News in AI/ML: what's happened in 2016 so far!
2016 will be a big one for a number of frontier technologies creeping into the mainstream. AI/ML is of course one of them. Here's what's happened so far this year!
Referred by a friend? Sign up here. Want to share? Give it a tweet.
Technology news, trends and opinions
New ‘moonshot’ effort to understand the brain brings artificial intelligence closer to reality. Harvard's Engineering, Brain Sciences, and Cellular Biology departments were awarded $28m to record activity in the brain's visual cortex, map its connections and reconstruct neural circuits in 3D to (somehow) inform the development of biology-inspired learning systems.
Yahoo releases 13.5TB of data under their Yahoo News Feed dataset, which includes 110B events describing anonymised user-news item interactions from 20M users on various Yahoo properties.
Education and underemployment in the age of machine intelligence. TL;DR "Where formal education continues to focus on inculcating basic skills, advanced computational technologies now demand entrepreneurs who are skilled in the creative application of new knowledge."
Baidu looks to artificial intelligence to reduce insurance risks. Ya-Qin Zhang, president of Baidu, shares this news. He also states his worries about the knock-on effects of AI on human behaviour and information retention. It's something I've been thinking about a lot. As AI improves search, for example, we're encouraged to retain less information in our memory because it's only a query away. This frees up mindshare, but what's filling that space?
Legendary filmmaker Werner Herzog is producing a documentary, Lo and Behold: Reveries of the Connected World, with topical content on AI. Here's the trailer.
Baidu's Silicon Valley AI Lab open sources an implementation of their supervised training algorithm for sequence prediction. This technology underlies their Deep Speech 2 end-to-end speech recognition system. Pop quiz: name the large technology company that hasn't jumped on the open sourcing bandwagon?
Research, development and resources
Emerging Technology from arXiv collates the "Best of 2015" research pieces - lots of great work, e.g. how deep learning beats humans in an IQ test or achieves international master level in chess with 72h of training.
Brain4Cars: Car That Knows Before You Do via Sensory-Fusion Deep Learning Architecture, Cornell University, Stanford University and Brain of Things Inc. This work produces a 1180 mile driving dataset (GPS, vehicle dynamics, videos, street maps) and shows how deep learning can be used to anticipate driving manoeuvres 3.5s before they occur in real-time.
Political Speech Generation, UMass Amherst. Using a language model for grammar and topic model for textual consistency trained on US congressional floor debate transcripts, the author is able to automatically generate speeches with either a supportive or opposing opinion on a particular topic.
Highly optimized artificial intelligence and machine learning library written in Swift. An ongoing project that currently includes a feed-forward neural network and a fast matrix library.
Label-free cell cycle analysis for high-throughput imaging flow cytometry (press release here), Swansea University and Broad Institute (MIT and Harvard). This work applies supervised machine learning on morphological features of cells in brightfield and darkfield images to identify cells at different phases of their cell cycle. This contrasts with traditional cell separation approaches that use protein markers to label cells for flow cytometry.
Accelerating AI with GPUs: A new computing model, Nvidia. On the topic of Nvidia, where are the other chip manufacturers in the AI hardware space?
Venture capital financings and exits
Quick summary for 2015? Here is it (more in-depth analysis to come):
In 2015, $3.06bn was invested into businesses using/developing AI through 328 deals around the world. The peak of activity was in Q3, which accounted for $1.57bn on its own.
55% of the deals were at the Seed/Angel stage (median $890k/deal) and 22.5% at Series A (median $5m/deal), 16% at Series B (median $15m/deal). 50% of the activity was in the US, with YC (11 deals), Techstars (7 deals), Data Collective (6 deals), a16z (5 deals), Khosla (5 deals) and NEA (5 deals) being most active.
There were 43 exits occurring on average 3-4 years since first fundraising (average raised $15.1m prior to exit). Apple, Google and Twitter each acquired 2 businesses.
Notable deals in 2016 to date:
Apple completes acquisition of Emotient, a startup analysing human facial expressions to infer emotion. The business raised $8m from Intel Capital and Handbag LLC.
IBM acquired a fraud prevention business IRIS Analytics, which serves card based payments, online banking, transaction banking and mobile payments.
AiCure, a New York-based company that uses facial recognition and motion-sensing technology to visually confirm medication ingestion, raised $12.25m Series A from New Leaf Venture Partners.
Skydio, a computer vision company writing software for drone navigation, raised their $25m Series A from Accel and a16z.
Pathway Genomics raised a $40m Series E from IBM Watson for diagnostic tests and personalised healthcare information product delivered to mobile devices that leverages patient data.
Anything else catch your eye? Just hit reply! I’m actively looking for entrepreneurs building companies that build/use AI to rethink the way we live and work.