In the papers no. 7

AutoGAN: Neural Architecture Search for Generative Adversarial Networks [arXiv:1908.03835v1] In this paper, we present the first preliminary study on introducing the NAS algorithm to generative adversarial networks (GANs), dubbed AutoGAN–Specifically, our discovered architectures achieve highly competitive performance compared to current state-of-the-art hand-crafted GANs, e.g., setting new state-of-the-art FID scores of 12.42 on CIFAR-10, and 31.01 on STL-10, respectively

Tag2Pix: Line Art Colorization Using Text Tag With SECat and Changing Loss [arXiv:1908.05840v1] The discriminator is based on an auxiliary classifier GAN to classify the tag information as well as genuineness. In addition, we propose a novel network structure called SECat, which makes the generator properly colorize even small features such as eyes, and also suggest a novel two-step training method where the generator and discriminator first learn the notion of object and shape and then, based on the learned notion, learn colorization, such as where and how to place which color.

On the Validity of Self-Attention as Explanation in Transformer Models [ arXiv:1908.04211v1] We investigate to what extent the implicit assumption made in many recent papers – that hidden embeddings at all layers still correspond to the underlying words – is justified. — we argue that attention visualizations are misleading and should be treated with care when explaining the underlying deep learning system.

Attention is not not Explanation (see above) [arXiv:1908.05164v1] In this work, we propose the Unconstrained Monotonic Neural Network (UMNN) architecture based on the insight that a function is monotonic as long as its derivative is strictly positive. — we argue that attention visualizations are misleading and should be treated with care when explaining the underlying deep learning system.

Predicting 3D Human Dynamics from Video [arXiv:1908.04781v1] In this work, we present perhaps the first approach for predicting a future 3D mesh model sequence of a person from past video input. — Our approach can be trained on video sequences obtained in-the-wild without 3D ground truth labels [site]

Temporal Collaborative Ranking Via Personalized Transformer [arXiv:1908.05435v1] we find our model is not only more interpretable but also able to focus on recent engagement patterns for each user. Moreover, our SSE-PT model with a slight modification, which we call SSE-PT++, can handle extremely long sequences and outperform SASRec in ranking results with comparable training speed, striking a balance between performance and speed requirements. Code and data are open sourced at this URL.

Multimodal Emotion Recognition Using Deep Canonical Correlation Analysis [arXiv:1908.05349v1] The experimental results indicate that DCCA has greater robustness. By visualizing feature distributions with t-SNE and calculating the mutual information between different modalities before and after using DCCA, we find that the features transformed by DCCA from different modalities are more homogeneous and discriminative across emotions.


Hiding from machines in fashion

SPF is not the only daily protection you need walking around anymore. You need to move with the fashions or be outcast put in a database.

Adversarial prêt-à-porter has arrived. The looks span from heinous utility to some workable pieces.

I maintain that the optimal protective gear will not have any unique visual component that makes you stand out. Hide in plain sight.

Its your obligation to protect your Fourth Amendment rights- since the high court ruled that if you do not – you have no reasonable expectation of privacy.

At the same time in what will be yet another dystopian catch-22 Some of these products may trigger anti-masklaws arguments.

When is a mask not a mask? When it hides more than your face. And that is the argument around facial recognition technology

There big gap between utility (unlocking your phone) like a key and a Swiss army knife that takes everything and makes assumptions with out any presumption of innocence.

Collected against your will and often without your knowledge. Placing you in databases you are not aware of – that feed profiles you can not see.

It’s more than the security, falsely implied – it’s supervision you never needed.

Today this post will establish a list of commercially available anti surveillance wearables

If you know of a company or related products please get in touch to add to this now living document.

As the list grows will start breaking them out into categories.

AH ProjectsHyperFace

Adversarial Fashion

CV Dazzle


Privacy Updates

London’s King’s Cross using facial recognition in security cameras [ft] Canary Wharf Group, the company that owns both private offices and public spaces in the area, is actively speaking to facial recognition suppliers to pilot the technology in an area traversed by 140,000 people daily, as part of its security systems.

British ethicists are challenging justice by algorithm [economist] Several use programs to predict where and when crimes are likely to occur. Cambridge University helped Durham Constabulary design an algorithm to estimate the likelihood of a suspect reoffending. It helps the authorities decide whether someone should be granted bail or qualify for rehabilitation as an alternative to prosecution. At least one force is keen to install microphones on “smart lamp-posts” to gather intelligence in crowds

Court rules Arizona residents have right to internet privacy [ArizonaTimes] Arizonans have a constitutional right to online privacy to keep police from snooping around to find out who they are without first getting a warrant, the state Court of Appeals has ruled. In what appears to be the first ruling of its kind in the state, the majority said Internet users have a “reasonable expectation of privacy” that the information they furnish about themselves to Internet providers will be kept secret. That specifically includes who they are and their home address.

This Tesla Mod Turns a Model S Into a Mobile ‘Surveillance Station’ [wired] At the Defcon hacker conference today, security researcher Truman Kain debuted what he calls the Surveillance Detection Scout. The DIY computer fits into the middle console of a Tesla Model S or Model 3, plugs into its dashboard USB port, and turns the car’s built-in cameras—the same dash and rearview cameras providing a 360-degree view used for Tesla’s Autopilot and Sentry features—into a system that spots, tracks, and stores license plates and faces over time. The tool uses open source image recognition software to automatically put an alert on the Tesla’s display and the user’s phone if it repeatedly sees the same license plate. 

Patel v. Facebook Opinion Federal Court Rules Facebook Users Can Sue Company Over Unlawful Use of Face Recognition Technology The U.S. Court of Appeals for the Ninth Circuit ruled today that Facebook users can sue the company over its use of face recognition technology. The ruling is the first decision of an American appellate court directly addressing the unique privacy harms posed by the face recognition technology being increasingly pushed on members of the public without their knowledge and consent. [aclu]

I Tried Hiding From Silicon Valley in a Pile of Privacy Gadgets – Within three weeks, I got tired of being careful. My faraday bag is somewhere in the back of my car, because I like to use my phone for GPS and playing podcasts. I stopped wearing my Reflectacles in public. I haven’t scrubbed my old tweets away with Jumbo in a while. I use my Visa to buy stuff from Amazon, which it delivers to my house. I plugged Alexa back in. [Bloomberg Businessweek]

Moratorium requested for Lockport facial recognition system [lockportjournal] The New York Civil Liberties Union has asked the New York State Education Department to place a moratorium the use of a facial recognition system in Lockport schools. The request comes less than 24 hours after the Lockport School Board approved revisions to the district’s policy for the application of the system. The revised policy included several changes from the original version. The major change eliminates the possibility of suspended student being added to the database tied to the system unless their actions are deemed by law enforcement to constitute a legitimate threat.

Facial Recognition Software Prompts Privacy, Racism Concerns in Cities and States [PEW] The study found live cameras were tracking the movements of tenants in apartment buildings and even patients coming and going from a medical center, which Detroit Police Chief James Craig denied in an interview with Stateline. Craig said his department does not use facial recognition software to track people.number of constitutional protections that limit what the government can do.”

In the papers 6

Neural Blind Deconvolution Using Deep Priors [arXiv:1908.02197] The process of neural optimization can be explained as a kind of “zero-shot” self-supervised learning of the generative networks, and thus our proposed method is dubbed SelfDeblur. Experimental results show that our SelfDeblur can achieve notable quantitative gains as well as more visually plausible deblurring results in comparison to state-of-the-art blind deconvolution methods on benchmark datasets and real-world blurry images. The source code is available at this https URL.

Word2vec to behavior: morphology facilitates the grounding of language in machines [arXiv:1908.01211v1] Enabling machines to respond appropriately to natural language commands could greatly expand the number of people to whom they could be of service.

From Crowdsourcing to Crowdmining: Using Implicit Human Intelligence for Better Understanding of Crowdsourced Data [arXiv:1908.02412v1] This paper presents CrowdMining. In particular, we observe that the knowledge hidden in the process of data generation, regarding individual/crowd behavior patterns (e.g., mobility patterns, community contexts such as social ties and structure) and crowd-object interaction patterns (flickering or tweeting patterns) are neglected in crowdsourced data mining.

In The papers 5

The Myths of Our Time: Fake News  1908.01760v1 [pdf] Seeking insights on the fear and desire hidden underneath these modified or generated stories, we use machine learning methods to generate fake articles and present them in the form of an online news blog. This paper aims to share the details of our pipeline and the techniques used for full generation of fake news, from dataset collection to presentation as a media art project on the internet.

Word2vec to behavior: morphology facilitates the grounding of language in machines 1908.01211v1 [pdf] Enabling machines to respond appropriately to natural language commands could greatly expand the number of people to whom they could be of service. Recently, advances in neural network-trained word embeddings have empowered non-embodied text-processing algorithms, and suggest they could be of similar utility for embodied machines. Here we introduce a method that does so by training robots to act similarly to semantically-similar word2vec encoded commands.

Yoga-Veganism: Correlation Mining of Twitter Health Data 1906.07668v1 [pdf] In this work, we explore Twitter data related to health. We extract the popular topics under  different categories (e.g. diet, exercise) discussed in Twitter via topic modeling, observe model behavior on new tweets, discover interesting correlation (i.e. Yoga-Veganism). We evaluate accuracy by comparing with ground truth using manual annotation both for train and test data. 

Learning to Train with Synthetic Humans1908.00967v1 [pdf] Neural networks need big annotated datasets for training. However, manual annotation can be too expensive or even unfeasible for certain tasks, like multi-person 2D pose estimation with severe occlusions. A remedy for this is synthetic data with perfect ground truth. Here we explore two variations of synthetic data for this challenging problem; a dataset with purely synthetic humans and a real dataset augmented with synthetic humans. 

Les jeux sont faits

Media M&A and musical chairs continue to fast track with news today of Tencent in talks to buy a 10% stake in UMG.

Music in the mix today as European Court Of Justice Rules On Three Big Copyright Cases –  At the heart of the case was the ever-present conflict between copyright and freedom of expression. And here, the court said copyright trumps freedom of expression and creativity, even when it’s a tiny 2 second music sample. The only exception is if you somehow distort the sample so that it’s not recognizable. “So, all you German musicians out there, make sure to distort your Kraftwerk samples.”

Libra Network Concerns – Data protection officials in the US, EU, UK, Australia and Canada have released a Joint statement on global privacy expectations of the Libra network (pdf) To date, while Facebook and Calibra have made broad public statements about privacy, they have failed to specifically address the information handling practices that will be in place to secure and protect personal information. – Libra Network operated by Facebook and 27 partners may “instantly become the custodian of millions of people’s personal information”.

Quad 9 From IBM +others — Quad9’s DNS promises to check websites visited by users against the IBM X-Force threat intelligence database, as well as 18 additional databases, and will detect and automatically block access to known malicious websites. Also promises not to collect, store, or sell any information about your browsing habits .

DNS Services Cloudfare Google Quad9

A Multimillionaire Surveillance Dealer Steps Out Of The Shadows . . . And His $9 Million WhatsApp Hacking Van —Intellexa. It’s a one-stop-shop, cyber arsenal for cops in the field. Alongside Android hacking tools, there’s tech that can recognize your face wherever you travel, listen in on your calls, and locate all the phones in an entire country within minutes, Dilian boasts. Every 15 minutes, he can know where you are, he says.

Live by creators die by creators

YouTube Terminated The Account Of A 14-Year-Old Star Over Her Anti-Gay Video — After her channel was taken down, Soph tweeted an image of herself with what appears to be an assault rifle, with the caption “youtube headquarters here I come.” She later removed the tweet. “Gun tweet obviosly a joke,” she wrote in a follow-up tweet.

– The Soph YouTube account had been spewing all sorts of bigoted hate for some time. YouTube CEO Susan Wojcicki continues to fail to execute. The service has languished for half a decade turning into a recommendation engine for hate and on ramp for kids to the worst programing ever (TWPE). This service curated correctly could rival all TV platforms- Its failure is cannon.

Social Share Privacy – Share with out tracking

A Brief Tour of Differential Privacy — lecture slides from a CMU course. 

Cinematic scientific visualization: the art of communicating science

Carlos Santana walks us through tripping though his woodstock set. [intv/nyt] Two hours after I took it, there was a face in my face that said, “You need to go on right now, otherwise you’re not going to play.” By this time I was really, really on it, you know? I just held on to my faith, and what my mom taught me. I asked, over and over, “Just help me stay in tune and on time.”

In the Papers

A subjective look at new computer science papers of note.

Tracking without bells and whistles The problem of tracking multiple objects in a video sequence poses several challenging tasks. For tracking-by-detection these include object re-identification, motion pre diction and dealing with occlusions. We present a tracker (without bells and whistles) that accomplishes tracking without specifically targeting any of these tasks, in particular, we perform no training or optimization on tracking data

Extract and Merge: Merging extracted humans from different images utilizing Mask R-CNN  In this work, we proposed an application utilizing Mask R-CNN (for object detection and mask segmentation) that can extract human instances from multiple images and merge them with a new background.

KiloGrams: Very Large N-Grams for Malware Classification we present a method to find the top-kk most frequent nn-grams that is 60×× faster for small nn, and can tackle large n≥1024n≥1024. Despite the unprecedented size of nn considered, we show how these features still have predictive ability for malware classification tasks.

Benefiting from Multitask Learning to Improve Single Image Super-Resolution In this paper, we introduce a novel approach using categorical information to tackle the SISR problem; we present a decoder architecture able to extract and use semantic information to super-resolve a given image by using multitask learning, simultaneously for image super-resolution and semantic segmentation. 

FSS-1000: A 1000-Class Dataset for Few-Shot Segmentation Both our approach and dataset are simple, effective, and easily extensible to learn segmentation of new object classes given very few annotated training examples. 

MoBiNet: A Mobile Binary Network for Image Classification  our MoBiNet focuses on binarizing the already-compressed neural networks like MobileNet without the need of a pre-trained model to start with. Therefore, our proposal results in an effectively small model while keeping the accuracy comparable to existing ones.

Computing the Value of Data: Towards Applied Data Minimalism  Using two types of recommender systems, we first demonstrate how much data is ineffective in both settings. We then present a general account of computing data value via sensitivity analysis, and how, in theory, individual data points can be priced according to their informational contribution to automated decisions. 

In The Papers

A collection of recent computer science papers…

Real-time Event Detection on Social Data Streams

iCartoonFace: A Benchmark of Cartoon Person Recognition

RadioTalk: a large-scale corpus of talk radio transcripts

Synthetic Image Augmentation for Improved Classification using Generative Adversarial Networks  

Music Recommendations in Hyperbolic Space: An Application of Empirical Bayes and Hierarchical Poincaré Embeddings

Dressing 3D Humans using a Conditional Mesh-VAE-GAN

I-Keyboard: Fully Imaginary Keyboard on Touch Devices Empowered by Deep Neural Decoder

Morning Link Roundup

Generalise, don’t specialise: why focusing too narrowly is bad for us Arturo Casadevall, an internationally renowned scientist, believes that increasing specialisation has created a “system of parallel trenches” in the quest for innovation. Everyone is digging deeper into their own trench and rarely standing up to look in the next trench over, even if the solution to their problem happens to reside there

Machine vision can spot unknown links between classic artworks  A new algorithm reveals a web of artistic connections by looking for humans posed similarly in different paintings.

‘Cute goddess’ revealed as middle-aged woman in live-stream glitch Chinese vlogger caught using filter to make herself look younger upsets [ — Real-time streaming video fakery – this is where we live now. – ED]

AP Explains: How big a threat is an electromagnetic attack? As Electromagnetic Pulse Techonlogy Ever Been Used?  In the 1962, during the Cold War, the U.S. detonated a nuclear weapon above the atmosphere over the Pacific Ocean, and the experiment — known as Starfish Prime — knocked out power to traffic lights and telecommunications in parts of Honolulu, illuminating the sky and even leading hotels to host viewing parties, according to news reports.

MIT Media Lab Disobedience Award   – Nomination Form  We are now accepting nominations for the third annual MIT Media Lab Disobedience Award, a no-strings-attached $250,000 cash prize. This award will go to a person or group engaged in what we believe is an extraordinary example of disobedience for the benefit of society. Award winner(s) will be announced in November, 2019.

‘Five Eyes’ nations discuss backdoor access to WhatsApp GCHQ, the UK agency which monitors and breaks into communications, has suggested that Silicon Valley companies could develop technology that would silently add a police officer or intelligence agent to conversations or group chats.

Microsoft, Dell, Concur: Here are all the tech companies doing business with ICE and how much they’re getting paid  Tech companies big and small are making money off Immigration and Customs Enforcement.

Tracking sex: The implications of widespread sexual data leakage and tracking on porn websites


Transhumanism narratives Summer 2019

  • Book – ‘Fall; or Dodge in Hell?’ (Neal Stephenson)
  • TV – ‘‪Years‬ ‪And‬ ‪Years‬’ (HBO/BBC)
  • Music – Transhuman Art Critics

Bridge behavior

If I read the current consensus correctly …

Half the population is planning on migrating to Mars and half are planning to upload their brains to the machine.

related reads –
Why nanotechnology could be the key to smarter machines (FT)