2022-03-18 11:46:39 RT @djstrouse: Unsupervised skill learning allows RL agents to learn useful behaviors in the absence of task-specific rewards. But beneath… 2022-03-16 00:39:20 RT @FrankRHutter: Announcing the late-breaking workshop track at #AutoML_Conf: paper submission by April 20, notification May 18 https://t.… 2022-03-12 08:11:00 CAFIAC FIX 2022-01-17 08:11:00 CAFIAC FIX 2022-01-12 21:12:40 @HolidayWrit I couldn't be entirely sure about that 2022-01-12 20:57:58 @HolidayWrit The poem is the only thing left of it 2022-01-12 20:55:16 @d_vasanthakumar Can be that all art starts out as noise and not needed at the time 2022-01-12 20:53:54 @DigantaMisra1 Nice one! 2022-01-12 20:53:10 @HolidayWrit Thanks! I like your interpretation 2022-01-12 20:51:57 @DecentGradient My neural network is generative 2022-01-12 12:10:28 @AnoopRKulkarni TIL, interesting! My dream was definitely not on that level, although maybe it's expressing a subconscious desire to go back to writing for fun again... 2022-01-12 11:58:27 @AnoopRKulkarni I don't get this reference, sorry 2022-01-12 11:04:15 Composed a poem in my dreams last night, woke up to see this written down, should I be concerned? My neural network is all I see Its weights majestic, biased beauty The delightful changes - there a 2, or is't 2.3? My neural network is all I see 2022-01-11 08:11:00 CAFIAC FIX 2022-01-05 23:29:15 @YFilmus Having been heavily involved in DEI efforts in this field for many years, I can confidently say you're incorrect. 2022-01-05 11:36:25 You’ve worked so hard to get here. You bring in something unique and valuable, and the hiring committee saw that. You absolutely belong. We need you. 2022-01-05 11:35:54 To sum up, there is no scalar value of “quality” when hiring. New perspectives are good. To people who think they’ve been offered a job they’re not qualified for because of perceived “positive discrimination”, I say: 2022-01-05 11:35:34 and that their worth in this field is not dependent on them looking like and doing the same things as those who’ve already succeeded. 2022-01-05 11:35:22 They decided that it was valuable to have new skills, perspectives, and experiences, so the makeup of who’s being hired has started to change. This is good and important, because it signals to newcomers in the field that bringing in new perspectives is valued, ... 2022-01-05 11:34:55 Hiring shouldn’t be about some kind of competition to see who can “AI the best”, and arguing over what percentage is more qualified. This is meaningless because hiring committees are all hiring for a multifaceted set of different reasons. 2022-01-05 11:34:42 This is especially true for AI, so that we can ensure that future technology benefits everyone, not just a small demographic. 2022-01-05 11:34:25 the “wisdom of crowds” depends on people being able to draw from a broad range of information. In other words, a candidate’s strength depends on them bringing in a relatively new perspective. 2022-01-05 11:34:02 (people’s relevant skills and knowledge fluctuate throughout their lives). It’s known that people tend to have unconscious bias toward those who are like them, which yields homogenous workforces if left unchecked. This is bad, because... 2022-01-05 11:33:03 However, it’s not possible to assign a single number to someone’s potential contributions, because their value to a company / department is multifaceted and dependent on who’s already there, and is not an intrinsic, constant quantity anyway 2022-01-05 11:32:41 It presumes that there is a single, constant, scalar value of “quality” that we can assign to all candidates, and that the goals of hiring should be to maximize this total “quality”, like some kind of score. This is where the “bar” comes from, ie a threshold on a scalar value. 2022-01-05 11:31:49 Ok, I have to say something. I’ve heard the argument many times that DEI efforts force us to “lower the bar” in order to be able to achieve diversity goals. This is always presented as being objectively logical (“it’s just numbers!”). Here’s why it’s wrong: 2021-12-27 08:20:00 CAFIAC FIX 2021-12-15 13:45:39 In fact, the central dichotomies of deep learning (train vs eval, weights vs activations) don't even make sense in biological systems, and could be holding us back, as far as continual learning and generalization are concerned. 2021-12-15 13:39:18 One of the main differences between natural and artificial intelligence is that biological systems are much more robust and able to adapt to changes in their environment. Could this be due to the fact that they're constantly self-organizing and growing? Interesting blog post! https://t.co/YXgpfvSXz0 2021-12-13 14:43:27 RT @neuro_data: Our @NeurIPSConf workshop "Out-of-distribution generalization and adaptation in natural and artificial intelligence" is hap… 2021-12-12 21:04:24 @hugo_larochelle @RaiaHadsell @kchonyc I have a feeling you might indeed soon need to worry about scaling! 2021-12-12 20:49:17 @hugo_larochelle @RaiaHadsell @kchonyc (sorry about all the questions, I'm really interested in this topic) Curious how you plan to be able to scale with the number of submissions? Will all authors be asked to review or suggest a reviewer? Will there be plans to train new reviewers or provide feedback about reviews? 2021-12-12 20:42:49 I'd already been feeling quite burnt out from the constant conference submission/review/org deadlines, so this couldn't have come at a better time! TMLR seems to have most of the changes I was hoping for, such as incentivizing reviewer quality and rolling submission. https://t.co/U2O0PYTsUI 2021-12-12 20:37:24 @hugo_larochelle @RaiaHadsell @kchonyc Nice! Will authors get a chance to revise and resubmit, with another round of reviews, like for a typical journal? Will there be desk rejects? 2021-12-12 20:14:32 @hugo_larochelle @RaiaHadsell @kchonyc Fantastic to see this! How do you plan to assign reviewers? Will it be entirely up to the action editors or area chairs? That is, no bidding? 2021-12-08 12:57:59 RT @djstrouse: Deep RL has had lots of success producing superhuman *opponents* in games, but how do we train great *teammates*? Check out… 2021-12-08 12:29:55 Super excited for this line of work led by @AndrewLampinen on how agents trained to explain tasks to themselves do much better in an abstract, relational (odd-one-out) RL task. https://t.co/HCr7IRZnli Click through for tweet thread that tells you why! https://t.co/8EEl7Grc3p https://t.co/aTHHukF5K2 2021-12-06 16:32:28 Excited to announce that Alchemy was accepted to #NeurIPS2021 Datasets/Benchmarks track, to be presented this Thursday. Wonderful colleagues and co-authors made this happen (unfortunately too many to name), so all due credit. https://t.co/KFjKtbzB3q Happy NeurIPS-ing! https://t.co/XxrxGOgmAJ https://t.co/AVnRSWU5JM 2021-12-06 13:12:39 RT @FrankRHutter: Exciting news: following a series of 8 AutoML workshops at ICML, we will hold the 1st International Conference on AutoML… 2021-11-22 14:33:41 Great opportunity to work on cognitive modeling and meta-reinforcement learning at CPI Lab! Co-supervised by Eric Schulz and myself, in collaboration with @zeynepakata and Matt Botvinick. Job ad here: https://t.co/uuyaCX4rNQ https://t.co/tbWo5rnPeo 2021-11-15 14:21:47 A very handy reference to have around on Bayesian modeling. I've always been impressed by people who can present complex topics in such clear, concise ways. Congrats to the authors! Draft here: https://t.co/lUjdyHPcbJ https://t.co/hSNc1a2srO 2021-11-06 23:20:00 CAFIAC FIX 2021-11-01 19:20:00 CAFIAC FIX 2021-08-19 18:11:46 RT @pgmid: What is learning, really? What counts as life? That, and @alonamarie, @janexwang, and @criticalneuro chat about algorithms, repr… 2021-08-12 12:31:32 RT @savvyRL: Earlier this summer Krystal and I are appointed as ICLR 2022 DEI chairs. Right away we started planning something tangible and… 2021-08-06 11:50:56 @csdashm @tamsioul That makes sense. Just thought I'd say something since they keep tagging me. 2021-08-05 08:34:04 @tamsioul Wrong Jane Wang, as far as I know I'm not involved in this 2021-07-22 17:56:56 You can play around here! https://t.co/nkdt9Xblk1 2021-07-22 17:51:46 The human proteome has been released, with 100 million more protein structure predictions to come, data openly available to the scientific community. Honestly in awe of the work that went into this, congrats to the AlphaFold team! https://t.co/upZWrAN9fm 2021-07-19 09:33:49 RT @WiMLworkshop: 1/5 Call for Nominations : #WiML is looking for nominations for our Senior Advisory Board. Please send nominations to… 2021-07-15 18:56:40 RT @demishassabis: Last year we presented #AlphaFold v2 which predicts 3D structures of proteins down to atomic accuracy. Today we’re proud… 2021-06-22 11:16:38 What are the inductive biases for evolution? And how did those come about? Was the descent from meta-learning to meta-physics always inevitable? I ponder on my lunch break, after having searched yet again for how to sample from np.random.random.randint.sample_choice or whatever https://t.co/b3w8Ne5cOZ 2021-05-20 18:46:51 @neuro_data @KordingLab @AToliasLab @behrenstimb It seems like any learning that leads to generalization will give you this effect? Task 1 and task 2 should presumably be related, and I'd guess learning on task 102 is likely to give you better backward transfer than task 2 2021-05-11 10:06:53 RT @NandoDF: In this tutorial I contrast (1) mimicking, aka behaviour cloning, and (2) selective imitation via offline reinforcement learni… 2021-05-07 15:16:22 RT @_kainoa_: Join us at the Learning-to-Learn workshop starting at 7AM PT today! How to join: https://t.co/UaycnNGhWC Invited talks + pa… 2021-04-23 22:57:34 @flennerhag @adamianou @andreialexrusu @rpascanu @joavanschoren Huge congrats!! It was/is awesome working with you! 2021-04-21 22:39:07 Apply for Neuromatch Academy on deep learning this summer! Looking forward to learning a lot from my fellow lecturers https://t.co/W6Gr3KAErm 2021-04-21 22:34:31 @KordingLab @neuromatch Thanks for inviting me! 2021-04-13 16:37:24 @awjuliani Thank you! And that's a good point about our documentation, we'll get that updated :) 2021-04-13 09:07:28 @awjuliani However I fully agree that there's a need for meta-RL environments that push agents to do better experimentation and hypothesis testing, and JRPGs seem like a great direction for this! 2021-04-13 09:04:15 @awjuliani It also comes with a Bayes ideal observer/actor, which allows for better interpretability and analysis. Such a comparison doesn't exist with more complicated environments like the ones you mentioned 2021-04-13 09:01:27 @awjuliani This is a neat benchmark, and thanks for calling out Alchemy in your blog post! However I just want to point out that Alchemy also comes with a symbolic version, which is much less computationally expensive to run. 2021-04-11 14:39:23 RT @sindero: @mmitchell_ai Umm. Right here. Significant contribs to DL since BEFORE it was DL. Notable R& https:/… 2021-04-07 12:11:31 @neuroecology Sorry to hear that you've been going through this. But having been in industry research for a while, I can say that overall I'm glad I left academia. Happy to discuss more if that'll help! 2021-04-06 20:08:00 @yael_niv So true! It's probably a positive feedback loop, where the more you bring a diverse group of people together, the stronger that community becomes, and the more it attracts and is appealing to new researchers from different backgrounds 2021-04-06 19:54:16 @ClopathLab Doina Precup Maybe this is just a (non-comprehensive) list of women scientists I admire :) 2021-04-06 19:51:35 Can't forget @luisa_zintgraf Linda Smith @RaiaHadsell @neuro_kim 2021-04-06 19:38:58 @itsharveen @OxfordPsychSoc Great idea! Started a list here: https://t.co/WdwqYIRHyt 2021-04-06 19:37:41 @criticalneuro Thank you Ida for speaking and bringing such interesting points of discussion! 2021-04-06 19:35:12 Tagging some of these aforementioned awesome speakers (please add on more, I'm sure I'm forgetting people): @ermgrant @chelseabfinn Ishita Dasgupta @AlisonGopnik @yael_niv @criticalneuro @drfeifei Kate Rakelly @FeryalMP @rosemary_ke Martha White @EmmaBrunskill Cate Hartley 2021-04-06 19:20:56 Maybe I just think everything is meta-learning 2021-04-06 19:17:43 Recently I organised a small group workshop on meta-learning and cogsci, and was really pleased to have so many awesome women speakers to choose from. Interesting that certain subfields tend to be more diverse, like meta-learning and NLP. Or do I just have a selection bias? 2021-04-02 14:29:07 RT @jhamrick: Model-based planning is often thought to be necessary for deep reasoning & 2021-03-31 21:07:15 This was a fun interview - thanks @hunter_abigail and @ldvcapital! https://t.co/AHOfIOcrXI 2021-03-31 17:00:52 @cryptexis @KordingLab @neuromatch Deep learning is just obfuscated nonlinear regression, change my mind :) 2021-03-31 15:13:20 @KordingLab @neuromatch What about how to deal with datasets? Cleaning, feature extraction, scaling, etc. This can also include data augmentation methods. 2021-03-27 22:48:10 @wojczarnecki It could be a system where you get feedback from authors, other reviewers, and AC. ACs can see this past feedback when assigning reviewers (in anonymous form), but not the public. 2021-03-27 16:19:10 I'm curious. For those who review for ML conferences: how do you think you would review differently if the quality of your reviews is tracked over time (while maintaining anonymity)? Would you prefer this system? https://t.co/vXSfmhQA6g 2021-03-27 16:17:03 @peter_richtarik @CsabaSzepesvari This is why we need reviewer accountability. Those bad reviews will follow the reviewer as well 2021-03-27 16:07:57 @CsabaSzepesvari @peter_richtarik It also seems like there's a lot of wasted reviewing effort among the conferences. Can we encourage better tracking of papers that get resubmitted, along with reviewer accountability? 2021-03-27 16:05:59 @CsabaSzepesvari @peter_richtarik We need mentorship in reviewing. For the last NeurIPS Meta-learning workshop we tried this out, it led to much better reviews but did take more time/commitment from the mentors. Happy to chat more if you're interested 2021-03-25 17:01:34 Clearly my next step is to apply neuroscience-inspired AI to string theory 2021-03-25 17:00:16 Real "physicist solving other field's problems" vibe here. I say this as a former physicist turned neuroscientist turned AI researcher https://t.co/nKIFYwKeNu 2021-03-25 16:50:50 @TheStalwart Have they tried turning off the Suez canal and turning it back on again? Idk usually works for me... 2021-03-21 20:24:34 RT @janexwang: Do you need financial assistance to participate in #ICLR2021? Applications are now open for free conference registration! St… 2021-03-20 13:55:53 Rest in peace Sebastian 2021-03-20 13:55:30 My first time experiencing the grief of pet loss. I didn't think it would hurt this much. This past year has been a terrible time for so many - my heart goes out to all of you who have lost someone or something irreplaceable. Hold your loved ones close. They are everything. https://t.co/3QuSBUEWzb 2021-03-18 17:47:57 @ldvcapital @hunter_abigail Thanks for having me, it was a fun discussion! 2021-03-09 00:43:16 RT @ldvcapital: On March 18th, we’ll host our very first live interview in the #WomenLeadingVisualTech series! @hunter_abigail will speak w… 2021-03-09 00:42:39 RT @KordingLab: You learned deep learning and are ready to help others learn it too? @neuromatch may run a large super didactic DL summer s… 2021-03-01 16:34:25 RT @WiMLworkshop: You can now apply to be an organizer for #WiML 2021 events at #NeurIPS2021, #ICML2021 and other conferences! Apply using… 2021-02-28 11:42:17 RT @jakhmack: E Simoncelli @criticalneuro @janexwang @neuro_kim @KordingLab @MatthiasBethge @MihaelaVDS O Schwartz P Dayan T Sejnowksi @v… 2021-02-28 11:42:09 RT @jakhmack: Champalimaud Research Symposium #CRS21 on #neural and #artificialintelligence-- will certainly happen in October, hopefully n… 2021-02-25 15:17:04 @aaossa Thank you for letting us know, this is not intended functionality and we've fixed the issue. Please let us know if it persists! 2021-02-24 17:05:43 @criticalneuro @neuroecology I miss it too! Cosyne in Colorado was the last time I traveled, back when the world was semi-normal still... 2021-02-23 12:55:05 Deadline April 1 2021-02-23 12:50:30 Do you need financial assistance to participate in #ICLR2021? Applications are now open for free conference registration! Students and people from diverse backgrounds are especially encouraged to apply: https://t.co/GjvHh3I6i4 2021-02-11 17:52:59 Also I should call out the wonderful collaboration we had with the DeepMind Worlds team, who were amazing and without whom this wouldn't have been possible! 2021-02-11 17:49:26 @ampanmdagaba @criticalneuro Sustainability is key! Different people have different optimal work styles, and that's okay, but most important is that we allow ourselves time to recharge and not feel constantly drained. 2021-02-11 16:31:00 We’re in the middle of a pandemic in which women and people of color have been disproportionately affected, how is this helpful right now? Please everyone, prioritize your mental health and rest. https://t.co/lbhNc4UCw9 2021-02-10 14:03:05 @egrefen @DeepMind @NicolasPorcel06 @bigblueboo @MaryKCassin @demishassabis @MinqiJiang Thanks for pointing out this paper Ed! It does seem relevant, but we were less focused on text-based/NLP based environments and more on creating an alternative meta-RL benchmark to e.g. MetaWorld, Retro, etc, which we do cite, although not in an explicit related works section. 2021-02-10 13:02:43 The fruition of a lot of work over the past year, a new benchmark task for meta-reinforcement learning: Alchemy! Train agents or play it yourself, this environment comes with interesting latent structure that encourages experimentation and structure learning https://t.co/2B7qz2Ll6t 2021-02-03 14:56:58 RT @joavanschoren: Today, @iddo and me are presenting a tutorial on meta-learning at #AAAI2021. For those who can't make it, we made all ma… 2021-02-02 09:39:38 @dougdegroot It works for me. Maybe try in a different browser? 2021-02-01 17:43:55 For more, I also discussed these topics recently at the Learning Salon https://t.co/NDhYjWV3yB with @criticalneuro @neuro_data @blamlab 2021-02-01 17:39:46 Meta-learning is a hot topic in AI, but how does it look in nature? I wrote a short review for Current Opinion in Behavioral Sciences that tries explores this question, hope you enjoy! Meta-learning in natural and artificial intelligence: https://t.co/68XHZ6Sw2t https://t.co/UzPq7aVdsr 2021-01-24 11:38:12 @neuro_data @beenwrekt @JohnCLangford @criticalneuro Not a cheat sheet, but many consider this to be the definitive book on RL, and a draft is free online: https://t.co/IzGHuqslHZ 2021-01-23 22:54:05 @neuro_data Thank you so much for organizing, it was a lot of fun! 2021-01-22 20:34:26 Happening in 30 min! Come join in the discussion :) https://t.co/ajJdTAjXTe 2021-01-22 16:25:33 RT @WiMLworkshop: *** WiML is hiring a full-time Operations Administrator *** The ideal candidate will be passionate about our mission and… 2021-01-20 19:57:16 What's the opposite of existential dread? 2021-01-19 14:30:50 Great listen! Paul's podcast is highly recommended. If you're curious about what neuroscience has to offer AI, you might also be interested in a recent NeurIPS tutorial I gave with @kevinjmiller10 and @AdamMarblestone on this topic https://t.co/7nwhdsxioM https://t.co/CttXDRP8fZ 2021-01-18 01:06:54 @neuro_data I meant more at the single agent/organism level, but I haven't thought about it in terms of an entire ecosystem, interesting! 2021-01-15 21:05:04 @neuro_data @criticalneuro @blamlab Looking forward to it! Although no promises for how awake I'll be, it's pretty late on a Friday night for me :) 2021-01-15 21:03:37 Something I've been thinking about lately: what does meta-learning actually look like in nature and neuroscience? https://t.co/65IBCnVFkX 2021-01-14 00:57:17 Interesting thread. Ultimately we need to rework how we communicate online though - social media and their incentive structures will always lead to this dilemma between ensuring responsible dialogue and giving people what they want - virality https://t.co/v6Vw41BUUW 2021-01-13 22:50:52 @solvay_1927 @arimorcos Ouch! 2021-01-13 15:26:07 @evgeniyzhe No, as in have a separate, rolling, centralized review process to assess the technical merits and accuracy, and then once papers have gone through this process, can be submitted to conferences, who will decide on importance / fit and whether to accept. 2021-01-13 15:21:09 Overall feedback was very positive, we wound up with excellent reviews, and it's clear that new reviewers learned a lot from their mentors and appreciated the opportunity to review under guidance. 2021-01-13 15:19:05 By the way, if any workshop organizers are interested in setting up a mentoring review system for ICLR/ICML, please let me know, would be happy to advise on our experience with the meta-learning workshop, which IMO was a great success! https://t.co/4ZyoKr98kT 2021-01-13 15:14:21 @gershbrain @dwf Also important: the ability to track / rate reviewers so that they have incentive to give better feedback. 2021-01-13 15:12:49 @gershbrain I had an idea a while back to have a collaborative, centralized review system that worked exactly this way, but there are actually a lot of practical and technical hurdles that have to be overcome, not to mention needing buy-in from at least 3 major conferences 2021-01-13 10:39:03 @FelixHill84 Yes true! We need better accountability for reviewers overall, no matter how well established they are. 2021-01-13 10:35:59 And also incorporate mentoring for junior reviewers, like we did for the Meta-learning workshop at NeurIPS https://t.co/QDxi1BzNKN 2021-01-13 10:31:05 Unfairly rejected papers don't necessarily make it in next time, I've had several papers in the past year go through multiple rounds of awful reviews, even when attempting to incorporate reviewer comments. IMO we need to separate the reviewing process from the conference cycles https://t.co/pyFzAdt6qN 2021-01-11 15:43:31 @MarlosCMachado @DeepMind Welcome! 2021-01-07 01:19:31 RT @BarackObama: Here’s my statement on today’s violence at the Capitol. https://t.co/jLCKo2D1Ya 2021-01-06 23:02:16 These people look terrified. This is happening in the heart of the capitol of the United States of America. https://t.co/ZaqHAO2s0c 2021-01-06 20:41:24 Whoa, what is happening in the capitol? Rioting and someone has been shot, but the DoD won't send in the national guard? https://t.co/grDwdfkTOA 2021-01-06 20:38:00 @Sci_Phile At some point, we all have to contend with reality 2020-12-29 13:31:18 @gershbrain yes, in the event that one is copied and the original destroyed, star trek transporter style, how should one report this on the NIH annual research performance progress report? 2020-12-13 22:15:18 @Rassh_RAJ https://t.co/6IUuBmhCcq 2020-12-13 16:12:50 @SarahHein0 I don't either in my personal life, but if you want to address me in a professional capacity, it's either Jane or Dr. Wang, although I will never stand for Mrs. Husband's Name in any situation (also never changed my name) 2020-12-13 14:10:52 For context: Wall Street Journal opinion piece suggests Dr. Biden should drop "Dr" https://t.co/ch2T58sv24 2020-12-13 11:53:26 @meetnavpk This is really beside the point. 2020-12-13 09:42:18 I don't usually ask people to call me "Dr" because I don't like the formality, but make no mistake. As with anyone who has earned a doctorate, my official designation is Dr. Wang. Not Mrs, not Ms, not Miss. I earned it. It means a lot to me. I expect you to respect that. 2020-12-12 00:04:20 @FrankRHutter @joavanschoren @ermgrant @FrancescoVisin @tmh31 @LouisKirschAI @luisa_zintgraf @KuanFang @jeffclune Yes so glad that's finally settled! 2020-12-11 20:38:52 @joavanschoren @drfeifei and @schwarzjn_ ! 2020-12-11 18:54:06 RT @ermgrant: The Meta-learning Workshop panel starts in < 2020-12-11 13:10:34 RT @AndrewLampinen: Excited that our work on "Transforming task representations to perform novel tasks" is now published online @PNASNews!… 2020-12-11 11:59:48 RT @LouisKirschAI: Excited about my first invited talk @ Meta Learn NeurIPS 2020 on #MetaLearning general-purpose Learning Algorithms. Inc… 2020-12-10 21:03:11 Meta-Learning workshop at #NeurIPS2020 is happening tomorrow! With invited speakers Frank Hutter, Luisa Zintgraf, Timothy Hospedales, Louis Kirsch, Fei-Fei Li, and Kate Rakelly https://t.co/wd1sncNqaF Please submit your questions for the panel here: https://t.co/MgImrizMRj 2020-12-09 19:50:39 Looking forward to participating in the @WiMLworkshop Mentorship Roundtables in just under an hour at #NeurIPS2020, come find me at table 46 in https://t.co/xnLLmLumAp! I'll be available to chat about non-traditional paths to machine learning 2020-12-09 01:21:49 @ajaysub110 @NeurIPSConf @AdamMarblestone Glad you enjoyed, it was fun to present! 2020-12-09 01:20:30 @SaberaTalukder I guess the hope is that something inspired by biology will someday be found to work better, by offering a new type of solution! 2020-12-07 10:07:51 @neuro_rish @tweetsatpreet Glad you enjoyed it! 2020-12-07 10:05:29 @Nickols53622834 @Plinz @kevinjmiller10 @AdamMarblestone You can register for the conference here: https://t.co/WXohQ2EEQz but the slides are also publicly available: https://t.co/GrUtmQbRRI… 2020-12-06 17:09:34 RT @WiMLworkshop: Dr. Timnit Gebru (@timnitGebru)’s technical contributions to our field have been hugely impactful & 2020-12-06 13:17:38 In case you can't make it (Monday Dec 7, 13:30 GMT), the video for our neuroscience/AI tutorial is now up: https://t.co/GrUtmQbRRI Please submit questions for us here: https://t.co/Qho9OT1GYb Or just respond to this tweet and I'll try to cover it in our Q& 2020-12-05 23:21:58 Have I been pronouncing "convolutional neural networks" wrong this whole time? 2020-12-05 23:18:49 Editing the captions for our #NeurIPS2020 tutorial on neuroscience and AI, which I know is important to do, but sad that now we will miss out on gems such as "excited Tory neurons" and "DeParle" systems... https://t.co/sBPUeFLDum 2020-12-03 12:38:33 RT @EmtiyazKhan: Happy to introduce https://t.co/vJkzSp60BE, a portal to easily connect mentors and mentees. Designed and created by @hen_s… 2020-12-01 12:47:52 RT @ChrisWLynn: Just out in @PNASNews. We review the exciting young field of graph learning. How do people learn and represent networks in… 2020-11-30 19:03:56 @mengyer I usually explain it as "any learning procedure that learns faster with more data, via either learned inductive biases or any other priors that enable faster learning on new tasks" 2020-11-30 18:01:40 The release in Nature about #AlphaFold has some amazing quotes: "The model from group 427 gave us our structure in half an hour, after we had spent a decade trying everything" https://t.co/OHQHvJFoDs 2020-11-30 17:45:53 RT @MoAlQuraishi: CASP14 #s just came out and they’re astounding—DeepMind looks to have solved protein structure prediction. Median GDT_TS… 2020-11-30 17:03:53 I don't know much about protein folding, but this seems like an amazing achievement - huge congrats to the AlphaFold team! https://t.co/6iTkHgJP6A 2020-11-30 10:42:02 Really looking forward to presenting a tutorial with @kevinjmiller10 and @AdamMarblestone at #neurips2020 on Where Neuroscience meets AI, Dec 7! https://t.co/7nwhdsxioM Please submit your questions for the live Q& 2020-11-25 20:36:17 @hugospiers Canceling plans? Greatest feeling ever as an adult 2020-11-24 13:42:39 Excited to read this! How the brain represents event structure to perform efficient inference, from @preston_lab https://t.co/X7YgLHqsM2 2020-11-24 13:38:18 RT @ProfData: Introducing the embedding space you didn't know you needed: Human similarity judgments for the entire ImageNet (50k images) v… 2020-11-24 13:35:30 RT @svpino: A great way to start with Deep Learning: https://t.co/dFuLyQQgYw A bunch of images of clothing A little bit of code to… 2020-11-22 12:56:59 @anne_churchland W is low for both first and last authors. Guess I should exit science 2020-11-19 12:35:57 @criticalneuro @DaniSBassett Exactly! And then to suggest that students should avoid female mentors is just irresponsible. I get angrier the more I think about it. 2020-11-19 10:43:12 This is how institutional biases get created and perpetuated. Really disappointed this got published @NatureComms https://t.co/hHTF7iy50Y 2020-11-18 13:25:39 RT @ELLISforEurope: The countdown is on! You have until December 1 to submit an application to the ELLIS Phd Program. Check out our video a… 2020-11-16 11:14:34 RT @summerfieldlab: "If deep learning is the answer, what is the question?" Very happy that our review with @SaxeLab and @steph_nelli just… 2020-11-14 21:06:36 @AmazonHelp @_chrisjones_ Clearly this is actually Amazon Help Total Landscaping 2020-11-13 10:54:40 @neuroecology Some valuable skills I wish I had learned early on: - how to write and maintain open source code - how to communicate ideas to a wider, lay audience (blogs and such) - how to manage large-scale projects and interface with different teams 2020-11-13 10:49:44 @neuroecology IMO there needs to be more training for non-PI positions, either during postdoc or grad school. There simply aren't enough TT positions to go around, and in a postdoc you have the opportunity to learn skills that would also be valuable outside of academia 2020-11-12 00:55:04 RT @jcrwhittington: Hello! Very happy to share that The Tolman-Eichenbaum Machine (TEM) has been published. Many thanks to @behrenstimb and… 2020-11-09 08:29:51 @neuro_data @bradpwyble @criticalneuro @blamlab @rajankdr @ZennaTavares Thanks for having me, it was so much fun! And sorry about the technical glitches, although as I understand they were entertaining :) 2020-11-07 23:57:36 RT @svlevine: New lectures released for the deep RL class: special lecture on open topics in algorithms by Aviral Kumer (#17), variational… 2020-11-07 17:56:15 Historic 2020-11-06 14:22:08 Welcome POTUS 46, it's so great to have you https://t.co/uKjlSq9hJi 2020-11-05 20:42:55 Nice read! https://t.co/18zu1DBof4 2020-11-05 08:47:38 Some background https://t.co/CPw0dieUa2 2020-11-05 08:46:41 Fascinating, apparently there's nothing in the rules preventing the VP from recognizing the minority senate leader instead of the majority, it's only precedent. So Kamala could call on Schumer first and break all the deadlock next session https://t.co/1xzwrAOsnd 2020-11-04 22:49:15 won, and of course, claimed for electoral purposes 2020-11-04 22:47:45 and in the end, haven't we all really won Pennsylvania, in our hearts? #IHaveWonPennsylvania 2020-11-04 17:38:12 @TechRonic9876 This doesn't explain why the numbers would be especially bad for Dem strongholds 2020-11-04 15:55:12 This seems like it's not getting enough attention... https://t.co/roVtixxSr0 2020-11-03 17:55:36 @neuroecology @Annika_Barber So accurate it hurts me 2020-11-03 17:53:02 @neuroecology @Annika_Barber It kinda worked! I read like 4 of them. Then I declared paper bankruptcy and saved the rest to disk 2020-11-03 17:41:33 I actually kind of love that we're all, collectively as a country, just a hot nervous mess right now. Nice to be a part of something. #VoteResponsibly all 2020-11-03 14:38:01 RT @TonyZador: Are you an AI researcher? Do you believe that neuroscience still has something to teach AI? If so, consider spending 2 y… 2020-11-02 19:22:33 I can't even tell what's election anxiety and what's covid anxiety anymore (forget about normal everyday anxiety), how's everyone else doing? also, please #VOTE #ElectionDay 2020-10-29 12:04:30 2020 is the year I realized that I won't have succeeded at life until one of my science papers gets retweeted by @MCHammer https://t.co/8b9xBL7A8x 2020-10-28 16:23:31 RT @AdaptiveAgents: Excited this work is finally out! Also, try the Colab tutorial, it's got a mini causal reasoning engine in it! #causali… 2020-10-27 19:47:57 RT @AriBerman: Very important: today is deadline USPS recommends for sending back your mail ballot to make sure it’s counted. Better yet, d… 2020-10-23 21:58:35 https://t.co/NbjgOmCEie 2020-10-22 21:26:56 "Neuroscience without behavior is a lost cause" Nicely said! https://t.co/WSJtwn5Q0F 2020-10-21 15:18:59 @neuroecology Or use a Twitter thread to add that touch of public humiliation: https://t.co/od29Jutw6Z 2020-10-16 12:06:03 "I know how tough it is some days to look with hope and confidence on the months and years ahead. But I would like to tell you what I often told you when you were much younger. I like you just the way you are." I needed to hear this today. What a legend. https://t.co/BANyqiwY9U 2020-10-13 15:15:11 We find that we’re able to outperform standard agents on hard exploration problems and provide analytical grounding for these results. 2020-10-13 15:14:54 We propose to estimate it via a distribution of temporal difference errors and treat exploration as a separate learning problem. 2020-10-13 15:14:28 An agent's uncertainty about the optimal policy can be a good signal for exploration, but estimating this uncertainty is often challenging and biased. 2020-10-13 15:14:00 New work out by the super talented @flennerhag and colleagues on exploration: Temporal Difference Uncertainties as a Signal for Exploration https://t.co/mfSu1VbQys w/@pablosprechmann, @FrancescoVisin, A. Galashov, S. Kapturowski, D. Borsa, N. Heess, @andre_s_barreto, R. Pascanu https://t.co/AsEri4ynka 2020-10-10 22:21:18 Happy #WorldMentalHealthDay, friends. I hope you're keeping safe, and taking care. 2020-10-10 22:01:03 Interesting tidbit after having read this: probabilistic inference can be thought of as serial computation across a sequence of agents, each of which gather more information from the environment. Distributed Bayesian inference! Downside: +2 net added to my tabs 2020-10-09 23:27:14 @iandanforth it also turns out to be an easy way to replicate tabs from one computer to another, so they can follow me everywhere 2020-10-09 17:57:06 Amazing to see this! Steps in the right direction https://t.co/yG4W6gmPM9 2020-10-09 17:26:00 @dosco that looks really cool! how do people collaborate and share notes? does each paper have a discussion section? 2020-10-09 09:27:53 @HannesStaerk Good idea! This thread has motivated me :) 2020-10-09 09:26:53 @schrepfler That's a lot of RAM... 2020-10-09 09:26:12 @isalirezag Wow! 2020-10-09 09:24:44 @ashimiblessing This is what leads to the much related reading group proliferation problem 2020-10-09 09:23:56 @barneyp Indeed! For me they seem to be papers that seem very relevant to my general interests but not to immediate projects, so perhaps good for long-term research planning? 2020-10-08 23:59:58 @tw_killian Writing a paper definitely increases the tabload many-fold, but at least it's very satisfying to close them all when you're done :) 2020-10-08 23:52:21 7. Facts as Experts: Adaptable and Interpretable Neural Memory over Symbolic Knowledge https://t.co/Gx05aJx3VI 2020-10-08 23:49:49 I'd actually be curious to hear what's in everyone else's Chrome tabs of "to-read"s (not that I need any more!) https://t.co/E8GVpuFzG9 2020-10-08 23:45:29 6. Action and Perception as Divergence Minimization https://t.co/s1SQgK8S3I 2020-10-08 23:44:31 5. Monte-Carlo tree search as regularized policy optimization https://t.co/UH1EgIXGDr 2020-10-08 23:43:43 4. Language as a Cognitive Tool to Imagine Goals in Curiosity-Driven Exploration https://t.co/7CMwgPLcdP 2020-10-08 23:42:19 3. Hopfield Networks is All You Need https://t.co/rseutfAK3y 2020-10-08 23:41:06 2. Understanding Human Intelligence through Human Limitations https://t.co/FIKPxC5fU4 2020-10-08 23:40:18 I have 35+ papers open in Chrome tabs that I've been meaning to read. Perhaps public accountability is the key. So here is a thread of a few of them: 1. Beyond dichotomies in reinforcement learning https://t.co/DleRwYy9ia 2020-10-07 14:03:14 RT @NobelPrize: BREAKING NEWS: The 2020 #NobelPrize in Chemistry has been awarded to Emmanuelle Charpentier and Jennifer A. Doudna “for th… 2020-10-06 15:33:27 RT @WorldAndScience: 'Your assumptions are your windows on the world. Scrub them off every once in a while, or the light won't come in' - I… 2020-10-05 16:00:34 RT @RevDrBarber: Today is the LAST DAY to register to vote in these 13 states: Arizona Arkansas Florida Georgia Hawaii Indiana… 2020-10-05 11:46:40 RT @IasonGabriel: The print-ready version of 'Artificial Intelligence, Values and Alignment' is now freely available from Minds and Machine… 2020-10-04 22:43:59 @scychan_brains @DeepMind @FelixHill84 @AndrewLampinen Fantastic, welcome! 2020-10-04 17:38:48 RT @FeryalMP: We have extended the deadline to **Wednesday October 7th**. We can't wait to recieve your submissions! :) https://t.co/jEGwkd… 2020-10-03 19:58:35 Interesting. I’m surprised at the institutional bias on the part of ACs, since if I recall they can’t see author names? But the gender gap seems to be explained by the fact that Bayesian papers have low acceptance rate, and they tend to have female authors. https://t.co/NIg0vxIS7k 2020-10-03 15:00:17 Deadline for the workshop on Meta-learning has been extended to Oct 4 6 pm PDT! https://t.co/rReJ9kW4jU 2020-10-03 14:29:56 @evanmiltenburg Definitely. What other sources of feedback do reviewers get? This is one of the best ways they'll improve. 2020-10-02 17:47:44 Shakespeare couldn't have written this better. This proves we're in a simulation https://t.co/4r2Z8vCvag 2020-10-02 14:49:09 @NateSilver538 The hottest of hot takes 2020-10-02 14:47:55 @KordingLab Sometimes it feels a bit like trying to do science on board the sinking titanic... like, I'll keep writing but should we do something about all this water coming in 2020-10-01 12:25:30 Please consider applying! It's a great opportunity to work at the intersection of experimental neuroscience and AI at DeepMind https://t.co/pB2NMbGsav 2020-09-30 19:06:09 RT @FelixHill84: New in #PNAS @PNASNews Our proposal for how to move from super-smart meta-learning neural language models like @OpenAI's… 2020-09-30 15:48:17 @neuro_kim Biting sarcasm on twitter is all I have to keep me holding on right now 2020-09-30 15:24:03 My head is telling me "increased funding for education and infrastructure will pay dividends down the line", but my heart is wondering if the dismantling of our voting system and democratic traditions is a good thing? #conflicted 2020-09-30 13:47:45 On the one hand, we do need to reconsider our tax systems, but on the other hand, maybe support violent white supremacy? How is one to reasonably decide? 2020-09-30 08:53:29 What a well-reasoned and civil debate! A lot of food for thought, as the arguments presented contained a lot of nuance and facts that'll take me some time to sort through. Still undecided 2020-09-29 12:32:10 @pfau @AustenLamacraft @egrefen if you were a bystander fixed in space, looking at the astronauts rotating, they would have downward velocity relative *to you*, but zero radial velocity (which is relative to the center of the earth) 2020-09-29 12:30:33 @pfau @egrefen i always thought of inertial reference frames as those in which all the laws of physics hold, so all the results of your experiments wouldn't be able to tell you you're accelerating. Everything floats because it's all got the same acceleration. 2020-09-29 12:21:54 @AustenLamacraft @egrefen i did not think this would be a controversial tweet! 2020-09-29 12:16:46 @pfau @egrefen inertial reference frame 2020-09-29 12:15:52 @egrefen Yes but my whole point with the original tweet is that from an *inertial reference frame*, which is not rotating, you're actually falling. Tangential velocity exists in a rotating reference frame. 2020-09-29 12:08:21 @egrefen depends on your reference frame. you have zero radial velocity, but from an inertial reference frame, you do have downward velocity, otherwise you'd go flying off in a straight line 2020-09-29 12:04:24 orbital mechanics is super trippy in general, but to understand why orbiting is the same as falling, i like this image from wikipedia: https://t.co/GRVV8G1wTr https://t.co/k7fn3PlifH 2020-09-29 12:01:08 @egrefen That's what orbiting is. You still have downward velocity, but you're moving so fast tangentially that you never hit the ground :) 2020-09-29 10:40:26 So crazy that these astronauts and the space station are all actually falling toward Earth, but the Earth is just *moving out of the way* https://t.co/XPYv0w5bGA 2020-09-28 12:08:01 Now that NeurIPS2020 decisions are out, consider submitting to the Meta-learning workshop! Deadline Oct. 2 https://t.co/o0j1c70Et0 https://t.co/XidgpfnUTq 2020-09-23 23:41:03 "Get rid of the ballots"??? It's an election what are you talking about??? 2020-09-23 23:38:48 OH OK THIS IS FINE https://t.co/v1f732ZoXr 2020-09-21 20:48:45 RT @zacharylipton: Why the frack is @google renting the top search result for "Joe Biden" out to ad hominem attack ads? Who is asleep at th… 2020-09-20 15:37:09 @kaymtye Yup, I used to be on the fence about a Tesla, but after all the crazy comments he's made, I don't want to put a dime in his pocket 2020-09-19 12:53:44 I realize how incredibly privileged I am to be able to afford a donation of that amount. Any amount helps, or even better would be spending some time to make phone calls / write emails. 2020-09-19 12:50:05 I'm sad, angry, and so so exhausted by this year. But I just donated $1000 to various senate races that have a chance of flipping Dem, and that helped. See links in the thread below. https://t.co/zLXg4gWg7N 2020-09-19 00:36:02 RT @kerrywashington: Her rest is earned. It is our turn to fight. 2020-09-19 00:26:37 floored and in shock. what a tremendous loss for us all. let's take a moment to mourn, collect ourselves, and then vote like everything depends on it. 2020-09-18 10:15:35 @mckaywrigley How can you ensure that what they say is factually true and logically consistent? GPT-3 generates text that sounds plausible, but is often semantically meaningless. Could Einstein actually teach me relativity, or would there just be a bunch of jargon? 2020-09-16 16:56:52 @OriolVinyalsML @ermgrant This is all very much still in the trial phase. If successful we hope it'll inspire other organizers to adopt something similar or build on it 2020-09-16 16:54:20 @OriolVinyalsML @ermgrant Area chairs don't/can't act as mentors for their reviewers since they have way too many and have a different mandate (make accept/reject decisions about their papers). It'd be great if we could change this up for the big conferences though! 2001-01-01 01:01:01