CogWrite Learning Analytics


Text Analytics, NLP, Machine Learning and Artificial Intelligence

Intelligent Systems for Authentic Learning, Instruction and Assessment


diagram of Galilean Moons from notes, 1610 and Starry Messenger

This item and this item are two discussions of the work of Galileo Galilei and his observations in early 1610 of the moons of Jupiter -- a favorite example of the power of evidence. A bit of evidence. An insight. Prediction. Validation. A new world becomes unavoidable.

Address:

5151 California Ave. STE 100, Irvine, CA 92617 -- Harry dot Layman at CogWrite dot com

Book a meeting or discussion with Harry here.

More info here

More info here





Machine-learning promises to shake up large swaths of finance - May 25, 2017

This article in The Economist talks about new application areas, including machine learning for text has some interesting take-aways. My thoughts are here.

AI Is China Outsmarting America in A.I.? - May 27, 2017

This piece in the New York Times looks at strong moves by China into AI technology in many forms.

China Bets on Sensitive U.S. Start-Ups, Worrying the Pentagon - March 23, 2017

Another NY Times article looks at Chinese state funding for AI and related technologies.

Rush of Chinese Investment in Europe’s High-Tech Firms Is Raising Eyebro - Sep 16, 2016

This earlier article in the New York TImes kicked off this occasional series on China's new focus on AI investing.

China’s Artificial-Intelligence Boom - Feb 16, 2017

Also making the same broad point in The Atlantic, this article with many Andrew Ng quotes. Author Sarah Zhang, a staff writer at The Atlantic, gave voice to a series of Ng's views on China and Chinese work in AI vs. the US / English-speaking world (yes, many aspects were collapsed into three or so broad generalizations) that would be interesting to learn more about. Some - intense competition - for example, rang true (but not as a unique feature...) while others seemed weak -- for example "Yet as the research matures in China, Ng says, it is also becoming its own distinct community. After a recent international meeting in Barcelona, he recalls seeing Chinese language write-ups of the talks circulate right way. He never found any in English.". Hard to imagine many international conferences on AI where write-ups of the talks were unavailable, but hey, "he never found any". Hmmm.

Interestingly, this article in MIT Technology review from January 4, 2017, 5 Big Predictions for Artificial Intelligence in 2017, calls the recent Neural Information Processing Systems conference in Barcelona the banner AI academic gathering - but other than talking up a big Chinese AI push (one of the 5 Big Predictions... perhaps a bit late), and a bizarre fake robot / AI startup launch party designed to draw attention to an excess of hype around AI (?), didn't shed much light on the event. But for technical content, it does seem to have been a cornucopia of content as usual...

The Algorithm will see you now - April 3, 2017

This article in The New Yorker (on the web site they changed the title to A.I. VERSUS M.D. -- to tweak doctors more gently?) is an exploration of the use of AI in medicine. Radiology is at the forefront of this, and the issues and reactions, the debates and discourse, reflects all that is good and bad about progress in generalpresented with a concise set of examples and issues. But for a few too many "straw man" arguments, many of the questions are interesting, even as the overall direction of how better systems will improve tools for doctors, just like they are doing for everyone else, is increasingly clear.

MD Anderson Benches IBM Watson In Setback For Artificial Intelligence In Medicine - Matthew Herper - Feb 19, 2017

This article chronicles an intriguing story about a gold-plated, IBM + Accenture + Oncology staff expert-led effort to apply AI to specific challenges in cancer diagnosis and treatment, but also (and you can see how this could be messy) a growing list of objectives. Sixty million dollars and a few short years later, outside expert reports about procurement practices and questionable side deals, limited actual use (or validation) of the solution and many exciting possibilities are quietly shelved as the effort is closed down, and various folks recede from the scene (usually by promotion). You can almost hear the board of directors standing out front saying Move along now. Nothing to see here... Except, of course, for a fairly public fiasco that makes clear just how challenging big projects, big egos, big companies and big money can be - even with experts at every turn. Will be interesting to see the fallout.

4 Ways to Tweak your data machine learning models - Dec 31, 2016

This note is a short piece that links to an interesting, clear and concise review of how bias and variance, and precision and recall, can point the way to improvements in existing prediction models.

Teleradiology firm announces $13 million investment for cloud-based imaging services - Aug 2, 2016

This piece in an industry publication points to the growth of both cloud-based medical services support and the ever-increasing use of medical imaging technology.

The Case for Machine Learning in Prescribing - Jan 18, 2016

This piece in an industry publication for healthcare executives offers some observations that may sound familiar. Although not a peer-reviewed journal article, the authors do cite an interesting study that documents how machine learning led to improved outcomes and lower costs.

AI Techniques Improving In Important, Enabling Ways - November 28, 2015

This piece in The New Yorker of (November 7, 2015) entitled "Google's New Autoreply Sounds Great!!!!" by Nicola Twilley illustrates improving AI capabilities under development at Google (and more generally elsewhere). As I read about the approach taken, it seems to me that across AI and Machine Learning in general, increasingly, the “vocabulary vectors” that drove the development of LSA techniques in the nineties [with text represented as vectors in a multi-dimensional vocabulary space] are morphing into semantic “thought vectors” which are beginning to allow representation of ideas in more abstract ways - creating new classes of tasks that AI engines may do with greater intelligence. This looks really useful, and directly applicable to my research.

AI Techniques Improving Medicine and Medical Outcomes - Sep 19, 2015

This piece in The Economist is part of a series on new uses of artificial intelligence. Work funded by the California HealthCare Foundation (CHCF) has applied the use of Machine Learning, and an ample supply of construct-relevant evidence (i.e., lots of image data), to the task of diabetic retinopathy diagnosis. The resulting Kaggle competition showed that machine learning can lead to higher quality diagnoses.

AI Techniques Improving Fine Wine Price Prediction - Aug 5, 2015

An audio piece in The Economist pointed to this research reported by University College London. With a good history of price and quality data, expert judgements, and their impact on price, could be judged. Could one interpretation be that price data in the marketplace suggest that "expert scoring" does not reflect actual consumer "scores" by consumers when "expert scoring" is not revealed? The details are interesting.

AI Techniques For Understanding Political Dynamics - Sep 15, 2015

If the very existance of Nate Silver and 538 weren't sufficient, this Economist piece shows how Big Data and Intelligent Systems are helping to understand shifts -- in this case to politic dynamics in California.

Very interested to read about TensorFlow. See here for more.

Seeing is Believing: Bringing Watson Image Analytics to Healthcare - October 19, 2015

This piece by By Shahram Ebadollahi on the "SmarterPlanet" web site is about brining AI and image analytics to Healthcare -- and relevant to IBM's recent acquisition of Merge Healtcare Incorporated and their massive image data bank.

Dysmorphology - Looking for answers, from The Economist - October 22, 2015

This piece, subtitled Face-recognition technology can diagnose developmental disorders from the Economist (October 10, 2015 print edition) highlights yet another way that artificial intelligence in the form of computer-based image analysis shows promise for use in diagnostic medicine.

On The Other Hands... The Economist Notes Data Science harbors a multiplicity of issues - October 24, 2015

This piece, subtitled Honest disagreement about methods may explain irreproducible results from the Economist (October 10, 2015 print edition) highlights how 30 groups of objective and esteemed scientists can look at the same "wodge" (great word!) of data and come to wildly different conclusions. Perhaps there is more to reproducing experimental results than lab technique... Excellent distillation of an issue.

Critical Thinking Resources from IBM.

This page is a gateway to Critical Thinking content on the "ThinkWatson" site. See also

Harry Layman: LinkedIn

Harry Layman's profile on LinkedIn is here. Harry's apps to help beginning readers (first or second language) with word recognition are here.


Mindstorm Tools: An Occasional Blog on technology, management, assessment and education.

The MindstormTools blog is here.

v3.05en 170801a 1.7 cn 151129b test

SGREL project here. Because rubrics.

1 SGREL what is it?.

2 SGREL compared to AES?.

3 SGREL is ... elaboration..