recentpopularlog in

Machinelearning

« earlier   
This health startup won big government deals—but inside, doctors flagged problems • Forbes
Parmy Olson:
<p>the spectacle of brash tech entrepreneurs making outsized claims for their products is hardly a new phenomenon. Neither would matter very much except for the fact that Babylon has two contracts with Britain’s National Health Service, which runs one of the world’s largest nationalized healthcare systems. Babylon’s GP At Hand app offers 35,000 NHS patients video calls and access to its triage chatbot for advice on whether to see a doctor. The NHS is also encouraging 2 million citizens in North London to use NHS 111: Online, an app from Babylon that primarily features a triage chatbot as an alternative to the NHS advice line. Neither uses Babylon’s diagnostic advice chatbot, but the company has talked about bringing this feature to its NHS apps, staff say.

The NHS’s motivations are clear and noble: It wants to save money and produce better health outcomes for patients. Britain will spend nearly $200bn on its national healthcare system in 2020, a sum equivalent to about 7% of GDP. That slice of GDP has doubled since 1950, and the country desperately needs to find a way to rein in costs while still providing a benefit that is seen as central to the UK’s social contract. 

Reducing emergency room visits is a logical step, since they cost the NHS $200 on average per visit, a total of $4bn in the past year, while waiting times are increasing and at least 1.5 million Brits go to the emergency room when they don’t need to. Babylon’s cost-saving chatbot could be a huge help. If it worked better. 

There are some doubts, for instance, about whether the software can fulfill one of its main aims: keeping the “worried well” from heading to the hospital. Early and current iterations of the chatbot advise users to go for a costly emergency room visit in around 30% of cases, according to a Babylon staffer, compared with roughly 20% of people who dial the national health advice line, 111. It’s not clear how many patients take that advice, and Babylon says it doesn’t track that data. </p>


Another amazing exposé; one of Babylon's biggest boosters is the current health secretary Matt Hancock. Perhaps he'll read this and think again.
health  babylon  ai  machinelearning 
3 hours ago by charlesarthur
Chainer: A flexible framework for neural networks
A Powerful, Flexible, and Intuitive Framework for Neural Networks

Chainer is a Python-based deep learning framework aiming at flexibility. It provides automatic differentiation APIs based on the define-by-run approach (a.k.a. dynamic computational graphs) as well as object-oriented high-level APIs to build and train neural networks. It also supports CUDA/cuDNN using CuPy for high performance training and inference.

Compare with tensorflow, pytorch, keras, CNTK.
machinelearning  python  framework  library  ai  ml  neuralnetwork 
4 hours ago by tobym
Microsoft Cognitive Toolkit
A free, easy-to-use, open-source, commercial-grade toolkit that trains deep learning algorithms to learn like the human brain.

Handles multidimensional dense or sparse data from Python or C++, and includes a wide variety of neural network types: FFN, CNN, RNN/LSTM, batch norm, and seq2seq with attention, for starters.

The Cognitive Toolkit supports reinforcement learning, generative adversarial networks, supervised and unsupervised learning, automatic hyperparameter tuning, and the ability to add new, user-defined, core components on the GPU from Python. It is able to do parallelism with accuracy on multiple GPUs and machines, and it can fit even the largest models into GPU memory.

Compare with chainer, pytorch, tensorflow, keras
ai  microsoft  machinelearning  ml  tool. 
4 hours ago by tobym
Kevin Systrom is on a mission to rid Instagram of its troll problem | WIRED UK
But machines are only as good as the rules built into them. Earlier this year Rob Speer, the chief scientist of text-analytics company Luminoso, built an algorithm based on word embeddings to try to understand the sentiment of text posts. He applied the algorithm to restaurant reviews and found, oddly, that Mexican restaurants seemed to do poorly. Stumped, he dug into the data. Ultimately, he found the culprit: "The system had learned the word 'Mexican' from reading the web," he wrote. And on the internet, the word "Mexican" is often associated with the word "illegal", which, to the algorithm, meant something bad.
AI  machinelearning  adtech 
5 hours ago by seatrout
Exploring LSTMs
this is a really good explanation
machinelearning 
6 hours ago by aparrish
Introduction to Conditional Random Fields
current math level: this mostly makes sense to me, like 90%
machinelearning  nlproc  statistics  language 
6 hours ago by aparrish
Software 2.0 – Andrej Karpathy – Medium
How neural nets will change the way we write software forever.
ai  machinelearning  programming  software 
7 hours ago by danielpcox
Neural networks as Ordinary Differential Equations
Recently I found a paper being presented at NeurIPS this year, entitled Neural Ordinary Differential Equations, written by Ricky Chen, Yulia Rubanova, Jesse Bettencourt, and David Duvenaud from the University of Toronto. The core idea is that certain types of neural networks are analogous to a discretized differential equation, so maybe using off-the-shelf differential equation solvers will help get better results. This led me down a bit of a rabbit hole of papers that I found very i...
machinelearning  ai  math 
7 hours ago by geetarista
This Health Startup Won Big Government Deals—But Inside, Doctors Flagged Problems | Forbes
- interesting inside tale of Babylon Health. Interesting especially in the light of IBM Watson Health's failure
health  machinelearning  totwitter 
10 hours ago by renaissancechambara
Neural networks as Ordinary Differential Equations
Recently I found a paper being presented at NeurIPS this year, entitled Neural Ordinary Differential Equations, written by Ricky Chen, Yulia Rubanova, Jesse Bettencourt, and David Duvenaud from the University of Toronto. The core idea is that certain types of neural networks are analogous to a discretized differential equation, so maybe using off-the-shelf differential equation solvers will help get better results. This led me down a bit of a rabbit hole of papers that I found very i...
machinelearning 
13 hours ago by colindocherty
Twitter
Our paper using and to detect in the was accepted and is up online!…
drones  oceans  MachineLearning  seaturtles  from twitter_favs
yesterday by kleinsound

Copy this bookmark:





to read