Issue 167: CognitionX Data Science, AI and Machine Learning

In medicine, “if you disrupt things too much, people die.”

Dr Jack Kreindler, a CognitionX advisor, tech evangelist, and medical practitioner (among other things) recently sat down with us for an interview and told us his thoughts about what healthcare should look like in the 21st century. He passionately said that “teaching about the importance of data, personalized responses, and AI to help us make important decisions is really essential” and that “it should be in every curriculum”.

Have you seen any AI companies or research groups focused on educating people for a world where healthcare is administered based on people’s data? Tweet at us @Cognition_X, we’d love to hear your thoughts.


Tabitha UntiltheBotsTakeOver Goldstaub


Assisting pathologists in detecting cancer with deep learning

A pathologist’s report after reviewing a patient’s biological tissue samples is often the gold standard in the diagnosis of many diseases. For cancer in particular, a pathologist’s diagnosis has a profound impact on a patient’s therapy. The reviewing of pathology slides is a very complex task, requiring years of training to gain the expertise and experience to do well.

To address these issues of limited time and diagnostic variability, Google researchers are investigating how deep learning can be applied to digital pathology, by creating an automated detection algorithm that can naturally complement pathologists’ workflow. They have used images (graciously provided by the Radboud University Medical Center) which have also been used for the 2016 ISBI Camelyon Challenge to train algorithms that were optimized for localization of breast cancer that has spread (metastasized) to lymph nodes adjacent to the breast.


Emerging machine intelligence clusters

Eze Vidra, CIO at Antitode, has written the second part in his series about machine learning in Israel. In this post, he discusses the state of machine learning clusters around the world.

He says that Europe has the largest cluster of machine intelligence startups outside of the US. Within Europe, the UK is the top hub, followed by Paris and Berlin). In his next post, he will dive deeper into the machine intelligence cluster in Israel, with 30 additional startups to watch in this space.

Exciting Opportunities

Prizes for robotics, AI and battery innovators to be announced

The Chancellor Philip Hammond will outline plans in Wednesday’s Budget to make hundreds of millions of pounds available to scientists and researchers to develop solutions to hi-tech challenges including AI and robotics, next generation batteries and new techniques for manufacturing medicines.

The Chancellor will also set out out further details on making sure the UK is at the leading edge of 5G mobile phone technology.

Mr Hammond is expected to allocate more than £500 million from the National Productivity Investment Fund (NPIF), which was created in last year’s autumn statement to help innovative UK companies lead the way in the new technologies set to transform the world.

Prediction of the Day

Five AI startup predictions for 2017

Bradford Cross, founding partner at DCVC, the world’s leading machine learning and big data venture capital fund, has written a piece on what he thinks the future holds for AI.

He says, “With AI in a full-fledged mania, 2017 will be the year of reckoning. Pure hype trends will reveal themselves to have no fundamentals behind them. Paradoxically, 2017 will also be the year of breakout successes from a handful of vertically-oriented AI startups solving full-stack industry problems that require subject matter expertise, unique data, and a product that uses AI to deliver its core value proposition.”

The 5 trends he predicts are: 1) bots go bust, 2) deep learning goes commodity, 3) AI is cleantech 2.0 for VCs, 4) MLaaS dies a second death, and 5) full stack vertical AI startups actually work.

Future of Health

Growing tissue grafts on humanoid robots: A future strategy in regenerative medicine?

Humanoid robots may enhance growth of musculoskeletal tissue grafts for tissue transplant applications.

Over the past decade, exciting progress has been made in the development of humanoid robots. The significant potential future value of humanoids includes applications ranging from personal assistance to medicine and space exploration. In particular, musculoskeletal humanoids (such as Kenshiro and Eccerobot) were developed to interact with humans in a safer and more natural way. They aim to closely replicate the detailed anatomy of the human musculoskeletal system including muscles, tendons, and bones.

With their structures activated by artificial muscles, musculoskeletal humanoids have the ability to mimic more accurately the multiple degrees of freedom and the normal range of forces observed in human joints. As a result, it is not surprising that they offer new opportunities in science and medicine. Here, the authors suggest that musculoskeletal robots may assist in the growth of musculoskeletal tissue grafts for tissue transplant applications.

Education and Advice We Rate

On the origin of deep learning

Haohan Wang and Bhiksha Raj have written a thorough 72-page review of the evolutionary history of deep learning models. It covers from the genesis of neural networks when associationism modelling of the brain is
studied, to the models that dominate the last decade of research in deep learning like convolutional neural networks, deep belief networks, and recurrent neural networks. In addition to a review of these models, this paper primarily focuses on the precedents of the models above, examining how the initial ideas are assembled to construct the early models and how these preliminary models are developed into their current forms.

For example, CNN is built on prior knowledge of biological vision system; DBN is evolved from a trade-off of modelling power and computation complexity of graphical models and many nowadays models are neural counterparts of ancient linear models. This paper reviews these evolutionary paths and offers a concise thought flow of how these models are developed, and aims to provide a thorough background for deep
learning. .

Chat Bots, yadda yadda yadda

TaleSpin Is bringing machine learning and  chat bots to physical retail stores 

TaleSpin is a young Delhi-based startup that wants to place chatbots inside stores. Not online stores though – physical retail outlets. The idea, according to TaleSpin’s product head Tanay Dixit, is to help bring some of the benefits of the online shopping experience – such as personalised recommendations – offline as well, at a low cost.

TaleSpin is sold as a software-as-a-service product to retailers, so they can add the technology to their stores at a very low cost, and it uses image recognition and machine learning to build a catalogue of the various products that are available in their store.

You can quickly browse through a visual menu like you would on any e-commerce store, or you can use the chat bot, and tell it what you’re looking for, Dixit explains. You can tell it that you’re looking for a men’s grey t-shirt with short sleeves, and you’ll see cards showing you some options. The bot can then immediately tell you if there is stock in your size, along with an item code, so you can then be directed to the correct product to try and then buy, or if it’s not there, you can be shown similar alternatives, or taken to an ordering page.


Moscow billboard targets ads based on the car you’re driving

Last November if you were driving a BMW x5 or a Volvo XC60 on the highway ringing Moscow, you might have noticed a digital billboard on the side of the road flash an ad just as you approached, one for a new SUV from Jaguar.

If it was evening, you saw an ad with a dark background, helping the car stand out. In bad weather, you saw it maneuvering in the snow.

Targeted advertising is familiar to anyone browsing the Internet. A startup called Synaps Labs has brought it to the physical world by combining high-speed cameras set up a distance ahead of the billboard (about 180 meters) to capture images of cars. Its machine-learning system can recognize in those images the make and model of the cars an advertiser wants to target. A bidding system then selects the appropriate advertising to put on the billboard as that car passes.

I’ve been making some changes based on Feedback. Would love to hear from more of you. Please do click to share your thoughts! 


Published in

Leave a reply

Thank you! Your subscription has been confirmed. You'll hear from us soon.

Log in with your credentials


Forgot your details?

Create Account