The Sequence Scope: The AI Platform Ecosystem is Getting Crowded
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.
(Core ML concepts + groundbreaking research papers and frameworks + AI news and trends) x 5 minutes, 3 times a week =…
📝 Editorial: The AI Platform Ecosystem is Getting Crowded
Artificial intelligence (AI) has been a very atypical technology trend. Traditionally, in emerging technology markets, startups disrupt incumbents from previous technology cycles until they becomes the new incumbents in the space. In AI, a lot of the innovation in recent years has not been driven by startups but by the research labs of technology giants such as Microsoft, Google, Amazon or Facebook. Those companies have created some of the top AI platform offerings in the market and have also been actively acquiring many early-stage AI startups to increase their pool of data science talent. All those factors have made it incredibly difficult for startups in the AI space to achieve meaningful traction. However, that’s slowly changing.
After some struggle, some areas of the AI market are steadily showing a strong presence of well-capitalized startups. Areas such as interpretability, data labeling or model monitoring seem to be leading the pack. Just this week, startups such as Anomalo( data validation), V7( training) and Truera ( explainability) raised sizable funding rounds which adds two more relevant companies to highly competitive field. Whether those fields remain standalone markets or become features of broader AI platforms remains to be seen. However, for now, the native complexities of the AI space together with this proliferation of well-capitalized startups makes it very difficult for data scientists and companies to keep up with the overall market. Despite the complexity and market fragmentation, the increasing number of AI platform startups are certainly pushing the boundaries of innovations in many segments of the market and finally challenging some of the incumbents platforms. Keeping up with the AI tech market can result both overwhelming and fascinating. Thankfully, there is a newsletter that can help 😉
🗓 Next week in TheSequence Edge:
Edge#49: an introduction to time-series forecasting models; how Uber uses neural networks to forecast during extreme events; Uber’s M3 time-series platform.
Edge#50: a deep dive into HiPlot and Polygames, two unique initiatives recently open-sourced by Facebook Research, that focus on advancing deep learning research with a specific focus on the PyTorch ecosystem.
Now, let’s review the most important developments in the AI industry this week.
🔎 ML Research
Privacy in Large Language Models
Several AI powerhouses such as Google, OpenAI, Apple and Stanford University collaborated in a new research study that shows some concerning security vulnerabilities in large language models such as GPT-3 ->read more on Google Research blog
Using Reinforcement Learning for ML Compiler Optimizations
Google Research published a paper detailing a technique that uses graph neural networks and reinforcement learning to optimize tasks in ML compilers ->read more on Google Research blog
Computer Vision in Small Devices
Microsoft Research published a paper presenting RNNPool, a pooling operator that reduces the size of image representations enabling computer vision models that can run in devices with small memory and computational resources ->read more on Microsoft Research blog
🤖 Cool AI Tech Releases
TensorFlow has released its new update with features that were much required ->read more on TensorFlow blog
💸 Money in AI
- Data intelligence platform BigID raised $70 million in Series D on a valuation of $1 billion. The company claims to be the first to combine ML-based classification, cataloging, correlation, and cluster analysis to provide unmatched insight across legacy and cloud data stores unlocking clients’ data value for data privacy, security, and governance. Hiring.
- Data validation startup Anomalo raised $5.95 million in venture capital. The company differentiates itself from competitors by using its own machine learning tools that let developers customize data validation and set rules to differentiate the company.
- AI-in-Sensors processors company AIStorm raised $16 million. The company creates high-performance processors that offers significant advantages for AIoT edge computing. Its unique approach is in the technology that allows the sensor to couple directly to popular convolutional neural networks.
- AI explainability platform Truera raised $12 million in Series A. Their tools allow to look into model predictions and gain insights into behavior, improving that way the development and operationalization. Model-agnostic, Truera works with all types of regression and classification models, including logistic regression, gradient-boosted and other tree ensemble models, and deep neural networks.
- ML-powered creative toolkit RunwayML raised $8.5 million Series A. They use deep learning techniques to bring a new paradigm to content creation with synthetic media and automation.
- Data training computer vision platform V7 Labs raised $3 million in funding. It offers a complete toolkit for creating robust computer vision AI, maintaining state-of-the-art performance at every step. The company claims that their V7 Darwin leverages automation to create pixel-perfect ground truth for neural networks 10x faster than other tools.