0
4
61
my career goal is to be important enough to get called out for following siraj raval
0
1
62
I'll be teaching some TensorFlow.js basics together with
@Lewuathe
at the Google office in Tokyo next week. Sign up if you're interested!
4
8
57
I personally think content loss is the best idea from the neural style transfer paper. We can preserve human-recognizable high level structure from a target image, without caring about exact pixel matching.
1
3
59
Writing a paper is so boring... I'd much rather be writing code and trying out new experiments.
I guess it's important to disseminate ideas, though 🙄
3
6
51
MNIST was easy, so I tried KMNIST. Simple chars like つ and ﾊ were easy but overall, the dataset is much harder😅. I'm confident it can be improved with more strokes + training steps, though. Stroke order is also all over the place.
@tkasasagi
@mikb0b
@hardmaru
@KitamotoAsanobu
2
9
51
Is trax/flax/stax the new tf.layers/tf.contrib.layers/tf.contrib.slim?
Flax: A neural network library for JAX designed for flexibility (pre-release)
11
111
410
1
4
48
At the end he smiled and said
"You guys have worn me out. It was great talking to all of you. It's nice to get a perspective of where the field is at." (from the person who basically invented the field)
Then we took this selfie.
0
3
44
Interesting
#NeurIPS2019
paper
Putting An End to End-to-End:
Gradient-Isolated Learning of Representations
They train a neural net without end-to-end backprop and achieve competitive performance.
We are putting an End to End-to-End!
Check out the camera-ready version of our paper selected for an oral presentation
@NeurIPSConf
:
W/
@BasVeeling
& Peter O’Connor
#neurips2019
#neurips
We also published our
@PyTorch
code:
3
34
124
2
11
44
Extremely grateful to
@distillpub
and
@ch402
for letting me be a part of their first experimental "discussion article"! Had a lot of fun interacting with the paper authors.
I think it's been a successful experiment so far. Looking forward to more of these in the future!
A Discussion of 'Adversarial Examples Are Not Bugs, They Are Features' - Six comments from the community and responses from the original authors.
6
200
613
2
2
42
Can we flip the question?
Can we automate CEOs? Why can't we progress as a human race with 1% of the CEOs we have?
4
17
34
☝️Can finally put this next to my name ☝️
Will be busy fighting jetlag for the next few days!!
2
1
36
@hardmaru
I like thinking that my constraints force me to try unusual things, but it's unclear to me if unusual == better. 🤔
3
2
31
What an elegant idea!
My new paper is on arXiv, Object Discovery with a Copy-Pasting GAN
6
71
327
0
5
31
💜
We have reached an agreement in principle for Sam Altman to return to OpenAI as CEO with a new initial board of Bret Taylor (Chair), Larry Summers, and Adam D'Angelo.
We are collaborating to figure out the details. Thank you so much for your patience through this.
6K
13K
66K
0
0
31
Based on this excellent work from Engstrom, et. al.
A single robustly trained classifier is all you need to perform image generation, image-to-image translation, inpainting, ML-assisted paint (and more)! Blog post (w/ paper and notebooks): (w/
@ShibaniSan
@tsiprasd
B. Tran
@andrew_ilyas
@logan_engstrom
)
3
102
373
1
2
29
Had a fun weekend working with fellow
@__MLT__
member
@the_asir
on our Kuzushiji OCR Web Interface for the
@junctiontokyo
hackathon.
@tkasasagi
we should plug in KuroNet to this interface!
1
10
27
I find the most amazing thing about this is how quickly it converges. The original SPIRAL paper shows it converges in 50M steps (using multiple multi-CPU/GPU instances). The MNIST-trained agent I show here used 11K steps and took ~10 hours to train on a free
@GoogleColab
GPU.
2
2
26
@gwern
Yup. One thing I want to try for the future is move from brushstrokes to something like paint splatters maybe? Anyway here's a grayscale one I tried real quick. As for regular style transfer, I did try but the results were just okay, not good enough to replace the original yet
1
0
24
Nice day to rent a bicycle and cycle alongside the Han River in Seoul
3
0
23
Anyway, I have to state here that I was not the first to think of this. While planning this tweet thread, I decided to look at some citations of the World Models paper and saw . Apparently, I am some months late to the party 😅
1
1
24
I also trained a full color 15-stroke agent on CelebA. The outputs seem on par with SPIRAL (orig. version). Unlike SPIRAL, my agent doesn't make strokes that are fully covered by future strokes. I attribute this to the ease of credit assignment when full gradients are available.
1
2
24
@jeremyphoward
@github
Nice! I code for a living and wish I had this guide before starting my github blog. I blog like twice a year and everytime I do I have to relearn ruby and jekyll and all the little idiosyncrasies of a template I chose over a year ago.
1
3
22
My OpenAI fellowship has changed my life in so many positive ways! I encourage anyone interested to apply. It's a great way to get into ML, especially for folks from non-traditional backgrounds!
We're excited to announce the OpenAI Residency! This new program offers a pathway to a full-time role at OpenAI for researchers and engineers who don't currently focus on AI.
The program will focus on recruiting from underrepresented groups in technology.
40
110
579
0
1
20
First time in San Francisco! Cycling the Golden Gate Bridge was fun!
1
0
22
Just arrived in the mail. This is the first AI book I've ever purchased. Time to explore the other side of the debate.
#Bookofwhy
#causality
2
1
22
I like to think of these as loopholes in nature found by evolution, like RL finding bugs in simulations
Turns out spiders can fly using natural electric currents. Even now we're still discovering amazing designs by nature.
0
8
44
1
1
21
everyone is secretly this way, we just never realize it until we try
types STRONGLY reduce cognitive load when reading/writing code
I've turned into someone who prefers strongly-typed programming languages and am not sure if Google did this to me or if I've always secretly been this way
54
31
566
1
1
19
On theoretical guarantees:
"I don't care about theoretical guarantees. Show me the empirical results. Throw the chicken into the engine"
Presumably referencing the chicken gun used to fire chickens to test a jet engine's robustness
1
1
20
Code:
Demo:
1
5
20
Python was the perfect beginner programming language for me. I couldn't understand memory management in C/C++, and Java's static types intimidated me.
Thank you Guido, and congrats!
It's bittersweet: I'm leaving
@dropbox
, and am now retired. I've learned a lot during my time as an engineer here -- e.g. type annotations came from this experience -- and I'll miss working here.
301
2K
9K
0
1
16
Had fun today visiting the Hokusai exhibit at Mori Art Museum. It was so crowded I arrived 10 minutes before opening but still had to wait 30 minutes to get in. Worth it!
1
2
17
@citnaj
LOL this is amazing. Basically what they want is coauthorship of Deoldify for the hard, hard work of entering it in a contest and describing it (apparently wrongly) in a pdf.
0
0
16
@runwayml
makes it easy to do silly things, like turning yourself into a map.
Takes <10 clicks to run CycleGAN on your webcam input
0
4
17
Adversarially robust models are going to be big for machine learning art.
Particularly interested in its effect on neural style transfer. Will better high level features result in better outputs?
Robustness goes beyond security: Representations induced by robust networks can align much more closely with human perception, and enable simple feature visualization and manipulation. See: (w/
@logan_engstrom
@andrew_ilyas
@ShibaniSan
@tsiprasd
B. Tran)
0
49
181
0
2
17
Cool! Let the environment constraints dictate how things should move instead of manually encoding things yourself.
At
@EmbarkStudios
we're experimenting with a new physically-based animation system using reinforcement learning. No manual animation was involved making these robots move. Thanks to using physics, the 100 ton robot learns a heavier, more lumbering gait than the smaller robot.
6
88
494
0
3
14
🌸🌸🌸 sakura 🌸 sakura🌸 sakura🌸 sakura🌸 sakura🌸 sakura🌸 sakura 🌸 sakura🌸 sakura🌸 sakura🌸 sakura🌸 sakura🌸 sakura 🌸🌸🌸
0
0
16
Approach motivated by how a student might solve math problems: by setting up intermediate equations, using a calculator to solve them, and using those results to construct further equations.
1
1
17
@bradcog
@mat_kelcey
Honestly making a bay area salary while living somewhere else is a dream
1
2
16
The approach is simple. I train a world model that can predict the output stroke of a real paint program for a corresponding action, in effect creating a fully differentiable paint program. I then train an agent to paint using the same adversarial method detailed in SPIRAL.
1
0
16
I answered my own question
Adversarially robust models are going to be big for machine learning art.
Particularly interested in its effect on neural style transfer. Will better high level features result in better outputs?
0
2
17
0
2
16
On causality:
"I don't know what they're talking about. I don't understand it. All I know are MDPs. The world is an MDP."
1
0
16
@GoogleColab
For anyone who wants to know more about this project, I wrote a blog post detailing my motivations, approach, and future work in this direction I'm excited about.
1
2
15
@GaryMarcus
This will be a pretty cool benchmark for math problem solving. Maybe a natural upgrade for
0
3
15
This is the reason it's time to move on from Imagenet baselines
1
0
14
CelebA 30-stroke agent trained for 26k steps. About ~3 days on
@GoogleColab
. The agent has decided on an interesting representation of eyes.
2
2
14
@jm_alexia
It might still be useful to put it out there! There might be insights that others can build on. I write blog posts for things I don't think are "paper-worthy".
0
0
12
I concur with this blog post. Independent research is not a good experience. The psychological toll failures take on you (rejection, failed expts, getting scooped) are magnified N times without other people to share it with.
After getting published in ICLR as an Independent Researcher, I have received nearly 100 messages from others who are looking to do the same. So I wrote a blog post on why I decided to do it and my advice to others.
47
491
2K
0
1
14
@hardmaru
Seems like a perfect candidate for porting to the browser (maybe minus the glove embedding). The models are already in magenta.js.
Anyway, awesome job
@ljvmiranda921
@thinkdatasci
!
1
1
13
His top 5 most important problems in RL
1. Improving core RL algorithms. They should learn better.
2. State construction
3. Planning
He couldn't remember the other two 😁
2
0
14