Glen Berseth
@GlenBerseth
Followers
3K
Following
4K
Media
78
Statuses
745
Assistant Prof at @UMontreal @Mila_Quebec @MontrealRobots. Creating generalist problem-solving agents for the real world. he/him/il
Montreal, QB
Joined January 2016
I am teaching a class on #FoundationalModels for #robotics and Scaling #DeepRL algorithms. This class expands on last year's class and my generalist robotics policies tutorial and code. I plan to share the lectures and code assignments. Starting with the first lectures below.
17
114
706
Sometimes I receive the baffling comment that no one has shown deep reinforcement learning working on real-world systems. Please contribute work to to help showcase the great work out there and disprove this sentiment.
6
33
229
I am excited to announce that I will be starting as an assistant professor at @UMontrealDIRO in August 2021 and a member of @Mila_Quebec! It has been a wonderful journey. I have many people to thank, especially my mentors @svlevine, @Mvandepanne and Petros Faloutsos. (1/2).
23
13
226
I am looking for more students! If you are interested in the practical and theoretical challenges of creating generalist problem-solving robots, apply to @Mila_Quebec
7
43
161
This year I am recruiting new students in #Robotics and #MachineLearning for the Fall of 2024. If you are interested in working with me, submit your application via Mila's supervision application process. To find out more, go here:
5
34
124
A student with multiple research papers is not necessarily a fantastic student (because of this access to resources others didn't have). It is about evaluating potential, which is not easy to judge. Would others have achieved more if they had the same access?.
If you have multiple papers before you even began a PhD, it likely means you had access that others didn't. I wish more PhD programs would take a step back and stop this absurd practice of favoring multiple papers before someone even begins a training program.
8
6
103
Thrilled to announce the first annual Reinforcement Learning Conference @RL_Conference, which will be held at UMass Amherst August 9-12! RLC is the first strongly peer-reviewed RL venue with proceedings, and our call for papers is now available:
3
25
102
Stitching is a great property that allows RL agents to combine trajectory segments together to learn a policy better than the expert trajectory. How well can conditional supervised learning methods (decision transformers) perform stitching between trajectories in RL? Not at all!
1
7
94
#NeurIPS is almost over. A common theme amoung results I have seen is that imitation learning is doomed to succeed. Compression is easier than discovery.
1
4
85
We @MontrealRobots are looking for more students! If you are interested in any of deepRL, perception, robotics, learning in the real world, generalization in RL, realtime RL, offline RL, autonomous driving, consider applying to Dec 1 deadline.
3
22
82
Finally, a conference focused on the best area of ML, reinforcement learning!.
Thrilled to announce the first annual Reinforcement Learning Conference @RL_Conference, which will be held at UMass Amherst August 9-12!.RLC is the first strongly peer-reviewed RL venue with proceedings, and our call for papers is now available:
2
9
78
My lab is looking for more postdocs in the areas of machine learning and robotics. Consider joining the 1,000 other researchers at Mila and develop methods to get machine learning working on real-world systems. Apply here
1
23
77
Offline RL aims to learn high-value policies from static data. We find that using latent diffusion models to sample from complex multimodal distributions over long horizons enables better credit assignment, leading to improved performance. More details in the 🧵 below:
1
9
88
Which objective is best for encouraging emergent behaviour in RL agents? Honestly we don't know, it depends on the environment. To address this we design an algorithm that can decide which objective is best, measured by computing a type of authority over controlling entropy.
1
6
68
I will be teaching a course on robot learning, starting in January 2022! This course will focus on deep reinforcement learning methods and their application to robotics. You can find more info on the course here:.
1
9
58
Our recent work accepted to #ICLR2024 indicates that supervised learning methods for RL (e.g. decision transformers) are not capable of stitching trajectories together on their own. We propose a simple fix to improve DT and a new dataset for better stitching evaluation.
What if I told you that there were a combinatorial number of tasks that Decision Transformer like methods overlook? Excited to share our paper which unwraps this by linking trajectory stitching to combinatorial generalization, leading to a practical data augmentation method.
0
7
56
Congratulations to everyone who attended, presented and helped organize the Montreal Robotics Summer School at @Mila_Quebec! Looking forward to another awesome show next year.
1
7
55
Looking forward to announcing some exciting news for the RL community tomorrow…stay tuned! .#reinforcementlearning.
0
3
49
Is there really no deep reinforcement learning workshop at #NeurIPS this year?! This workshop has been a great place for deepRL researchers to gather since 2015. It was bad enough ICML had hardly any RL workshops.
9
3
49
Dealing with #Neurips2023 reviews right now feels like talking to a wall. Paper claim: Cures cancers A and C. Reviews critisisms: Yes, but have you tried a larger transformer?. #UnrealisticExpectations.
2
0
47
I am incredibly honoured to join this group of amazing AI researchers. I look forward to building the next generation of inteligent robots.
CIFAR is pleased to announce 5 researchers at @Mila_Quebec are our newest Canada CIFAR AI Chairs!. Congratulations to @GlenBerseth, @huangcza, Foutse Khomh @SWATLab, @BangLiuUdeM & @dhanya_sridhar. A thread about their leadership & work 🧵:.
8
2
47
Merci beaucoup for visiting @Mila_Quebec, showing off the new hardware #stellaquadruped and answering all of our (@MontrealRobots) questions.
You attended our presentation today @Mila_Quebec ? Let’s continue the conversation! Share your thoughts and tag us #stellaquadruped #robotics #research.
0
3
43
Very excited to announce the first keynote for @RL_Conference this year. @PeterStone_TX has been performing #reinforcementlearning research for decades. I am excited to hear his insights.
0
4
41
New work on fully autonomous real-world robotic learning with mobile manipulators. One of the exciting parts of this work is how we can enable robots to train in the real world without any human support or resets. website: Paper:
1
4
39
Very excited about kicking off the Montreal robotics summer school at @Mila_Quebec!.
Wishing everyone @Mila_Quebec an amazing 1st day at the Robotics Summer School 2022! .Exciting week of #robotics #learning and #coding Ahead. #quadrupedrobot #research #Universities #MachineLearning #AI #ArtificialIntelligence #DeepLearning #stellaquadruped #MilaRSS2022 #RSS2022
0
4
37
We (@marcgbellemare and I) are looking for a Postdoc to study risk-sensitive decision-making with applications to robots (real and simulated). We invite all candidates with experience in reinforcement learning and/or robotics to apply. Details:
0
11
34
What else can a good RL agent do? We conduct 100+ experiments with different RL algorithms, pre-training methods, and molecular grammars to search the vast space of molecular graphs, because discovering useful drugs can provide significant value.
1
7
37
The skill of being able to imitate others from just minimal visual observation is challenging to learn but an extremely beneficial skill most intelligent animals possess. We develop an agent that can learn to imitate an expert from only a single video demonstration—a 🧵.
1
5
30
#Neurips2023 was a huge and lots of fun. It was great catching up with everyone and seeing all the great research. I look forward to seeing everyone again at our more focused @RL_Conference.
1
1
33
I am incredibly honoured to join this group of amazing AI researchers. I look forward to building the next generation of intelligent robots.
Mila is proud to welcome five new Canada CIFAR AI Chairs among its faculty. These world-class researchers will receive long-term funding to support their research activities.
2
0
31
It is genuinely great to work at Mila. Being in the community at Mila provides an incredible space to support research and education. I look forward to learning much more in the coming years. #thankstomystudents @MontrealRobots.
@GlenBerseth Core Academic Member of Mila, explores how machine learning helps train more adaptable robots that could help humanity tackle its most pressing challenges. Watch the video:
0
4
30
I had a great time talking with @ahead_io while they were showing off their new robot Stanley at @UMontrealDIRO. I am looking forward to getting one of these for my research lab.
2
5
30
Getting robots and people to work together can be tricky. In our new work we enable human and robot interaction by developing a "task space" that allows a learning robot to quickly identify the users intentions and help the users reach their goals.
RL can help humans control robots: our new method, ASHA (ICRA '22) does this by pretraining autonomously to learn a "task space", then rapidly learning to map high-dim signals from a human interface into this task space to perform user-desired tasks: 🧵->
1
3
28
The great people at @wairobotics are organizing a hackathon for women (students) to learn about robotics. I am organizing a branch at Mila Nov 4-6. If you are interested, please sign up. #Robotics .
3
13
29
In this work we designed a robotic system that can learn autonomously without interventions from people (besides changing the batteries). I learned a lot from this project that takes steps towards robotic systems that solve real world tasks and collect large data themselves.
Some perspectives on autonomous lifelong RL for real-world robotics: This post, by @jendk3r, Charles Sun, @ColinearDevin & @GlenBerseth discusses how we built an RL system to train a robot to clean its room (or your room?). 👇
1
7
25
Our new work trains an agent that learns to discover and stabilize sources of uncertainty in partially-observed environments using intrinsic reward signals defined by its beliefs in a world model #NeurIPS2021 .arXiv:
Intrinsic motivation allows RL to find complex behaviors without hand-designed reward. What makes for a good objective? Information about the world can be translated into energy (or rather, work), so can an intrinsic objective accumulate information? That's the idea in IC2. A 🧵:
1
1
26
I am looking for more students! If you are interested in the practical and theoretical challenges of creating generalist problem-solving agents for the real world, apply to @Mila_Quebec.
1
1
26
I'll be arriving at NeurIPS on Wednesday for the @RL_Conference party at the holwing wolf! I am also looking for more #ReinforcementLearning students to join the lab so reach out if you want to chat!.
1
4
28
Congratulations again @agarwl_ ! Well done and love love RL. #weareonlydoingRL #reinforcementlearning #everythingisRL.
I had tons of fun doing my PhD jointly at Mila and Google Brain / DeepMind. Posting acknowledgements from my thesis below to express my gratitude.
0
1
27
Writing is a complex skill. It is also a vital skill to succeed in research and is a valuable skill in general. I created a writing workshop at Mila that focused on principles for writing clearly, correctly, and quickly. Content available here.
3
5
26
Want to know how meta-reinforcement learning can enable fast adaptation for continual reinforcement learning? Check out our new paper, Continual meta-policy search (CoMPS).
Can we meta-train on one task at a time, and accelerate learning of each new task? Continual meta-policy search (CoMPS) enables online meta-RL, so that each new task learns faster by learning to learn w @GlenBerseth @WilliamZhang365 G. Zhang @chelseabfinn
0
3
23
Are we using the best representations for OfflineRL? We found that using latent diffusion models work better at capturing the complex multi-modal distribution of Q-values of Offline RL datasets. More details at the #robotlearning workshop. Dec 16 Hall B2 10-11 and 14:45-16:15.
Offline RL aims to learn high-value policies from static data. We find that using latent diffusion models to sample from complex multimodal distributions over long horizons enables better credit assignment, leading to improved performance. More details in the 🧵 below:
0
2
26
It is because of issues like this in many current reviewing system that we at @RL_Conference created a processes that is focused on evaluating technical correctness, to avoid the issues of estimating potential future impact.
My most-cited, never-accepted, ArXiv-only paper has over 1880 citations. "Deep Convolutional Networks on Graph-Structured Data".Mikael Henaff, Joan Bruna, Yann LeCun.
0
4
21
One of the strongest reasons for submitting to @RL_Conference, amoung many, is much better reviewers from people that live for #reinforcementlearning.
First blog post about RLC covering (1) why we need a new RL conference and (2) details about how we're trying to fix reviewing for RL papers!.
0
5
23
@marcgbellemare, and I are looking for another postdoc. If you or someone you know has experience in the area of #deepreinforcementlearning and/or #robotlearning, consider applying here.
1
4
23
The time for new graduate student applications is fast approaching. Tip 1, In the SOP, focus on discussing research interests and why you will be passionate about them for many years. Also, don't repeat items from the CV. More tips are available here (.
0
3
24
RL training has high variance over final policy performance. Our new work @corl_conf #Agilerobotics workshop augments RL by learning to estimate where the agent is underperforming and forcing the agent to learn along this boundary. W/ @velythyl, @shoddy_robots, @duckietown_coo
0
4
23
Learning is the step before doing. The #DLRL Summer School aims to keep the community up to date on the most important topics for DL and RL. I am excited to present on large models and Offline RL this Friday @Mila_Quebec.
Today was the first day of @CIFAR_News's #DLRL Summer School 2023! Mila is beyond excited to host the program this year in beautiful #Montreal. We are proud to see the global scientific community gathered around deep learning and reinforcement learning this week.
1
7
23
The best classes have lots of questions and I use this slide in my class to encourage questions. For agents to learn they should perform interventions (ask questions) to improve their causal model. Best of luck, and happy learning to all of the #RL agents out there!
0
0
23
@pmddomingos We are curating a growing list of real-world applications of reinforcement learning, . Others are welcome to add examples via pull requests.
1
3
22
If you are interested in getting involved in next year's edition let me know!.
That's a wrap! Thank you to all our students/postdocs, employees and to our amazing partners @ahead_io and @nvidia who helped prepare and run the first edition of our Robotics Summer School, organized by Professor @GlenBerseth!
3
2
21
I am looking forward to working with great students and postdocs in machine learning and robotics! If you are interested in working with me please apply!.(2/2).
9
1
22
@UMontrealDIRO (my department) has 3 openings in machine learning. Strong candidates may also join @Mila_Quebec. More detailed are available here.
0
10
18
The lab is also looking for more postdocs in robotics. If you have experience in machine learning and robotics and are interested in making generalist robots and other embodied agents, you should apply!.
This year I am recruiting new students in #Robotics and #MachineLearning for the Fall of 2024. If you are interested in working with me, submit your application via Mila's supervision application process. To find out more, go here:
0
3
20
I am looking forward to having a chat about where machine learning and robotics jive well together. Also, message me to meet while I am visiting.
🗓️ Join us in-person or online for our next Robotics Institute seminar on August 25 at 3PM (EST) with @GlenBerseth from @UMontreal to learn more about solutions and new methods for overcoming limitations of robot learning systems 🤖 Learn more:
0
7
19
We are having a party to mingle more with the community! Come by and have a chat. To get in, there will be a RL skill-testing question. #reinforcementlearning #rlskills.
For folks @NeurIPSConf, we're having a social at Howling Wolf from 7 to 10 on Dec. 13th to celebrate the launch of RLC and get ideas and feedback from the community. Come by, hang out, talk about RL!.
0
1
20
🏆Quebec is a global force in artificial intelligence and ranks 7th among the world's most successful NATIONS, Canada is 4th. We are strong in talent and Gov strategy. For details, Tortoise Media (, PwC (. @Mila_Quebec @CIFAR_News.
0
4
17
Day 57 of being a professor: Responding to 1-on-1 emails in a timely manner has become unsustainable. Instead, I have taken to social media where I can communicate 1-to-n. Still, if you don't hear from me in 5 days send help.
0
0
18
On Fridays we at @MontrealRobots have a meeting called spaghetti throwing session where we talk about unclear research questions. This is an artist's interpretation of the meeting using. This is more accurate than I originally thought.
3
6
20
There are many real-world problems we can use reinforcement learning methods to solve. This video presents some of my recent research using RL for robotics and design.
1
0
19
Want to work with world-class researchers working on developing generalist learning robots? Consider applying to my lab's full-time open robotic and software developer position. For more details see
0
4
16
We are now accepting submissions for the @RL_Conference!.
The RLC submission page is now open! Get your submissions ready:.
0
7
17
RL really should not start from scratch all the time. Check out this workshop that is exploring where RL should start from.
Announcing the #ICLR2023 workshop on "Reincarnating Reinforcement Learning". Have you ever wondered why do we almost always train RL agents from scratch? Our @iclr_conf workshop instead focuses on reusing prior computational work in RL. See for details.
1
3
18
Many thanks to The Canada Foundation for Innovation (CFI)! This support will go towards further research in developing real world planning systems.
@GlenBerseth, Core Academic Member of Mila, was awarded funding by the Canada Foundation for Innovation to purchase robotics hardware, which will allow him to further his research on embodied machine learning. Read the announcement :
0
0
18
This new work finds a unique combination of representations to enable efficient finite horizon exploration. This exploration requires estimating the state history so the agent can properly determine what states are left to explore.
📢Come check out our poster on “How can we design an algorithm to learn policies that can search through the state space efficiently?” tomorrow at the Frontiers4LCD workshop #ICML2023. Paper: with Lucas Lehnert, @irinarish and @GlenBerseth.
1
4
15
Underactuated vs overactuated, closed-loop vs open loop, MDP vs POMDP, single vs multi-agent? What makes an environment rich enough for agents to develop human-level intelligence? Join the @EcoTheoryRL workshop tomorrow to find out!.
0
1
14
Very helpful suggestions. Generally the more detailed the input you give to your advisor the more detailed and helpful the output.
How to work with your advisor(s)?. Working effectively with your advisor is the no doubt the key to success of your research! However, junior grad students often don't have a clear idea on how to do so. Sharing some tips that I found useful. 👇.
0
0
14
Very excited about this continued collaboration! If you are interested in generalization or creating algorithms to develop new materials consider applying to my lab.
Intel strengthened its co-innovation collaboration with @Mila_Quebec by renewing a 3-year commitment to #ResponsibleAI. Together we'll develop new AI techniques for problems such as climate change, new materials discovery, and digital biology. Learn more.
0
4
14
The first day at @NeurIPS was exciting. Others that are interested in getting RL working on real world problems should reach out. Including potential students and postdocs.
3
0
14
Congratulations @pcastr! If you don't know Pablo, he does fantastic work and will be a great supervisor.
This year I became an adjunct prof @UMontreal and a core industry member at @Mila_Quebec , and am planning on taking students for fall 2024. Would love to have a diverse set of candidates from which to choose from, so please apply if interested, and spread the word! 👇🏾.
1
1
14
I am extremely excited about this opportunity and greatful for the contribution from the Courtois foundation. This contribution will allow my lab to explore more impactful research, such as learning how to automate scientific experiments with robots.
[Les acteurs au cœur de l’Institut Courtois]. Fraîchement débarqué de Californie, le professeur @GlenBerseth se consacre à sa passion: comprendre l'intelligence des machines… et faire en sorte que les robots fassent le ménage à la maison.
0
1
13
@Mila_Quebec Generally, I am looking for students interested in research on reinforcement learning (RL) including batch RL, multi-task RL, model-based RL, Sim2Real, hierarchical RL, generalization in RL, multi-Agent RL, meta-RL, imitation learning, and unsupervised RL (2/3).
1
0
12
@ashleyruba_phd Being a prof isn't for everyone. However, I would add that the academic culture in Canada is nicer than American.
0
0
11
In the finals for this year's Montreal Robotics Summer School. The competition is very competitive this year.
This week, Mila hosted its second Robotics Summer School, where participants took part in a final challenge where the goal was to help the robots navigate autonomously through an obstacle course.
0
0
13
Janarthanan is great. I have been collaborating with him on some solid RL research and I encourage students to apply to with with him.
Pleased to share that I will be joining @DalhousieU as an Assistant Professor in Jan '24. :) I have openings for Master’s & PhD positions in Deep Reinforcement Learning. Details: Please share with potential candidates and within your network. Thanks!
1
1
13
This is really great work lead by @ZhongyuLi4. I am still impress how much multi-task learning helps Cassie learn even more dynamic skills.
Our goal is to push the limits of the agility of legged robots. Using a single RL-based multi-task policy, a biped Cassie learns to perform various challenging jumps, including standing long jump, high jump, and/or multi-axes jumps, with significant robustness, in the real world.
0
0
12
Congratulations @MassCaccia on passing your defense today! Well done! Also félicitations @lcharlin and thank you @apsarathchandar and Bing Liu for being on the jury.
I'll be defending my thesis tmr 🛡️.Swing by if you want to learn about different flavors of continual learning!. Title: Ascending the Continual Learning Problem .Hierarchy: Towards General-Purpose, Autonomous Learning. Date: October 20th, 11:30 ET. Link:
1
1
10
Are you interested in how to train RL agents that can search more efficiently? Come check out our work at NeurIPS today to understand how predecessor and Successor reps can be used to learn non-Markovian exploration strategies.
Excited to be at #NeurIPS2023 this week, showcasing ηψ-Learning-- a method for training agents to navigate the state space optimally. w/ Lucas Lehnert, @irinarish, @GlenBerseth. Poster: Join us for a chat at Poster #1915 on Tue, Dec 12, 10:45 AM CST.
0
2
13
Most computer vision models are specific to image resultion. In our recent work we create a residual network block that can be integrated into common network designs for creating networks that can gracefully adapt image resolution based on Laplacian residuals and dropout.
2
3
12
Are you tired of tuning PD gains for each robot and change in the task? Using torque-based control eliminates these issues and helps reduce the sim2real gap because of torque control's natural compliance!. 1/n.
1
1
12
Congratulations Khimya! This just keeps getting better and I look forward to more amazing research thanks to you.
I am now an external affiliate member at @Mila_Quebec , and am looking to co-supervise 1-2 PhD students for fall 2024. Please apply if interested, and spread the word! #WiML #WomenInTech #STEM.
1
3
12
We could consider writing skills the unsupervised objective function to optimize before we select the task we want to fine-tune. I gave a workshop on technical writing that others can watch to help them write and present their ideas.
0
1
12
@AravSrinivas The bottleneck is not needing a VLM. Is there some logic that says once we have a VLM, we will also have a compositional generalization?.
0
0
11
The robots and students have had a long and busy week of preparing for the showdown today. I am looking forward to the results as we start early today!.
The first edition of our Robotics Summer School, organized by Professor @GlenBerseth in collaboration with our partners @ahead_io and @nvidia, is in full swing!. The ultimate challenge will take place tomorrow. Stay tuned!
0
1
11
TIL: The first deep imitation learning paper used on real cars was also published at the first #NeurIPS in 1988! How things have hardly changed :).
2
2
12
This is great competition. We need to work on hard problems like these to discover our current limitations.
We are excited to announce that the 2021 @NeurIPSConf MineRL BASALT Competition on Learning from Human Feedback has officially started! Sign up to participate here: . Keep reading to learn more about the competition! . 1/6
0
1
11
Want to know how you can train agents to imitate experts from a single video demonstration? Come chat we me today at the 2pm poster session at #ICML. Hint: the secret is how to train your model to give us better signal from sparse data. Talk here: .
1
0
11
In this new project we have learned how to incorporate vision to allow a quadruped to learn how to kick a football to target locations.
A fun project on training quadrupedal robots to score a goal: w/ Yandong Ji, @ZhongyuLi4, Yinan Sun, @xbpeng4, @GlenBerseth, Koushil Sreenath.
2
0
9
Want to learn about how to apply RL to more than just video games? We are presenting this work at AI4materials workshop at NeurIPS today on how to use RL for protein design.
What else can a good RL agent do? We conduct 100+ experiments with different RL algorithms, pre-training methods, and molecular grammars to search the vast space of molecular graphs, because discovering useful drugs can provide significant value.
0
0
10
Great points. Making a better sponge should also help us collect more data and explore more ways to make RL spongy.
Recent articles about RL such as and touch upon a concept that we found to be crucial to consider in robot learning:. Data sponge 🧽.🧵
1
0
11
@abhishekunique7 In similar ways to "simulation is doomed to succeed" is used to describe that we make simulations we can solve. When we collect data to imitate we already know how to solve the problem. The goal should be to discover useful skills we don't already know.
0
0
10
I am looking forward to @JasonMa2020's talk tomorrow!.
Hello! We have @JasonMa2020 from UPenn giving a talk at this week's robot learning seminar (Thursday 11:30am EST online). Hope to see you all there!.Title: Foundation Reward Models for General Robot Skill Acquisition. #Robotics #MachineLearning.
0
2
11
How can we train one policy that can solve the task without additional training, no matter the morphology/structure of the agent? Our (@brandontrabucco and @mphielipp) work learns how to accomplish this using transformers and even works if half of the robot's sensors are corrupt.
On the Mila Blog, PhD student at @CarnegieMellon @brandontrabucco shares his latest work with Mila Prof @GlenBerseth and Dr. Mariano Phielipp of @IntelAI on AnyMorph: a RL method that learns one policy to control many agent designs.
0
1
8
Very nice work! Showing that collecting more data in the real work is important for the best performance is important. Real interaction is the best.
We are thrilled to share our groundbreaking paper published today in @Nature: "Champion-Level Drone Racing using Deep Reinforcement Learning." We introduce "Swift," the first autonomous vision-based drone that beat human world champions in several fair head-to-head races! PDF
1
0
10
This work has been a great process of getting to the details of the best methods for Sim2Real transfer for biped robots. Truly impressive work by @ZhongyuLi4 and collaborators.
Interested in making your bipedal robots to be athletes? We summarized our RL work to create robust & adaptive controllers for general bipedal skills. 400m-dash, running over terrains/against perturbations, targeted jumping, compliant walking, not a problem for bipeds now.🧵👇
0
0
9
Today at @TechaideMTL 3 pm EST, I am giving a talk on my research plans to make generalist robotic scientists. This research is strengthened by the new Courtois Institute (a building designed to automate scientific experiments using robotics).
0
1
8
Thank you for the great presentation and lunch at @Mila_Quebec today. I lookforward the future of the group. This is a great program and women interested in #ai or #robotics should reach out.
🥗 🥤 🍕Come join us on July 25th at noon for lunch & networking with @wairobotics at @Mila_Quebec .#ai #robotics #womeninai #womeninrobotics.@GermanyInCanada @BaviereQuebec @ForumIAQuebec @QuebecScience @McGillRobotics @mcgillu @Concordia @etsmtl .
0
0
9
@EugeneVinitsky Is that good though? This separates the deep RL community into many smaller slices and takes some focus away from the main problems/goals. I saw this at #ICML. @yayitsamyzhang was saying we should make our own RL conference to help build the community.
3
0
9
We need an AI system that can take as input a CV and autofill the 6790 random boxes in a grant application that just wants the information in a different format. .
0
1
9