Name that tune, Now Playing!

A recent note on the Google AI blog discusses the company’s use of a deep neural network for music recognition on mobile devices. As it brings extreme-scale noodling (convolution) to bandwidth limited devices (smart phones) it could be a breakthrough on par with MPEG and JPEG, which dramatically transformed music distribution beginning in the 1990s. It’s known as Now Playing, and it can use a sequence of embeddings that run your music against its network and recognize the song, while conserving energy on the device. Each embedding has 96 to 128 dimensions. An embedding threshold is raised for obscure songs – which is the town where I live. I guess when you look at what Google has done with Search, it shouldn’t be that surprising – but the idea that so much of the work occurs on the Thing (device), is pretty astounding. I  asked it ‘what’s that song’ and it got it right. Slam dunk. “Ride Your Pony” by Lee Dorsey. Now, Shoot! Shoot! Shoot! Shoot!  Jack Vaughan

RELATED 

https://ai.googleblog.com/2018/09/googles-next-generation-music.html  - Google AI Blog, Sept 14, 2018

Speaking of Name That Tune – why not a little vignette from the time when Humans Walked the Earth?

Opaque algorithms with singular purpose

Today's article  by DiResta:

"Opaque algorithms with their singular purpose—“keep watching”—coupled with billions of users is a dangerous recipe

...

Most RT viewers don’t set out in search of Russian propaganda. The videos that rack up the views are RT’s clickbait-y, gateway content: videos of towering tsunamis, meteors striking buildings, shark attacks, amusement park accidents, some that are years old but have comments from within an hour ago. This disaster porn is highly engaging; the videos been viewed tens of millions of times and are likely watched until the end. As a result, YouTube’s algorithm likely believes other RT content is worth suggesting to the viewers of that content—and so, quickly, an American YouTube user looking for news finds themselves watching Russia’s take on Hillary Clinton, immigration, and current events. These videos are served up in autoplay playlists alongside content from legitimate news organizations, giving RT itself increased legitimacy by association.

...

The social internet is mediated by algorithms: recommendation engines, search, trending, autocomplete, and other mechanisms that predict what we want to see next."

https://www.wired.com/story/free-speech-is-not-the-same-as-free-reach/

The Cavalcade of Falsehood

Notes - The Fog falls on Facebook-When Deception becomes the norm - Manipulation of public opinion is nothing new. But assorted characteristics of the social media platform Facebook make of public opinion a very new mixer. But for context let’s not go back as far as the Lusitania – lets go to the Ukraine.

The downing of a Malaysian jet liner in July 2014 may have pretty immediately been laid at Russia’s door, but it did not catch the Bear without the wherewithal to respond in ingrained fashion. A disinformation campaign was soon launched, with cascading and contradictory alternative stories of lies, half-truths and some truths.

The formula is becoming familiar now in America, but in 2014 it was already very familiar to countries bordering Putin’s dark lair. Reading the report “Fog of Falsehood: Russian Strategy of Deception and the Conflict in Ukraine” bears this out. It is in fact an analysis of the Kremlin kiddos strategic deception, which has had incredible influence in America, and which is at the point of  shaking the foundations of friends and families.

Like others, the Fog of Falsehood editors, Katie Pynnoniemi and Adras Racz, point to Putin’s pre-history, the Soviet Era, to gain historical view on deep trickeration as a tact.  -Jack Vaughan

Demolition Derby World of Data

Weapons of Math Destruction by Cathy O’Neil cuts to the chase when it comes to big data and its very dark side, which she saw first hand working as a quant in the run up to 2008, and thereafter as a data scientist in e-commerce.

What she saw was the housing crisis the collapse of major financial institutions ... all had been aided and abetted by mathematicians wielding magic formulas.

That was 2008. But there was no let up there after.
   

New mathematical techniques were used to churn through petabytes of information much of its created from social media or you e- commerce websites -mathematicians studied desires movements and spending power they were predicting trustworthiness and calculating potential.

But, as O’Neil, author of the Mathbabe blog, documents ably: The models encoded human prejudice.

She enumerates the differences between a study of small classroom and the big data they work on at Google. This is something I see regularly, as I cover big data as it pertains to business enterprises. People see  what Google does and mistakenly extrapolate the company’s proven success to their own potential outcome. They feel good, because they think they  are doing something akin to what the great disruptor of advertising did.

Systems like that can be improved via Feedback but systems like the one she discusses in the Washington school system and which she says is similar to other weapons of mass destruction she considers in her book, generally lag in terms of feedback. They also create fail-safe false premises for their syllogism.

Author writes:

You cannot appeal to a WMD that’s part of their fears sun power they do not listen nor do they bend their DEF not only to charm threats and cajoling but also to logic.
...they define their own reality and use it to justify their results this type of model is Self perpetuating highly destructive and very common. [p.10]

A great example of unfairness is the use of credit scores to decide who gets a job. That has a way of enforcing failures that would cause Horacio Alger a troubled sleep. For me it rather recalls the great moloch of Search Engine Optimization, a dark cottage industry that sells “Google know how” but which is an amazing indisputable black box of Oz.

But the story really begins with the economic crisis of 2008. And the creation of math models that packaged assorted mortgages (as buckets of risk called securities) in ways that proved lethal, complex and resistance to unravelment an underlying assumption was as familiar as any disaster that had come before:

The risk models were assuming that the future would be no different than the past. [p.41]

Subsequently, O'Neal becomes a data scientist for Intent, working on algorithms to predict the better prospects among web sites' visitors. The leap from math models for futures and math models on web sites put her firmly in the realm of big data, which is where Weapons of Math Destruction really begins.
- Vaughan


https://www.amazon.com/Weapons-Math-Destruction-Increases-Inequality/dp/0553418815

Gaming platforms

With Facebook we see algorithms have replaced editorial boards... a lot of people welcome that... but they may have not have entirely thought through the implications. The Facebook and Twitter platforms have been gamed/amplified by clever/nefarious state-backed programmers. A lot of positive work done to engineer the Internet, has, like the snake eating its tail, begun to devour itself. Her work is not "light reading" but Renee DiResta is someone who I find has really thought through this stuff, and is thinking several steps ahead of the bad guys. - Vaughan

Related

https://www.c-span.org/video/?449088-1/senate-intelligence-committee-holds-hearing-social-media-foreign-influence

https://www.youtube.com/watch?v=QOdqXZWs3yg

https://www.nytimes.com/2017/11/12/technology/social-media-disinformation.html

Facebook fights broadcasts of confusion

Facebook continues to be used as a vehicle for disinformation. This is done by publishing provocative news (not always fake, but certainly presented with nefarious gusto)  under false pretenses to influence the division in large population. Facebook said on Tuesday that it had identified several new Iranian and Russian influence campaigns on its platform designed to mislead people in different countries and regions. Able Renee DiResta of the New Knowledge Research Group said "malicious narratives are spreading to mislead people around the world". The news comes on the same day as reports that Microsoft has found Russian government affiliated websites that masquerade as websites of prominent American conservative think tank websites. The saw of confusion cuts in all directions. Know your links, or you may be sharing falsehoods that have suspicious origin and negatively disruptive intention. - Vaughan

How well can neurals generalize across hospitals?

Which features in any quantity influence a convolutional neural network’s (CNN’s) decision? To find the answer in radiology, work is needed, writes researcher John Zech on Medium. The matter gains increased importance as researchers look to ‘go big’ with their data, and to create models based on X-rays obtained from different hospitals.

Before tools are used to crunch big data for actual diagnosis "we must verify their ability to generalize across a variety of hospital systems" writes Zech.

Among findings:

that pneumonia screening CNNs trained with data from a single hospital system did generalize to other hospitals, though in 2 / 4 cases their performance was significantly worse than their performance on new data from the hospital where they were trained.

he goes further:

CNNs appear to exploit information beyond specific disease-related imaging findings on x-rays to calibrate their disease predictions. They look at parts of the image that shouldn’t matter (outside the heart for cardiomegaly, outside the lungs for pneumonia). Initial data exploration suggests they appear rely on these more for certain diagnoses (pneumonia) than others (cardiomegaly), likely because the disease-specific imaging findings are harder for them to identify.

These findings come against a backdrop: An early target for IBM’s Watson cognitive software has been radiology diagnostics. Recent reports question the efficacy thereof. Zech and collaborators’ work shows another wrinkle on the issue, and the complexity that may test estimates of early success for deep learning in this domain. - Vaughan

Related
https://arxiv.org/abs/1807.00431
https://medium.com/@jrzech/what-are-radiological-deep-learning-models-actually-learning-f97a546c5b98
https://en.wikipedia.org/wiki/Convolutional_neural_network
https://www.clinical-innovation.com/topics/artificial-intelligence/new-report-questions-watsons-cancer-treatment-recommendations

DeepMind AI eyes ophthalmological test breakthrough

Eye ball to eye ball with DeepMind.

DeepMind, the brainy bunch of British boffins whom Google pickedup to carry forward the AI torch, has reported in a scientific journal that it succeeded in employing a common ophthalmological tests to screen for many health disorders.

So reports Bloomberg.

DeepMind’s software used two separate neural networks, a kind of machine learning loosely based on how the human brain works. One neural network labels features in OCT images associated with eye diseases, while the other diagnoses eye conditions based on these features.

Splitting the task means that -- unlike an individual network that makes diagnoses directly from medical imagery – DeepMind’s AI isn’t a black box whose decision-making rationale is completely opaque to human doctors, [a principal said].

The group, which encountered controversy over its use of patient data in the past, said it has cleared important hurdles and  hopes to move to clinical tests in 2019.

Related
https://www.bloomberg.com/news/articles/2018-08-13/google-s-deepmind-to-create-product-to-spot-sight-threatening-disease

these deep neural nets just sort of keep getting deeper and bigger

Making intelligence intelligible with Dr. Rich Caruana https://t.co/s3jGUR4QRV

— Jack Vaughan at TT (@JackVaughanatTT) June 5, 2018

hard to open up those many layered neurals.

to wrap your head around a hundred million weights.

that's harder to udnerstand compared to linear regression.

these deep neural nets just sort of keep getting deeper and bigger.

cc: ummings