Radical Reads

Applying NLP to healthcare bottlenecks with BirchAI

By Editorial Team


Image Source: BirchAI; Radical Ventures

We are excited to announce Radical Ventures’ latest investment: we are leading a $3.1 million seed round in BirchAI, a Seattle-based startup that recently spun out of the Allen Institute for AI (AI2).

BirchAI is building a cutting-edge natural language processing (NLP) solution to automate call centre activity in healthcare, with an initial focus on pharmaceutical, insurance and medical device customers. These sectors spend tens of billions of dollars annually on call centres whose operations remain largely manual, frustrating and ineffective for patients and customers.

Though still at the seed stage, BirchAI is already seeing remarkable market pull, with multiple Fortune 500 companies signed up as customers.

CTO/cofounder Yinhan Liu, one of the industry’s foremost NLP researchers, leads BirchAI’s technical team. Previously at Facebook AI Research, Yinhan was the lead author on the RoBERTa and BART papers, among the most essential models in NLP today.

We are delighted to be investing alongside AI2, leading healthcare VC Flare Capital Partners, and Washington Research Foundation. Our Bay Area-based Partner, Rob Toews, led the deal for Radical and will join the BirchAI board.

AI News This Week

  • Listen: David Rolnick on how machine learning can help tackle climate change  (The Robot Brains Podcast)

    David Rolnick, a machine learning for climate change pioneer working in Canada, sits down with Pieter Abbeel, a renowned AI researcher, professor at UC Berkeley, and co-founder of Radical portfolio company Covariant, to discuss use cases for AI in weather simulations, ecological monitoring, and predicting natural resource depletion. David organized the first AI event at the United Nations Climate Change Conference, was named a top innovator by the MIT Technology Review, and is currently an Assistant Professor at McGill University, a Core Academic Member at Mila, and a Canada CIFAR AI Chair. (Radical Ventures is an investor in ClimateAi, a climate prediction platform for enterprises. ClimateAi’s industry-leading modelling products are capable of delivering accurate weather and climate predictions beyond current capabilities.)

  • How AI conquered poker  (The New York Times)

    Unlike in chess, backgammon or go, in which the board and the pieces are shared information, in poker, a computer has to interpret its opponents’ bets despite never being certain what cards they hold. Good poker players have always known that they need to balance bluffing and playing it straight – a dynamic that presents challenges for artificial intelligence. In recent years, however, AI-powered systems have made significant strides in tackling the game, and the world of high-stakes poker may be forever changed.

  • Meta’s ‘data2vec’ is a step toward One Neural Network to Rule Them All  (ZDNet)

    There is a flurry of recent activity in the development of multi-modal AI systems, capable of processing multiple data types, such as image, text, and speech audio. The aim is to create a single model that can be applied to image recognition, natural language understanding, or speech detection. The latest achievement is called “data2vec” and comes from researchers in Meta’s AI division. To enable the AI to learn many different tasks, the researchers are building on a standard version of a Transformer model, creating a system that can process multiple data types. (Radical Ventures is an investor in Cohere, a Transformer-based company, led by one of the creators of Transformers, that enables businesses of all sizes to benefit from world class machine learning models.)

  • ‘We Were Blown Away’: How new AI research is changing the way conservators and collectors think about attribution  (Artnet)

    A new study applies deep learning to identify artworks created by different artists, not by looking at the content of a painting but by looking at the 3D texture of its surface. The study improves the AI system’s performance not by developing more advanced algorithms but by using a more informative data type. The initial research is still limited in scope, but the researchers argue that this method is more accurate than previous methods that directly analyzed the coloured image. The researchers hope to identify forgeries and extend their approach beyond oil and acrylic paintings to drawings, watercolours, and sculptures.

  • Moon’s hidden depths uncovered with new algorithm   (Scientific American)

    Researchers have produced a deep-learning algorithm to shed new light on the moon’s permanently shadowed regions. “Our images enable scientists to identify geologic features, such as craters and boulders … as small as three meters across for the first time — a five-to-10-fold increase in resolution compared to previous efforts.” Recent studies suggest these dark areas of the moon contain rich ice reservoirs that could reveal details about the early solar system.

Radical Reads is edited by Leah Morris (Senior Director, Velocity Program, Radical Ventures).