In Nightfall’s Up to Speed on AI and Deep Learning, we summarize the latest news, research, technology, and applications of AI and Deep Learning. In this week's edition:
Learn about emerging hardware that will impact the AI and deep learning ecosystem.
Read about techniques that researchers have deployed to harden deep learning networks from inversion attacks.
See how “troll” bots will be used in the future of disinformation campaigns.
Read these stories and other timely AI and Deep Learning news below.
Announcements
Automatically Detect Computer Generated Text With This Chrome Extension
(Lifehacker)
Following the full release of OpenAI’s GPT-2, a large-scale unsupervised language model that generates coherent paragraphs of text, Giulio Starace created a browser extension to detect if a body of text was computer generated. It works by checking whether the generator’s output is similar to the text you are curious about.
AI Is Tearing Up the Dancing Floor Again
(Synced)
For decades machines have been able to understand simple musical features like beats per minute. Now AI is boosting their abilities to the point that they can not only figure out what particular genre of music is playing, but also how to appropriately dance to it.
Intel Unveils oneAPI: What Is it?
(Analytics India Magazine)
At the recently-concluded Supercomputing 2019 event, Intel made its vision for AI loud and clear. oneAPI marks an evolution from today’s proprietary programming approaches to an open standards-based model for cross-architecture developer engagement.
The Cerebras CS-1 computes deep learning AI problems by being bigger, bigger, and bigger than any other chip
(Techcrunch)
The CS-1 is a “complete solution” product designed to be added to a data center to handle AI workflows. It includes the Wafer Scale Engine (or WSE, i.e. the actual processing core) plus all the cooling, networking, storage, and other equipment required to operate and integrate the processor into the data center. It’s 26.25 inches tall (15 rack units), and includes 400,000 processing cores, 18 gigabytes of on-chip memory, 9 petabytes per second of on-die memory bandwidth, 12 gigabit ethernet connections to move data in and out of the CS-1 system, and sucks just 20 kilowatts of power.Blaize AI Emerges From Stealth
(Forbes)
Earlier this month, an AI hardware startup named “Blaize,” previously named “Thinci,” emerged from stealth, having already reached key milestones in four areas: innovative hardware, a comprehensive software stack, a staff of over 325 employees, and most importantly, 15 pilot projects underway in the USA, Europe and Asia.
Research and Tutorials
Domain Knowledge Aided Explainable Artificial Intelligence for Intrusion Detection and Response
(arXiv)
Artificial Intelligence (AI) has become an integral part of modern-day security solutions for its capability of learning very complex functions and handling ”Big Data”. However, the lack of explainability and interpretability of successful AI models is a key stumbling block when trust in a model’s prediction is critical. This leads to human intervention, which in turn results in a delayed response or decision. While there have been major advancements in the speed and performance of AI-based intrusion detection systems, the response is still at human speed when it comes to explaining and interpreting a specific prediction or decision. In this work, we infuse popular domain knowledge (i.e., CIA principles) in our model for better explainability and validate the approach on a network intrusion detection test case. Our experimental results suggest that the infusion of domain knowledge provides better explainability as well as a faster decision or response. In addition, the infused domain knowledge generalizes the model to work well with unknown attacks, as well as open the path to adapt to a large stream of network traffic from numerous IoT devices.
Adversarial Learning of Privacy-Preserving and Task-Oriented Representations
(arXiv)
Data privacy has emerged as an important issue as data-driven deep learning has been an essential component of modern machine learning systems. For instance, there could be a potential privacy risk of machine learning systems via the model inversion attack, whose goal is to reconstruct the input data from the latent representation of deep networks. Our work aims at learning a privacy-preserving and task-oriented representation to defend against such model inversion attacks. Specifically, we propose an adversarial reconstruction learning framework that prevents the latent representations decoded into original input data. By simulating the expected behavior of adversary, our framework is realized by minimizing the negative pixel reconstruction loss or the negative feature reconstruction (i.e., perceptual distance) loss. We validate the proposed method on face attribute prediction, showing that our method allows protecting visual privacy with a small decrease in utility performance. In addition, we show the utility privacy trade-off with different choices of hyperparameter for negative perceptual distance loss at training, allowing service providers to determine the right level of privacy-protection with a certain utility performance. Moreover, we provide an extensive study with different selections of features, tasks, and the data to further analyze their influence on privacy protection.
HAL: Improved Text-Image Matching by Mitigating Visual Semantic Hubs
(arXiv)
The hubness problem widely exists in high-dimensional embedding space and is a fundamental source of error for crossmodal matching tasks. In this work, we study the emergence of hubs in Visual Semantic Embeddings (VSE) with application to text-image matching. We analyze the pros and cons of two widely adopted optimization objectives for training VSE and propose a novel hubness-aware loss function (HAL) that addresses previous methods’ defects. Unlike (Faghri et al. 2018) which simply takes the hardest sample within a minibatch, HAL takes all samples into account, using both local and global statistics to scale up the weights of “hubs”. We experiment our method with various configurations of model architectures and datasets. The method exhibits exceptionally good robustness and brings consistent improvement on the task of text-image matching across all settings. Specifically, under the same model architectures as (Faghri et al. 2018) and (Lee et al. 2018), by switching only the learning objective, we report a maximum R@1 improvement of 7.4% on MS-COCO and 8.3% on Flickr30k.
AI and ML in Society
The Risks of Using AI to Interpret Human Emotions
(Harvard Business Review)
A lot of companies use focus groups and surveys to understand how people feel. Now, emotional AI technology can help businesses capture the emotional reactions in real time. The ultimate outcome is a much better understanding of their customers — and even their employees. But such practices aren’t without risk given the subjective nature of emotions and biases in sampling the data needed to determine what counts as emotional content.
Self-driving trucks likely to hit the roads before passenger cars
(CNBC)
As the hype over self-driving vehicles begins to wear a bit thin, it looks like the technology will come to trucks more quickly than passenger cars. Chinese autonomous driving company Pony.ai, which also has an office in California, has focused on applying the technology to passenger vehicles. Its latest funding round in April brought in $50 million, according to Crunchbase.
To See the Future of Disinformation, You Build Robo-Trolls
(Wired)
Facebook and other social networks already fight against propaganda and disinformation campaigns, whether originating from terrorist groups like ISIS or accounts that are working on behalf of nation-states. All evidence suggests those information operations are mostly manual, with content written by people. Jason Blazakis says his experiments show it’s plausible that such groups could one day adapt open source AI software to speed up the work of trolling or spreading their ideology. “After playing with this technology, I had a feeling in the pit of my stomach that this is going to have a profound effect on how information is transmitted,” he says.