RT @firstname.lastname@example.org: If there's one thing you could change about the media and public conversation on AI, what would it be? Here's mine: stop talking about AI as if it has agency, because it doesn't — it's the people building and deploying it who do.
Cool paper alert! "CURIOUS: Intrinsically Motivated Multi-Task, Multi-Goal Reinforcement Learning," Colas et al.: https://arxiv.org/abs/1810.06284
Cool work! Still feel that UVFA is underpreciated (as I argued here https://thegradient.pub/how-to-fix-rl/).
[Fwd from @Miles_Brundage@twitter.com (for images) ]
Another awesome demonstration that AI will aid the creativity of people, not "replace" artists! This immediately strikes me as a good tool for trying random stuff out and discovering new ideas.
RT @email@example.com An extremely cool AI project from Google Magenta: the Piano Genie, which lets you improvise fluently on a piano by bashing just eight buttons https://www.theverge.com/2018/10/16/17982596/google-magenta-ai-piano-genie-improvisation-neural-networks
Super cool indeed! Always felt that domain randomization as done now is far too ad hoc, nice to see it being done in a more well motivated manner.
RT @firstname.lastname@example.org Super cool new paper from the @NvidiaAI@twitter.com Robotics group in Closing the Sim-to-Real Loop: Adapting Simulation Randomization with Real World Experience [Paper https://buff.ly/2pWMLUg , video:https://buff.ly/2pWqXbq ]
Something neat - got around to posting recording of interview I did on @KPOORADIO@twitter.com about the AI PhD life and state of AI, as part of @email@example.com's cool Science Envoy program (https://wonderfest.org/science-envoy/).
Check out more interviews: http://www.vicesbyproxy.com/kpoo_play/lmtym.html
RT @OlaRosling@twitter.com: Study contradicting common sense: Exposure to opposite ideas may actually lead to increased polarization, if communication style disallow patiently listening and dialog. Solution: Don’t let everything become politics! https://twitter.com/timharford/status/1051453819798913024
RT @Bananaft@twitter.com: I just published “My journey into fractals” https://medium.com/p/my-journey-into-fractals-d25ebc6c4dc2
RT @Noah_K_Whiteman@twitter.com: I showed the tree of life in a talk to @UCBerkeley@twitter.com alumni today. I asked them close their eyes and come to understand that they are the product on an unbroken chain of cell divisions that started 4 billion years ago! So much extinction and so lucky to be here. #Homecoming2018
RT @firstname.lastname@example.org: BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding:
SOTA on 11 tasks. Main additions:
- Bidirectional LM pretraining w/ masking
- Next-sentence prediction aux task
- Bigger, more data
It seems LM pretraining is here to stay.
Since it is #WorldMentalHealthDay, seems like a good time to post this little blog post again. Take care of yourself!
RT @email@example.com: Excited to propose Model Cards for Model Reporting as a pressing step forward in transparency, accountability, and the responsible democratization of AI technology. https://arxiv.org/abs/1810.03993
-- with @timnitGebru@twitter.com and others whose twitter handles I'm not sure of!
"The best way to fight for the carbon fee and dividend is to join @firstname.lastname@example.org Lobby, which now has more than 90,000 members but needs more, especially young people. CCL members are appropriately polite and respectful as they cajole politicians in Washington. "
hey we should all support 'free' stuff that refreshingly does not get funded by advertising!
I just donated to @Wikipedia@twitter.com. Support free knowledge! #iloveWikipedia https://donate.wikimedia.org/?utm_medium=SocialMedia&utm_campaign=ThankYouPage&utm_source=Twitter&uselang=en
"Economists agree: If 100 percent of this fee is distributed uniformly to the public, the economy will be spurred, GNP will rise, and millions of jobs will be created. Our energy infrastructure will be steadily modernized with clean energies and energy efficiency."
"As long as fossil fuels are cheap, they will be burned and emissions will be high. Fossil fuel use will decline only if the price is made to include costs of pollution and climate change to society."
when will we finally wake up... @email@example.com's fee and dividend also good
RT @firstname.lastname@example.org Thirty years later, what needs to change in our approach to climate change https://www.bostonglobe.com/opinion/2018/06/26/thirty-years-later-what-needs-change-our-approach-climate-change/dUhizA5ubUSzJLJVZqv6GP/story.html?event=event25 via @BostonGlobe@twitter.com
"We propose CAML, a meta-learning method that partitions the model parameters into two parts: context parameters that are adapted on individual tasks, and shared parameters that are meta-trained and shared across tasks."
surprised this has not been done yet, but cool!
RT @Miles_Brundage@twitter.com "CAML: Fast Context Adaptation via Meta-Learning," @email@example.com et al.: https://arxiv.org/abs/1810.03642
RT @firstname.lastname@example.org: Imitation from watching videos (in this case, from YouTube) by virtual characters -- directly imitating acrobatic skills, dance, etc. via deep RL + HMR for pose recovery: https://bair.berkeley.edu/blog/2018/10/09/sfv/ https://twitter.com/xbpeng4/status/1049679153916665856
CS PhD at Stanford Vision Lab. Creator/lead editor http://skynettoday.com 🤖 Into AI, robots, cinema, coding, photography. Russian.
Scholar Social is meant for: researchers, grad students, librarians, archivists, undergrads, academically inclined high schoolers, educators of all levels, journal editors, research assistants, professors, administrators—anyone involved in academia who is willing to engage with others respectfully.
We strive to be a safe space for queer people and other minorities, recognizing that there can only be academic freedom where the existence and validity of interlocutors' identities is taken as axiomatic.
"A Mastodon profile you can be proud to put on the last slide of a presentation at a conference"
(Participation is, of course, optional)
Scholar Social features a monthly "official" journal club, in which we try to read and comment on a paper of interest.
Any user of Scholar Social can suggest an article by sending the DOI by direct message to @email@example.com and one will be chosen by random lottery on the last day of the month. We ask that you only submit articles that are from *outside* your own field of study to try to ensure that the papers we read are accessible and interesting to non-experts.