Week Ending 06.09.19
RESEARCH WATCH: 06.09.19
This week was active for "Computer Science", with 1,133 new papers.
The paper discussed most in the news over the past week was by a team at Samsung: "Few-Shot Adversarial Learning of Realistic Neural Talking Head Models" by Egor Zakharov et al (May 2019), which was referenced 175 times, including in the article This technology can make the Mona Lisa talk (sort of) in World Economic Forum. The paper author, Egor Zakharov (Samsung), was quoted saying "Effectively, the learned model serves as a realistic avatar of a person". The paper also got the most social media traction with 61188 shares. On Twitter, @catovitch observed "I wonder if this can/will be used with video compression. If it can be done in real time, the decoder could just be told "build a face our of this frame, then rotate it to these 3D positions for the next 5 frames"".
Leading researcher Yoshua Bengio (Université de Montréal) published "How to Initialize your Network? Robust Initialization for WeightNorm & ResNets".
The paper shared the most on social media this week is by a team at DeepMind: "Generating Diverse High-Fidelity Images with VQ-VAE-2" by Ali Razavi et al (Jun 2019) with 750 shares. @ikrimae (ikrima ✈ E3) tweeted "1. So impressive! 2. Naive take: I wonder if we're the proverbial animal encountering a mirror for the 1st time & being amazed at the glass creature while a R^{1000} being would laugh at our awe 3. Counter take: Hoomans are just ML+environment sampling loop trained over decades".
This week was active for "Computer Science - Artificial Intelligence", with 128 new papers.
The paper discussed most in the news over the past week was "Automated Speech Generation from UN General Assembly Statements: Mapping Risks in AI Generated Texts"by Joseph Bullock et al (Jun 2019), which was referenced 14 times, including in the article AI model creates fake UN speeches that are scarily real in CNET News. The paper got social media traction with 64 shares. A Twitter user, @AI_RRI_Ethics, commented "How easy to impersonate diplomats? To generate #fake speeches with open #ML tools and $7.80? Need for strategic cross-sector & global strategy based with scenarios, #HRs & content-monitoring Congrats".
Leading researcher Yoshua Bengio (Université de Montréal) published "Do Neural Dialog Systems Use the Conversation History Effectively? An Empirical Study"@apsarathchandar tweeted "Attention MAY NOT be all you need. You also need recurrence. Interesting work by Chinnadhurai Sankar and Sandeep Subramanian: #ACL2019".
This week was active for "Computer Science - Computer Vision and Pattern Recognition", with 227 new papers.
The paper discussed most in the news over the past week was by a team at Samsung: "Few-Shot Adversarial Learning of Realistic Neural Talking Head Models" by Egor Zakharov et al (May 2019), which was referenced 175 times, including in the article This technology can make the Mona Lisa talk (sort of) in World Economic Forum. The paper author, Egor Zakharov (Samsung), was quoted saying "Effectively, the learned model serves as a realistic avatar of a person". The paper also got the most social media traction with 61188 shares. On Twitter, @catovitch commented "I wonder if this can/will be used with video compression. If it can be done in real time, the decoder could just be told "build a face our of this frame, then rotate it to these 3D positions for the next 5 frames"".
Leading researcher Oriol Vinyals (DeepMind) published "Generating Diverse High-Fidelity Images with VQ-VAE-2", which had 81 shares over the past 4 days. @ikrimae tweeted "1. So impressive! 2. Naive take: I wonder if we're the proverbial animal encountering a mirror for the 1st time & being amazed at the glass creature while a R^{1000} being would laugh at our awe 3. Counter take: Hoomans are just ML+environment sampling loop trained over decades". This paper was also shared the most on social media with 750 tweets. @ikrimae (ikrima ✈ E3) tweeted "1. So impressive! 2. Naive take: I wonder if we're the proverbial animal encountering a mirror for the 1st time & being amazed at the glass creature while a R^{1000} being would laugh at our awe 3. Counter take: Hoomans are just ML+environment sampling loop trained over decades".
Over the past week, 23 new papers were published in "Computer Science - Computers and Society".
The paper discussed most in the news over the past week was by a team at University of Washington: "Defending Against Neural Fake News" by Rowan Zellers et al (May 2019), which was referenced 8 times, including in the article To Catch a Fake: Machine learning sniffs out its own machine-written propaganda in ZDNet. The paper author, Rowan Zellers (University of Washington), was quoted saying "Our work does suggest that there is an arms race between the adversary and the verifier". The paper also got the most social media traction with 262 shares. On Twitter, @Thom_Wolf commented "If you want a sneek-peek in and co-workers work on GROVER (a 1.5 billion param GPT-2-like model), check this live tweet 👇 Interesting hints, results, and analysis! Paper: Demo".
This week was active for "Computer Science - Human-Computer Interaction", with 24 new papers.
This week was extremely active for "Computer Science - Learning", with 508 new papers.
The paper discussed most in the news over the past week was by a team at Samsung: "Few-Shot Adversarial Learning of Realistic Neural Talking Head Models" by Egor Zakharov et al (May 2019), which was referenced 175 times, including in the article This technology can make the Mona Lisa talk (sort of) in World Economic Forum. The paper author, Egor Zakharov (Samsung), was quoted saying "Effectively, the learned model serves as a realistic avatar of a person". The paper also got the most social media traction with 61193 shares. On Twitter, @catovitch said "I wonder if this can/will be used with video compression. If it can be done in real time, the decoder could just be told "build a face our of this frame, then rotate it to these 3D positions for the next 5 frames"".
Leading researcher Yoshua Bengio (Université de Montréal) published "How to Initialize your Network? Robust Initialization for WeightNorm & ResNets".
The paper shared the most on social media this week is by a team at DeepMind: "Generating Diverse High-Fidelity Images with VQ-VAE-2" by Ali Razavi et al (Jun 2019) with 754 shares. @ikrimae (ikrima ✈ E3) tweeted "1. So impressive! 2. Naive take: I wonder if we're the proverbial animal encountering a mirror for the 1st time & being amazed at the glass creature while a R^{1000} being would laugh at our awe 3. Counter take: Hoomans are just ML+environment sampling loop trained over decades".
Over the past week, 12 new papers were published in "Computer Science - Multiagent Systems".
The paper shared the most on social media this week is by a team at Massachusetts Institute of Technology: "Finding Friend and Foe in Multi-Agent Games" by Jack Serrino et al (Jun 2019) with 66 shares. The investigators develop the DeepRole algorithm, a multi - agent reinforcement learning agent that they test on The Resistance : Avalon, the most popular hidden role game. @kevinakwok (Kevin Kwok) tweeted "AI x Resistance! So fascinating Hope applied to Avalon eventually! - will meta improve 10x? - Is China style truly dominant to US-West? - So many questions! Thanks and If ever in SF come join us for Avalon Great find by".
This week was active for "Computer Science - Neural and Evolutionary Computing", with 36 new papers.
Leading researcher Jianfeng Gao (Microsoft) published "Budgeted Policy Learning for Task-Oriented Dialogue Systems" The authors present a new approach that extends Deep Dyna - Q (DDQ) by incorporating a Budget - Conscious Scheduling (BCS) to best utilize a fixed, small amount of user interactions (budget) for learning task - oriented dialogue agents.
This week was active for "Computer Science - Robotics", with 48 new papers.
The paper discussed most in the news over the past week was "EVDodge: Embodied AI For High-Speed Dodging On A Quadrotor Using Event Cameras" by Nitin J. Sanket et al (Jun 2019), which was referenced 2 times, including in the article AI helps drones dodge fast-moving objects in Venturebeat. The paper was shared 4 times in social media.
Leading researcher Sergey Levine (University of California, Berkeley) published "Off-Policy Evaluation via Off-Policy Classification" The investigators consider the problem of model selection for deep reinforcement learning (RL) in real - world environments. This paper was also shared the most on social media with 57 tweets.