Discover and read the best of Twitter Threads about #DeepLearning

Most recents (24)

Introducing RepNet, a model that counts repetitions in videos of *any* action

w @yusufaytar, @JonathanTompson, @psermanet and Andrew Zisserman


#CVPR2020 #computervision #deeplearning
Here's an overview of RepNet's architecture.
An integral part of RepNet is the temporal self-similarity matrix (TSM) which not only makes it easy to count repetitions but also drives generalization to actions and domains not seen during training.
Read 9 tweets
#Doctors and #Medicine need an #AI update.
Bold and convincing book.

#ArtificialIntelligence has the ability to bring the fourth industrial revolution. Healthcare is deemed to be an industry which can get transformed the most.

@EricTopol is a #cardiologist, historian of the present and a #MedicalFuturist.
He believes that the way we practice medicine now ( #ShallowMedicine ) relies too heavily on human inputs and is thus plagued by human biases and imperfections.

Read 23 tweets
6 weeks ago we announced our #Covid19 Open #ArtificialIntelligence initiative (#COAI)¹, opening-up our #DeepLearning platform for doctors/hospitals dealing with #Coronavirus patients to facilitate breakthrough medical discoveries & actionable findings to fight the pandemic.
Few days ago, #radiologists @GustaveRoussy, @APHP, @Inria_Saclay & researchers @ BioMaps used our #AI to develop a high-fidelity model capable of predicting #Covid_19 patient complications based on patient's CT scans².
Our state-of-the-art #radiology model is open for all researchers, practitioners and/or medical institutions working with #coronavirus patients. We can unveil new discoveries from #medicaldata!
Read 4 tweets
For students and others interested in expanding their knowledge of AI during this period, we thought it might be helpful to ask our researchers what they consider to be the most impactful and insightful resources available to use #AtHomeWithAI (1/9)
Today’s suggestions are from research scientist Feryal Behbahani @feryalmp! #AtHomeWithAI
Want to know more about reinforcement learning? She recommends @EmmaBrunskill’s’ online lecture series from @StanfordEng. Watch them here: #AtHomeWithAI
Read 9 tweets
We need new antibiotics to counter resistance. Now @MITdeptofBE's James Collins,@BarzilayRegina @ItsJonStokes @MIT_CSAIL @WyssInstitue, @broadinstitute & team have trained a
#deeplearning #AI model to help the path of discovery…
@CellCellPress @AIHealthMIT
@MITdeptofBE @BarzilayRegina @ItsJonStokes @MIT_CSAIL @broadinstitute @CellCellPress @AIHealthMIT @CellPressNews When there's any report about an #AI and drug discovery effort, my "go to" is @Dereklowe for perspective.
When he says "So overall, this is an impressive paper," it has gravitas:…
@MITdeptofBE @BarzilayRegina @ItsJonStokes @MIT_CSAIL @broadinstitute @CellCellPress @AIHealthMIT @CellPressNews @Dereklowe "The study is remarkable, says Jacob Durrant @LabDurrant ... The team didn’t just identify candidates, but also validated promising molecules in animal tests."… @NatureNews by @JoMarchant May be the best #AI-drug story yet, especially for unmet need.
Read 3 tweets
We are very excited to share our SHANEL method making the intact human organs transparent, just published at Cell.… by @shan_heather et al.
See short tweetorial for details:
#clearing #imaging #3D #deeplearning #AI @CellCellPress
Mapping the human brain or other organs is a crucial step forward deciphering how they function in health and disease.
Tissue Clearing methods work well on rodent tissue but haven been poorer on stiff & aged human organs. Making the whole human organs transparent required a new approach.
Read 13 tweets
In 2017, I wrote the book 'The Deep Learning Playbook' that described 3 pillars of research that will drive innovation: Meta-Learning, Modularity and Game Theory.
In 2017, we had a nascent understanding of these 3 and how they relate to each other. Today, we now have a much better understanding of their relationship and this will lead to accelerated innovation. #deeplearning #ai.
The overarching model is near completion and we only need to fill in the blanks.
Read 17 tweets

"Qui l’arrivo di VR e di AR compiono come mai immaginato prima, né sognato prima, il sogno di Edward Bernays"

Researchers at the Samsung #AI Center have produced a system that can recreate realistic motion from just a single image of a person

"We believe that telepresence technologies in AR, VR and other media are to transform the world"
Graphene could soon make your computer 1000 times faster

“If you want to continue to push technology forward, we need faster computers to be able to run bigger and better simulations for climate science, for space exploration, for Wall Street”
Read 9 tweets
I disagree. It's difficult to understand and fix bias through #AI algorithms for #DeepLearning These are highly non-linear blackbox models. they amplify biases. many works show that trying to fix superficially is like putting lipstick on a pig. We need fresh thinking
In a recent work we report an intriguing finding to detect bias : assess hardness of different samples for #DeepLearning model. It's a first step. Using it to fix bias is much harder
More importantly our work shows that #DeepLearning model is deciding which samples are hard for it to classify and introducing bias. Here I use bias to mean disparate treatment not statistical bias. Model leaves harder example with worse accuracy and more vulnerability to noise
Read 9 tweets
High time to read: "Causality for Machine Learning"… by Bernhard Schölkopf #ai
"Machine learning often disregards information that animals use heavily: interventions in the world, domain shifts, temporal structure"
"The transformation, however, really started already in the mid 20th century under the name of cybernetics. It replaced energy by information."
Read 20 tweets
In the pursuit of more advanced AI (or even AGI), one must be able to articulate a model of intelligence. That model doesn't have to be absolutely correct, but without one is like navigating unfamiliar territory without a map. You will miss important landmarks. #ai #deeplearning
The Achilles heel of younger AI researchers is the lack of exposure to alternative models of cognition. One has tunnel vision when optimization is one's framework for cognition. Cognition is very messy and acknowledging this should be the first step.
Behaviorism in psychology and reductionism in neuroscience are examples of pursuits that purposely ignored the messiness of the brain. Fortunately, we are moving to more holistic approaches in these areas. The same should happen in AI research.
Read 22 tweets
The @StanfordHAI (Human-centered #AI) Conference on #AiEthics, Policy, & Governance is getting started, live stream below. Speaker list is fantastic. Will be tweeting.

#machinelearning #data4good
@StanfordHAI "Silicon Valley, we have a problem" opens @erikbryn

Speaks to the 'great decoupling' - where productivity and GDP expand with technology, but don't take household incomes with them.

Part of the problem is lower-paid occupations have more automate-able tasks.
@StanfordHAI @erikbryn Both @Susan_Athey and @erikbryn speaking to ML adoption in private sector and the important distinction between labor replacing vs. labor augmenting #automation.

The "Suitability for Machine Learning" or SML scores from this
@MIT_IDE is useful here:… #ai
Read 22 tweets
The oddest thing about Artificial Neural Networks is that they actually work despite being based on a completely false model of a biological neuron. #ai #deeplearning
Why Artificial Neural Networks work remains a mystery. Understanding that "Why" can inform us why real biological neurons might work. This is what's exciting, we got something synthetic that does work!!
Dennett uses the term 'Inversion of Reasoning' or 'Competence without Comprehension' to describe a universal characteristic of many kinds of evolution. Artificial Neural Networks share this common characteristic with biological neurons.
Read 26 tweets
Impressive life science research that would not have been possible w/o machine vision #AI (that they developed)…
400,000 fly videos-> whole #brain maps for gene expression + anatomical basis of 14 traits @CellCellPress @HHMIJanelia Kristin Branson 1/3
@CellCellPress @HHMIJanelia Beyond DeepLabCut, new #AI software to do that is DeepPoseKit… @eLife #deeplearning from flies to giraffes 2/3
@CellCellPress @HHMIJanelia @eLife This week @NatureNews reviewed these #AI machine vision tools: "Deep learning powers a motion-tracking revolution"
by @robertakwok #openaccess… 3/3
Read 3 tweets
Philosophically, it (Deep Learning) is every rationalist’s nightmare: the triumph of the inexplicable over the clear, and that of the probabilistic over the deterministic...…
"Here Charles Sanders Peirce’s tychism and Georg Simmel’s dynamism will help us understand what these (Deep Learning) developments mean for the future of human (and machine) subjectivity." #deeplearning #ai - Peirce formulates four reasons in support of his hypothesis: Growth and increasing complexity, Variety and diversity, Regularity (laws of nature),
Read 7 tweets
1/ While this will play well (and get cited a lot) among the anti-#deeplearning holdouts, I was left a bit underwhelmed. I wanted to find some interesting edge cases where DL is not working (so we can work out solutions), but instead got a set of pretty unreasonable comparisons
2/ The deep learning models are tiny (4 conv layers) with justification that it works for MNIST. Everything works for MNIST! Linear regression works for MNIST!…

We know in complex images deeper and more complex is vastly better, and does less overfitting!
3/ The linear and non-deep models are not "apples to apples" either though. This isn't deep learning vs simple models, it is deep learning vs incredibly complex feature engineering built up over decades of research.
Read 14 tweets
DeepMind now researching the connection of Sleep to Deep Learning:… #ai #deeplearning
Biological brains require sleep, but what exactly does sleep have to do with cognition?…
But why do brains require re-experiencing events to synthesize its understanding of these events? It seems that what humans remember about experiences are the summaries of these experiences played backward. That is how we build our causality networks!
Read 12 tweets
Black Mirror.
Social Credit Score.

“They don’t fear you”
Black Mirror on Netflix does an episode where a social credit score dystopian society ruins a young woman’s life.

Just fantasy?
Or just [their] fantasy?

They don’t hide it.
They don’t fear you.

TV shows (Homeland, Black Mirror, etc), movies (Eyes Wide Shut, Matrix, etc) are their way of laughing in our face.
Read 8 tweets
Deep Learning (DL circa 2012) big breakthrough is that it demonstrated a new kind of Artificial Intuition. Prior to DL, computers provided programmable Artificial Logic (circa 1945).
AlphaGo big breakthrough is that it demonstrated that with the use of DL one could beat the best human in a game of intuition (i.e. Go). .
AlphaZero big breakthrough is that you could train a system from scratch (without human knowledge) and it could be the best human system Stockfish that was programmed using logic.…
Read 6 tweets
🆃🅴🅾🆁🅴🅼🅰 (de representación de Arnold-Kolmogorov)

Tiene que ver con #deeplearning, #machinelearning, AI, #bigdata y todas esas palabras cool (y es una excusa para hablar de Arnold).

Dedicado a @wsosaescudero y los fanáticos del libro verde.
🗣️ porfa RT así difundimos #TeRegaloUnTeorema
Dice así: toda función continua, aunque dependa de muchas variables se puede escribir como combinación de funciones (continuas) que dependen de una sola variable.
Read 13 tweets
A "worrying analysis":

"18 [#deeplearning] algorithms ... presented at top-level research conferences ... Only 7 of them could be reproduced w/ reasonable effort ... 6 of them can often be outperformed w/ comparably simple heuristic methods."


[Updates worth tweeting]

There is much concern about #reproducibility issues and flawed scientific practices in the #ML community in particular & #academia in general.

Both the issues and the concerns are not new.

Isn't it time to put an end to them?
There are several works that have exposed these and similar problems along the years.

👏👏 again to @Maurizio_fd et al. for sharing their paper and addressing #DL algorithms for recommended systems (1st tweet from this thread).

But there is more, unfortunately:
Read 18 tweets
“He absorbs knowledge like a sponge,” IBM Watson’s chief technology officer Rob High tells CNBC, “and is highly driven to learn even more about #deeplearning & #AI.”

Read about Tanmay Bakshi, 14-year-old self taught #ArtificialIntelligence expert for IBM:…
“I’m really passionate about #AI, and I’m excited to share my knowledge with the rest of the community,” Bakshi explains.
“Why should other people need to rediscover fire or reinvent the wheel just because I didn’t want to share what I had already done?”
His expertise in #AI has already garnered him praise.
After speaking at the 2017 Knowledge Summit in Dubai, he was given the Knowledge Ambassador Award from the King of Dubai’s Mohammed bin Rashid Al Maktoum Foundation.
He's also an IBM Cloud champion & an honorary Cloud adviser
Read 7 tweets
[THREAD] 1/ In light of the #humantrafficking cases of Jeffrey Epstein & R. Kelly this week, we want to share some of the #hightech ways orgs are working to catch & prosecute #traffickers while protecting children. A summary of our white paper on 40 #ToolsAgainstChildTrafficking.
2/ This #whitepaper presents a list of technologies combating #childtrafficking. By providing this resource, we hope to consolidate core knowledge and enable law enforcement, government agencies, and nonprofits to select and combine strategies with maximum effectiveness.
3/ A prominent #hightech method to fight trafficking is #ImageRecognition. Orgs are able to identify victims (@thorn), create #videofingerprints to filter & #eliminateCSAM (@_project_vic, @twohathq), and scan the internet for photos of missing children in ads (@marinusai).
Read 11 tweets

Related hashtags

Did Thread Reader help you today?

Support us! We are indie developers!

This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3.00/month or $30.00/year) and get exclusive features!

Become Premium

Too expensive? Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal Become our Patreon

Thank you for your support!