Explaining Deep Learning
June 30, 2018
I read an interesting analysis of one approach to deep learning. “The Matrix Calculus You Need For Deep Learning.” This is a useful write up and one worth reading. I want to point out that I noted one statement as quite interesting.
Here’s the statement I highlighted:
How does the product xy change when we wiggle the variables?
The operative words are:
- Change
- Wiggle
- Variables
My thoughts ran along this rather simplistic line:
If humans wiggle variables to induce change, then automated systems need a cowpoke to ride herd.
If software operations wiggle variables to induce change, then inputs can have an impact.
Black box methods are difficult to figure out because humans are not keen on revealing what the cowpokes do to get the system operating in an acceptable manner. If the black box is automatic and really smart, humans have a tough time figuring out what occurrence A generated action B.
Figuring out if algorithms and numerical recipes are biased may prove to be a challenge. Explaining reveals how. Not explaining may reveal that smart methods are only as smart as the cowpokes who herd the numbers.
Stephen E Arnold, June 30, 2018
Amazon: Information and Its Pharma Play
June 29, 2018
Amazon sells quite a few health related products. One I found interesting is Amazon’s hydrogen peroxide. That’s an interesting chemical, and I wondered who has order a higher concentration version. I assumed that Amazon could answer that question, among others; for example, who bought certain books describing the use of the compound.
I thought about Amazon’s health products when I read “Amazon to Buy Online Pharmacy PillPack, Jumping Into the Drug Business.” I think the deal is an interesting one, but my view is different. Hey, I live in rural Kentucky, one of the states associated with opioid abuse.
The news about the deal had an immediate impact on the outfits dedicated to putting store fronts selling soft drinks, batteries, and snacks on every corner. Oh, I almost forgot to mention that these Walgreen-type outfits are distribution points for medications prescribed by doctors. Amazon seems to be serious about disrupting how many people pick up their medicines, fill out forms for insurance, and provide endlessly repetitive details like their date of birth, insurance data, and home address.
Amazon assumes that it can make this traditional doctor-pharmacy-consumer business more efficient and make a buck along the way.
Several thoughts crossed my mind as I read the NYT story. Not surprisingly, these have not be mentioned in the NYT story or the other coverage I have scanned.
Let me share a handful of questions which struck me after reading the NYT write up.
- Will Amazon be able to determine what individuals are acquired specific medicines and place these data on a timeline?
- Will Amazon be able to determine what medications are flowing to specific geographic regions; for example, specific zip codes within the Commonwealth of Kentucky or any other geocoded area?
- Will Amazon be able to pinpoint the physicians, dentists, etc. who are prescribing specific medicines and array those data on a timeline or a compound output like those available from Palantir Gotham or IBM Analysts Notebook?
- Can Amazon cross correlate these medicine related data with other specific Amazon customer behavior?
- Can Amazon provide insights about possible improper script issuance, medical fraud, or other similar activity?
I assume that Amazon may not have these questions. Amazon sells books.
I want to raise a final question:
What if Amazon can process these drug related data in a way that reveals patterns, identifies abusers, or provides data to flag medical fraud?
- If the answer is no, that’s okay with most Amazon customers.
- If the answer is yes, that suggests a number of other questions.
Amazon is an interesting company indeed.
Stephen E Arnold, June 29, 2018
Search History? No Big Deal Maybe
June 29, 2018
What you search for leaves a digital footprint, or more accurately, a fingerprint. So much identifying data is left behind in your search history. However, there are some angles to this predicament many people are overlooking. We realized just how much bad information people are getting after reading a recent Pagal Parrot article, “Searching These Five Things Can Make Trouble For You.”
This odd little story seems to really give some elementary advice on what not to search for, like:
“#2 Your Name- It’s not a big secret that in this era of the internet our privacy questioned. If you try to Google most probably you will get stumble upon some unpleasant results, bad photos of you, outdated information, irrelevant content. we take such things way too seriously. If you find something like this, you want to delete it.”
This is a little obscure, considering there are such worse implications of your search history. For one, it informs all the bots what is sent through your social media feed. So, for example, a simple search about fake news might just land you with a glut of bogus stories. Thankfully, there is better advice out there than not searching your name, like how to wipe your Facebook and Google search history so that you aren’t fed to the algorithm monsters. Much more practical, in our book!
Patrick Roland, June 29, 2018
IBM Demo: Debating Watson
June 29, 2018
IBM once again displays its AI chops—SFGate reports, “IBM Computer Proves Formidable Against 2 Human Debaters.” The project, dubbed Project Debater, shows off the tech’s improvements in mimicking human-like speech and reasoning. At a recent demonstration, neither the AI nor the two humans knew the topics beforehand: space exploration and telemedicine. According to one of the human participants, the AI held its own pretty well, even if it did rely too much on blanket statements. Writer Matt O’brien says this about IBM’s approach:
“Rather than just scanning a giant trove of data in search of factoids, IBM’s latest project taps into several more complex branches of AI. Search engine algorithms used by Google and Microsoft’s Bing use similar technology to digest and summarize written content and compose new paragraphs. Voice assistants such as Amazon’s Alexa rely on listening comprehension to answer questions posed by people. Google recently demonstrated an eerily human-like voice assistant that can call hair salons or restaurants to make appointments…But IBM says it’s breaking new ground by creating a system that tackles deeper human practices of rhetoric and analysis, and how they’re used to discuss big questions whose answers aren’t always clear. ‘If you think of the rules of debate, they’re far more open-ended than the rules of a board game,’ said Ranit Aharonov, who manages the debater project.
The demo did not declare any “winner” in the debate, but researchers were able to draw some (perhaps obvious) conclusions: While the software was better at recalling specific facts and statistics to bolster its arguments, humans brought more linguistic flair and the power of personal experience to the field. As for potential applications of this technology, IBM’s VP of research suggests it could be used by human workers to better inform their decisions. Lawyers, specifically, were mentioned.
Keep in mind. Demo.
Cynthia Murrell, June 29, 2018
What Has Happened to Enterprise Search?
June 28, 2018
I read “Enterprise Search in 2018: What a Long Strange Trip It’s Been.” I found the information presented interesting. The thesis is that enterprise search has been on a journey almost like a “Wizard of Oz” experience.
The idea of consolidation, from my point of view, boils down to executives who want to cash in, get out, and move on. The reasons are not far to seek: Over promising and under delivering, financial manipulations, and positioning a nuts and bolts utility as something that solves information problems.
Some, maybe many, licensees of proprietary enterprise search systems may have viewed their investment as an opportunity that delivered an unexpected but inevitable outcome. Where is that lush scenery? Where’s the beach?
The reality is that enterprise search vendors were aced by Shay Banon. His Act II of Compass: A Finding Story was Elasticsearch and the company Elastic. Why not use free and open source software. At least the code gets some bugs fixed unlike old school proprietary enterprise search systems. Bug fixes? Yep, good luck with your Fast Search & Retrieval ESP platform idiosyncrasies.
The landscape today is a bit like the volcanic transformation of Hawaii’s Vacationland. Real estate agents will be explaining that the lava flows have created new beach views, promising that eruptions are a low probability event.
The write up does not highlight one simple fact: Enterprise search has given way to “roll up” services or what I call “meta-plays.” The idea is that search is tucked inside systems like Diffeo, Palantir Gotham, and other “intelligence” platforms.
Why aren’t these enterprise grade solutions sold as “enterprise search” or “enterprise business intelligence and discovery solutions”?
The answer is that the information retrieval nest has been marginalized by the actions of vendors stretching back to the Smart system and to the present with “proprietary” solutions which actually include open source technology. These systems are anchored in the past.
Consider Diffeo?
Why offer enterprise search when one can provide a solution that delivers information in context, provides collaboration tools, and displays information in different ways with a single mouse click?
Amazon Intelligence Gets a New Data Stream
June 28, 2018
I read “Amazon’s New Blue Crew.” The idea is that Amazon can disintermediate FedEx, UPS (the outfit with the double parking brown trucks), and the US Postal Service.
On the surface, the idea makes sense. Push down delivery to small outfits. Subsidize them indirectly and directly. Reduce costs and eliminate intermediaries not directly linked to Amazon.
FedEx, UPS, and the USPS are not the most nimble outfits around. I used to get FedEx envelopes every day or two. I haven’t seen one of those for months. Shipping vis UPS is a hassle. I fill out forms and have to manage odd slips of paper with arcane codes on them. The US Postal Services works well for letters, but I have noticed some returns for “addresses not found.” One was an address in the city in which I live. I put the letter in the recipient’s mailbox. That worked.
The write up reports:
The new program lets anyone run their own package delivery fleet of up to 40 vehicles with up to 100 employees. Amazon works with the entrepreneurs — referred to as “Delivery Service Partners” — and pays them to deliver packages while providing discounts on vehicles, uniforms, fuel, insurance, and more. They operate their own businesses and hire their own employees, though Amazon requires them to offer health care, paid time off, and competitive wages. Amazon said entrepreneurs can get started with as low as $10,000 and earn up to $300,000 annually in profit.
Now what’s the connection to Amazon streaming data services and the company’s intelligence efforts? Several hypotheses come to mind:
- Amazon obtains fine grained detail about purchases and delivery locations. These are data which no longer can be captured in a non Amazon delivery service system
- The data can be cross correlated; for example, purchasers of a Kindle title with the delivery of a particular product; for example, hydrogen peroxide
- Amazon’s delivery data make it possible to capture metadata about delivery time, whether a person accepted the package or if the package was left at the door and other location details such as a blocked entrance, for instance.
A few people dropping off packages is not particularly useful. Scale up the service across Amazon operations in the continental states or a broader swatch of territory and the delivery service becomes a useful source of high value information.
FedEx and UPS are ripe for disruption. But so is the streaming intelligence sector. Worth monitoring this ostensible common sense delivery play.
Stephen E Arnold, June 28, 2018
Looking for News Like the Hawaii Volcano Eruption?
June 28, 2018
With the problem of fake news online, the news itself has often made headlines of late. We’ve noticed a couple different news-related moves from big players: TechCrunch reports on Google’s recent project in, “Google Experiments in Local News with an App Called Bulletin.” We learn that Apple, meanwhile, plans to integrate its recent acquisition, magazine aggregator Texture, into Apple News and an upcoming subscription service in, “Apple Said to Plan a ‘Netflix for News’ in Latest Push” at the Daily Herald.
Google’s Bulletin is a place for members of a community to post local news and event notices. TechCrunch’s Sarah Perez suspects it’s also another attempt by Google to squeeze into the Social Media space. She observes:
“The move to delve into local news would have Google competing with other services where people already share news about what’s happening locally. Specifically, people tend to tweet or live stream when news is breaking …. Meanwhile, if they’re trying to promote a local event … it’s likely that they’ll post that to the business’s Facebook Page, where it can then be discovered through the Page’s fans and surfaced in Facebook’s Local app. And if Google aims to more directly compete with local news resources like small-town print or online publishers or Patch, it could have a tougher road. Hyperlocal news has been difficult to monetize, and those who have made it work aren’t likely interested in shifting their limited time and energy elsewhere.”
Over at the Daily Herald, reporters Mark Gurman and Gerry Smith Bloomberg note that Apple cut 20 Texture workers shortly after acquiring the company, but we’re cautioned against reading too much into that. The article notes:
“An upgraded Apple News app with the subscription offering is expected to launch within the next year, and a slice of the subscription revenue will go to magazine publishers that are part of the program, [sources] said. … A new, simplified subscription service covering multiple publications could spur Apple News usage and generate new revenue in a similar manner to the $9.99 per month Apple Music offering.”
Will enough folks pay per month for news, like they do for (other) online entertainment? Perhaps now, when it is prudent to be skeptical, people are willing to pay up to 10 bucks a month for a trusted name. We shall see.
What’s clear is that when one looks for “news” about the Hawaii volcano, few of the online news services are useful. In order to keep up to date, old fashioned search, review, and read processes are the norm. Want current videos about the eruption on YouTube? Good luck with that too. Comprehensiveness may be impossible with free or low cost services, but chronological tags and spam content filtering could be helpful.
For slow moving lava, what’s the rush?
Cynthia Murrell, June 28, 2018
Health Care: Data an Issue
June 28, 2018
Healthcare analytics is helping doctors and patients make decisions in ways we never could have dreamed. From helping keep your heart healthy to deciding when to have major surgery, analytic numbers make a big impact. However, that data needs to be perfect in order to work, according a recent ZD Net story, “Google AI is Very Good at Predicting When a Patient is Going to Die.”
According to the story:
“As noted, 80 percent of the effort in creating an analytic model is in cleaning the data, so it could provide a way to scale up predictive models, assuming the data is available to mine…. “This technique would allow clinicians to check whether a prediction is based on credible facts and address concerns about so-called ‘black-box’ methods that don’t explain why a prediction has been made.”
This really illustrates how powerful clean data can be in the health field. However, cleaning data is just about the most misunderstood wallflower in the often tedious world of machine learning and data science—not just in healthcare. According to Entrepreneur magazine, the act of filling in blanks, removing outliers, and basically looking at all the data to make sure it will be accurate, is the most important part of the process and also the hardest role to fill on a team.
Garbage in, garbage out. True decades ago. True today. How do we know? Just ask one of IBM Watson’s former health care specialists. Querying patients who were on the wrong end of a smart output may be helpful as well.
Patrick Roland, June 28, 2018
What Can We Learn from Chatbots?
June 27, 2018
I assume that many people will learn from chatbots in general and Microsoft’s in particular.
Digital assistants and chatbots are only as smart as their programmed AI. They are also incredibly easy to hijack, so they do lewd and racist things. Take for instance, the 2016 Microsoft fiasco discussed in the MIT Technology Review article, “Microsoft’s Neo-Nazi Sexbot Was A Great Lesson For Makers Of AI Assistants.”
Microsoft designed Tay, an AI chatbot, as an example of how lifelike and smart artificial intelligence has become. Within a day, users figured out how to manipulate Tay’s AI, so the chatbot became racists, a neo-Nazi, and misogynist. I bet Papa Bill Gates was proud! While this was a PR nightmare, Yandex Head of Research and Machine Intelligence Misha Bilenko says it teaches us an important lesson.
“Bilenko said Tay’s bugs—like the bot’s vulnerability to being gamed into learning or repeating offensive phrases—taught great lessons about what can go wrong. The way Tay rapidly morphed from a fun-loving bot (she was trained to have the personality of a facetious 19-year-old) into an AI monster, he said, showed how important it is to be able to fix problems quickly, which is not easy to do. And it also illustrated how much people tend to anthropomorphize AI, believing that it has deep-seated beliefs rather than seeing it as a statistical machine.”
Bilenko explained that chatbots have grown in leaps and bounds since 2016, but do not expect to be chatting up your phone for an intelligent conversation in the next five years. Humans are still the more advanced technology…er…species and our communication is sign of how complex we are. If only we could show off the best of what are species has to offer, instead of Internet trolls.
Whitney Grace, June 27, 2018
Google: No, No More Government Work. Sort Of.
June 27, 2018
Google would have us believe it’s rebuffing the military’s advances, but the PR ploy may amount to no more than semantics. Pando tells us, “Google Distances Itself from the Pentagon, Stays in Bed with Mercenaries and Intelligence Contractors.” Reporter Yasha Levine observes:
“Earlier this week, Google made a big show of refusing DARPA funding for two robotics manufacturers it purchased, even though the companies themselves were financed with plenty of DoD cash. It’s a nice gesture, and one that was welcomed by those who want Silicon Valley to be free of government interference. Unfortunately, while a crowd-pleasing announcement is good for Google’s public image, it does nothing to change the company’s long and ongoing history of working closely with US military and surveillance agencies.”
Levine details Google’s extensive history with intelligence agencies around the world and its more recent dealings with military contractors from the likes of Lockheed Martin to much smaller operations. Meanwhile, the company ramped up its lobbying efforts and hired some folks with ties military and intelligence experience into its sales department. See the article for more details, including discussion of the mysterious Blackbird. In his closing, Levine emphasizes:
“It’s important that we — the millions of people who trust our data to Google every day — understand what Google is: It isn’t a traditional Internet service company. It’s not even, as mild cynics are fond of saying, an advertising company. Google is a whole new type of beast: a global for-profit surveillance company with a mission to funnel as much of our daily life in the real and online world through its servers. The purpose: to track, analyze and profile us as deeply as possible — who we are, what we do, where we go, who we talk to, what we think about — and then constantly figure out ways to monetize that intelligence.”
We wonder if the “real” journalists pay any attention to Google’s teaming with In-Q-Tel. Probably not important.
Cynthia Murrell, June 27, 2018
-
- Subscribe to Beyond Search
Feature archive
News archive
-
