🖼 AI Incumbents and Their Favorite ML Frameworks

Relevant ML news, Research Papers, Cool Tech, and Money in AI

📝 Editorial 

Traditionally, open-source innovation in technology markets is targeted to challenge the incumbents in the field and boost a new generation of startups that can capitalize on the innovations of contributors to open-source projects. From the OS wars to recent cloud or big data trends, technology markets are full of examples in which an open-source project challenges big technology incumbents, which then react by creating their own alternative. Machine learning (ML) has challenged those conventional dynamics. In the ML space, large incumbents have not only been the innovation behind some of the most popular open-source stacks but they continue to actively invest in growing those projects.  

Let’s take the example of ML development frameworks, which are one of the foundational components of the artificial intelligence (AI) ecosystem. Despite the proliferation in the number and types of ML frameworks in the market, the last couple of years have concentrated the race in a handful of stacks that have become the favorites of data scientists and machine learning engineers. Interestingly enough, these frameworks have received the endorsements of some of the biggest AI labs in the world, which are investing a lot of resources to advance these open-source machine learning stacks. Here is a quick view of how big AI incumbents are supporting open source ML frameworks:  

  • Facebook, Microsoft, OpenAI have thrown their support behind PyTorch. 

  • Google is all in with TensorFlow and Keras. 

  • Amazon is very invested in MXNet. 

This level of support from large AI labs has been super important for advancing open-source ML frameworks. Not only AI incumbents are actively committing to engineering and financial resources to support their favorite ML stacks, but they are also adopting those frameworks in some of the most complex ML scenarios in the world.      


🔺🔻TheSequence Scope – our Sunday edition with the industry’s development overview – is free. To receive high-quality content about the most relevant developments in the ML world every Tuesday and Thursday, please subscribe to TheSequence Edge 🔺🔻

🗓 Next week in TheSequence Edge:

Edge#95: DQN reinforcement learning models; DeepMind’s RL agent that masters Quake III; OpenAI Gym – one of the most important technology stacks for modern reinforcement learning solutions. 

Edge#96: deep dive into Molecula, a cloud-based enterprise feature store.

Now, let’s review the most important developments in the AI industry this week

🔎 ML Research

Amazon Research published an insightful blog post detailing how Alexa leverages self-learning to correct errors and improve its knowledge ->read more on Amazon Research blog

GANs for Robotics 

Google Research introduced two GAN models for robotic training ->read more on Google Research blog

A Dataset About the Human Cortex 

Google Research published a paper describing the H01 dataset, a petabyte rendering of brain tissue->read more on Google Research blog

📌 Events

Awesome Summer Marathon by DataTalks.Club with two tracks: Career in Data and ML in production. Listen to specialists from Amazon AI, Databricks, YData, Shopify, and others. It’s online and entirely free!


🤖 Cool AI Tech Releases

PyTorch at Facebook 

Facebook AI Research (FAIR) published a detailed blog post explaining the use of PyTorch across the organization ->read more on FAIR team blog

💬 Useful tweet

Very cool!

Follow us on Twitter

💸 Money in AI

ML news

  • MLOps startup Iterative.ai raised $20 million in a Series A round led by 468 Capital and Florian Leibert. The company develops data science and AI engineering workflows, building Data Version Control (DVC), Continuous Machine Learning (CML), and other developer tools for ML. They just launched DVC-Studio, an open-source version control system for ML Projects.

AI in real world

  • 🦄 Execution management company Celonis raised a $1 billion Series D round. Celonis provides a proprietary process mining AI engine, allowing it to extract data from clients’ operational systems, create a "Digital Twin" and then process a model that helps understand the business processes with the power of AI.

  • Revenue intelligence platform Gong.io raised $250 million (with valuation at $7.25 billion) in a Series E funding round led by Franklin Templeton. Capturing every customer interaction, Gong uses AI to dig insights that help revenue teams improve their results by making data-based decisions.

  • Low-code conversational AI startup Cognigy raised a $44 million Series B funding round led by Insight Partners. The company powers intelligent voice and chatbots, enabling enterprises to have natural language conversations with their users on any channel and in any language.

  • Global wildfire detection startup OroraTech raised $7 million in a Series A round led by Findus Venture and Ananda Impact Ventures. OroraTech’s platform uses a custom wildfire detection algorithm and data from 14 different satellites to notify users of fires in their area of interest and to support immediate firefighting.

  • Drug synthesis AI platform Molecule.one raised a $4.6 million round led by Atmos Ventures. Its ML systems are used for synthesis planning – the task of predicting how to physically make a given molecule, which is crucial for creating new drugs and treatments.