Bringing AI to the World's App Teams partnered with to help enterprises automate core user experiences faster than ever. Just like humans, the bots naturally recognize screen elements and learn more about your app each time they test. On their YouTube channel, you can watch AI-powered tests executed by bots testing apps from Uber, Lumosity, Porsche, Disney, Candy Crush, and more.

Watch PinkLion Demo Videos

Google AI and DeepMind researchers introduced a new self-supervised learning method called Temporal Cycle-Consistency Learning (TCC) that can produce granular labeling for multi-step processes from videos. TTC leverages temporal alignment between videos to break down continuous actions in videos to develop a semantic understanding of each video frame. Yuqing Li of Synced Review digs into it in Google AI & DeepMind Temporal Cycle-Consistency Learning.

A new paper by Liyuan Liu and other AI researchers introduces RAdam, or “Rectified Adam”, a new variation of the classic Adam optimizer that provides an automated, dynamic adjustment to the adaptive learning rate based on their detailed study into the effects of variance and momentum during AI training. Less Wright of FastAI believes RAdam holds the promise of immediately improving every AI architecture compared to vanilla Adam as a New State of the Art AI Optimizer.

Myle Ott and other researchers from the University of Washington published a paper on RoBERTa, a robustly optimized BERT pretraining approach. RoBERTa achieves state-of-the-art performance on GLUE without multi-task fine tuning, on SQuAD without additional data (unlike BERT and XLNet), and on RACE.The authors show that rigorous tuning of hyperparameters and dataset size can play a decisive role in performance. The study highlights the importance of proper evaluation procedures for all new machine learning techniques. Read more in RoBERTa (arXiv).

Unity’s launch of Obstacle Tower, a video game that provides an effective test of AI expertise in areas like computer vision, virtual locomotion, and planning. The ultimate goal is that these types of new, specially tailored pieces of software will help create smarter AI agents that can learn more complex skills at ever-accelerating rates. Nick Statt of The Verge digs into it in Unity Developed a Video Game Designed to Test AI Players.

The AI-specific report,Who Is Winning the AI Race: China, the EU or the United States?, released by the Center for Data Innovation, details data collected on how AI is doing in six categories: talent, research, development, adoption, data, and hardware. Based on a 100-point scale, researchers found that the U.S. led overall with 44.2 points, China was second at 32.3, and the European Union placed third with 23.5. Chris O’Brien of VentureBeat covers more of the report’s results in U.S. Leads AI Race, with China Closing Fast and EU Lagging.

The National Institute of Standards and Technology, in responding to a February executive order, issued its roadmap for developing AI standards that would guide federal policy and applications. The 46-page document seeks to foster standards strict enough to prevent harm but flexible enough to drive innovation. The plan describes a broad effort to standardize in areas as disparate as terminology and user interfaces, benchmarking and risk management. It calls for coordination among public agencies, institutions, businesses, and foreign countries, emphasizing the need to develop trustworthy AI systems that are accurate, reliable, secure, and transparent. Here’s the full report called U.S. Leadership in AI.

OpenAI plans to release a version of GPT-2, an advanced conversational AI model, that has 774 million parameters compared to a previous version with 124 million parameters. Along with GPT-2, they also shared an open-source legal agreement to help companies that create large AI models to establish their own model-sharing agreements. This may not seem like a big deal, but with more guidance around sharing models with other companies could have a positive impact overall and promote more industry collaboration. Khari Johnson of VentureBeat reports in OpenAI Releases Curtailed Version of GPT-2 Language Model.

Huawei pulled back the curtain on its own proprietary operating system, HarmonyOS, which will be used in an upcoming smart TV and soon other smart home products from the company. They also unveiled in its Shenzhen headquarters a new AI chipset for IoT devices called Ascend 910, in yet another move by the company to reduce its reliance on U.S. components and compete with Qualcomm and Nvidia. Ben Sin of Forbes reports on it more in Huawei Launches Proprietary AI Chipset.

Re-Shaping the Test Pyramid for App Teams, August 28 (webinar)

The Top Ten Misconceptions About AI-First Testing, October 2 (webinar)

StarWEST, September 29 - October 4 in Anaheim, CA

Pacific Northwest Software Quality Conference, Oct. 14 - 16 in Portland, OR

Agile + DevOps East, November 3 - 8 in Orlando, FL

Join thousands of AI and software professionals reading the AI and Software Testing newsletter. A digest of timely, must-read posts by thought leaders and industry media. Enter your email below for future updates.