The Sequence Scope: Learning from Real Machine Learning Practitioners
Weekly newsletter that discusses impactful ML research papers, cool tech releases, the money in AI, and real-life implementations.
The Sequence Scope is a summary of the most important published research papers, released technology and startup news in the AI ecosystem in the last week. This compendium is part of TheSequence newsletter. Data scientists, scholars, and developers from Microsoft Research, Intel Corporation, Linux Foundation AI, Google, Lockheed Martin, Cardiff University, Mellon College of Science, Warsaw University of Technology, Universitat Politècnica de València and other companies and universities are already subscribed to TheSequence.
📝 Editorial: Introducing TheSequence Chat
There is nothing more inspiring to machine learning practitioners than to learn from the thought leaders who are doing real work in this field. Many times, it’s just hard to associate a specific piece of machine learning research or technology with the creators behind the scene. However, learning about the experience gained by researchers, engineers and entrepreneurs doing real machine learning work can result in a great source of knowledge and inspiration. We’d like to introduce to you TheSequence Chat, dedicated to learning from people doing practical machine learning work. Every few weeks, we will publish one of these short interviews, bringing you closer to real machine learning practitioners.
Please meet Justin Harris, the Senior Software Developer at Microsoft Research. We’ve covered his paper Decentralized & Collaborative AI on Blockchain in Edge#35. Justin is currently using his experience in machine learning and crowdsourcing to implement a framework for ML in smart contracts in order to collect quality data and provide models that are free to use. We asked him why decentralization is important for the future of AI. Justin shared with us his vision about incentive mechanisms for decentralized AI architectures. We also spoke about federated learning, the challenges of implementation and its dependence on mobile deep learning, and some other exciting things.
No subscription is needed.
🔺🔻TheSequence Scope — our Sunday edition with the industry’s development overview — is free. To receive high-quality educational content every Tuesday and Thursday, please subscribe to TheSequence Edge 🔺🔻
🗓 Next week in TheSequence Edge:
Edge#45: the concept of Encoder-Decoder Architectures; how Amazon uses Encoder-Decoder to teach Alexa to chat more naturally; and Tf-seq2seq.
Edge#46: deep dive into Pro-ML, the architecture powering machine learning at LinkedIn.
Now, let’s review the most important developments in the AI industry this week.
🔎 ML Research
This week DeepMind unveiled the groundbreaking research behind AlphaFold, a deep learning model that provides a solution to the famous protein folding problem ->read more on DeepMind blog
GANs for Causal Modeling
Microsoft Research published two papers proposing techniques that leverage generative adversarial neural networks for uncovering causal relationships in datasets ->read more on Microsoft Research blog
Facebook AI Research (FAIR) unveiled ReBeL, a bot that mastered several perfect information games such as poker ->read more about on FAIR blog
🤖 Cool AI Tech Releases
The new release of TensorFlow’s library for building recommendation models was released this week ->read more on TensorFlow blog
During the re:Invent conference, AWS unveiled a new chip optimized for machine learning training ->read more in this article from ZDNet
AWS SageMaker New Additions
This week AWS also unveiled important additions to its SageMaker platform, including a new feature store, data preparation services, and a new CI/CD stack ->read more on TechCrunch
AWS QuickSight Q
Another AWS release this week was QuickSight Q, a new machine learning capability that allows users to interact with datasets using natural language->read more about on AWS blog
💸 Money in AI
- This week Element AI, a startup that builds AI services for enterprises, was acquired by cloud-based IT services company ServiceNow for the price of around $500 million, according to TechCrunch. The goal is to boost the AI capabilities of its applications and become a bigger player in the world of automation and AI for enterprises.
- Visual data labeling startup Scale AI has raised $155 million, hitting $3,5 billion in evaluation. Recently, Scale AI also launched Nucleus, an AI development platform that helps organize, curate and manage massive data sets, aiming to develop beyond data labeling.
- Digital freight provider Flock Freight raised a $113.5 million round to accelerate developments of its algorithmic pooling technology for truckload shipping and logistics.
- Drug discovery startup Genesis Therapeutics raised a $52 million A round. The company works at the intersection of modern deep neural network approaches, biophysical simulation and massively scalable computing infrastructure to achieve industry-leading performance in molecular generation and property prediction.
- AI-powered solar software Aurora Solar raised a $50 million round. Using a combination of lidar sensor data, computer-assisted design, and computer vision, the software optimizes solar panel installations and predicts the amount of power the panels will produce as well as the potential energy savings.
- Autonomous robotics for industrial uses startup ANYbotics raised a $22,3 million round A. They are currently working on the third generation of their ANYmal robot, integrating computing units and sensors of increasing sophistication.
- Сustomer service automation platform Ultimate.ai raised $20 million in a Series A round. They present themselves as an AI-first, ‘no code’ tool, helping clients design virtual agents to automate up to 80% of support interactions.
- AI-based observability platform for data pipelines Databand raised $14.5 million in round A. It offers a central monitoring hub and inventory system to collect all pipeline run information and metadata so that users can collaborate and exchange information in an integrated repository for executions, code, data, experiments, and metrics.