/* ---- Google Analytics Code Below */

Monday, July 20, 2020

Facebook Reveals Holographic Optics for Thin, Light VR Headsets

Technically interesting  next step via Facebook.  Still apparently in early stages. 

Facebook Reveals Holographic Optics for Thin, Light VR Headsets
VentureBeat
Jeremy Horwitz

Facebook has unveiled a holographic optical framework for thinner, lighter virtual reality (VR) headsets, which it expects to be incorporated within high-performance devices for VR and augmented reality. The "pancake optics" system combines thin layers of holographic film with a laser projection system and directional backlights, in order to project either flat imagery or volumetric holograms. The optical system can range from 11 millimeters to less than 9 millimeters thick, while each prototype eye display has a resolution of about 1,200 by 1,600 pixels, with a field of view that is either a 93-degree circle or a 92-by-69-degree rectangle. The researchers said the device is not ready for near-term deployment. .. " 

Sunday, July 19, 2020

AI, CoronaVirus and Investing

I am guessing probably because most learning models depend on machine learning, and the data is not there to predict rare events.  A classic caution for any AI work ...

Even the Best AI Models Are No Match for the Coronavirus in Wired
Many so-called “quantitative funds” that mine historical data to make trading decisions fared poorly in March, when stocks fell sharply amid coronavirus fears.

THE STOCK MARKET appears strangely indifferent to Covid-19 these days, but that wasn’t true in March, as the scale and breadth of the crisis hit home. By one measure, it was the most volatile month in stock market history; on March 16, the Dow Jones average fell almost 13 percent, its biggest one-day decline since 1987.  .... " 

Sankey Diagrams for Visualization

A reader reminded me that I had posted on the Sankey Diagram, a form of 'pie' chart with room for lots of relationships.    A kind of concept diagram too.    Even link to a blog with lots of examples.  Here is a re-post of that, with lots of new examples.

Also, I had forgotten this, Sankey Diagrams are available in Google Charts.

SANKEY DIAGRAMS
A Sankey diagram says more than 1000 pie charts

(Below one example, then lots more at the link) 

SAMPLES
EU Food Flows 2011

This comprehensive and well-structured Sankey diagram on food production, waste and consumption is featured in an article  ‘Quantification of food waste per product group along the food supply chain in the European Union: a mass flow analysis by Carla Caldeira et al. (published as open access article under CC BY license in: Resources Conservation and Recycling · June 2019). The paper “presents a high-level top-down approach to food waste accounting in the European Union.” ... " 

Virtual Try Ons

An idea with a long history that we also examined as part of a broader smart-store effort.  Would not say it has been generally successful to date, but with increased contact-less interaction?  And once the consumer is comfortable, what are the other implications?  During and after pandemics.

Virtual Try-Ons Replacing Fitting Rooms During the Pandemic
The Washington Post
Abha Bhattarai
July 9, 2020

The Covid-19 pandemic has spurred consumers to substitute fitting rooms with augmented reality (AR) using their smartphones and computers. Jewelry brand Kendra Scott is allowing shoppers to try on earrings virtually with their iPhone browsers, while AR provider Zeekit is about to launch the world's largest virtual fitting room, letting shoppers drag hundreds of thousands of clothing items onto photos of themselves. Zeekit co-founder Yael Vizel said shoppers who use the company's artificial intelligence to try on clothing virtually are five times more likely to buy—and keep—the items. Some retailers are pursuing contactless fittings in stores, with mall operator Brookfield Properties recently announcing it will add three-dimensional body scanners that can direct consumers to the best-fitting clothing brands and sizes. Privacy experts worry virtual try-ons leave an exploitable trail of personal and biometric data, and recommend consumers check whether applications are recording this information  ... "

Has AI Lived up to its Hype?

I say No and Yes.   No because it cannot live up to the implied magic of the term AI.   People still expect it to do far more than it can do currently.   I hear the equivalent of: " .. But of course AI could do that.."  Without backing that up.    And I say Yes because it can do far more than what we expected in the 90s ... and actually doing magic that we did not expect it could ever do.     But even these abilities have come with unintended consequences we had not thought of.   Some producing considerable risk for those that trust the algorithms.     All this is not much different from computing in general.   It can all be written into science fiction,  but the further editing into practice can be difficult. 

Has artificial intelligence (AI) really lived up to the hype?
Nikolas Kairinos, CEO and Founder, Fountech discusses how far artificial intelligence (AI) has come over the years and whether it has lived up to the hype surrounding it

Regardless of the industry you work in, you’ve no doubt heard about artificial intelligence (AI) and its potential in changing the world around us. The technology has been a source of debate in the private and public sectors for more than 50 years, and yet it has only been in the last decade that we’ve begun to really see momentum build in the AI space.  ... " 

AI Agents Learning to Communicate

Machines learning and adapting using language.   As suggested,the environment/context is also important.

AI agents can learn to communicate effectively  by University of Gothenburg

A multi-disciplinary team of researchers from Chalmers and University of Gothenburg has developed a framework to study how language evolves as an effective tool for describing mental concepts. In a new paper, they show that artificial agents can learn how to communicate in an artificial language similar to human language. The results have been published in the scientific journal PLOS ONE.

This research lies on the border between cognitive science and machine learning. There has been an influential proposal from cognitive scientists that all human languages can be viewed as having evolved as a means to communicate concepts in a near-optimal way in the sense of classical information theory. The Gothenburg researchers' method for training the artificial agents is based on reinforcement learning, which is an area of machine learning where agents gradually learn by interacting with an environment and getting feedback. In this case, the agents start without any linguistic knowledge and learn to communicate by getting feedback on how well they succeed in communicating a mental concept.

Reconstructing colors

"In our paper we have studied how agents learn to name mental concepts and communicate by playing a several rounds of a referential game consisting of a sender and a listener. We have especially focused on the color-domain which is well studied in cognitive science. The game works as follows; the sender sees a color and describes it by uttering a word from a glossary to the listener which then tries to reconstruct the color.

Both agents receive a shared reward based on how precise the listener's reconstruction was. The words in the glossary have no meaning at the outset; it is up to the agents to agree on the meaning of the words during multiple rounds of the game. We see that the resulting artificial languages are near-optimal in an information-theoretic sense and with similar properties as found in human languages," says Mikael KÃ¥gebäck, researcher at Sleepcycle, and whose Ph.D. dissertation at Chalmers contained some of the results presented in the paper. .... " 

Saturday, July 18, 2020

Amazon Alexa-Live Next Week

Am an invited remote participant at the Amazon Skill Developer event:  Alexa-Live on July 22 next week.  In particular to represent a project with voice assistant application potential.    See more about this event.    If you have anything you would like me to be on the lookout for at this meeting, send me a note.    I will likely post some items of general interest here, especially about novel applications, or ones that indicate changes in direction.

Capital One's AI Assistant

Had not seen this assistant example before.   Much more at the link.

How Capital One’s AI assistant achieved 99% NLU accuracy
VB Staff in VentureBeat

Like almost every other bank, Capital One used to have a basic SMS-based fraud alert system, asking customers if unusual activity that was detected was genuine. But in about 15% of cases, people didn’t text back a simple yes or no.

As Ken Dodelin, VP of Conversational AI Products at Capital One explained in conversation with VentureBeat’s Founder and CEO, Matt Marshall at Transform 2020, a good number of people would write back other explanations — things like “Yeah, that was me” or “Yep. [thumbs-up emoji]” or “Yes, that was last weekend when I was in Philadelphia visiting my sister.”
“We didn’t understand any of it [then],” said Dodelin. “And so that kind of hit the lightbulb for us — wouldn’t it be great if we could understand natural language coming back to us from customers?”

Since then, the vision of building an AI assistant that takes complexity out of money for Capital One customers, and makes money management easier, has been relentless. The intentionally non-gender-specific-named ENO requires no opt-in. Through SMS or push notification, ENO will let you know if you were charged twice for the same amount within a couple of seconds; it might check if an exorbitantly large tip you left was intentional, or if the decimal place was in the wrong spot. Or it could alert you that the free trial you signed up for (and clearly forgot about) is about to expire. ... 

Implications of Working Without an Office

Worth while look based on a broad survey, analysis of the results, with some surprising results. 

The Implications of Working Without an Office   HBR Big Idea
HBR Big Idea
By Ethan Bernstein, Hayley Blunden, Andrew Brodsky, Wonbin Sohn, Ben Waber

In early 2020, the world began what is undoubtedly the largest work-from-home experiment in history. Now, as countries reopen but Covid-19 remains a major threat, organizations are wrestling with whether and how to have workers return to their offices. Business leaders need to be able to answer a number of questions to make these decisions. Primary among them is “What impact has working from home had on productivity and creativity?”

To help answer that question, we decided to explore how employees have fared since they began working virtually. To that end, we started surveying a diverse group of more than 600 U.S.-based white-collar employees during the second half of March and have continued to do so every two weeks since then. (This article is based on results collected through May.) Approximately half of our respondents are women, and half are men; they hail from 43 states; nearly half are married; and more than a third have children. About 40% hold management positions. We have been asking them about their job satisfaction, work engagement, perceptions of their own performance, conflicts with colleagues, stress, negative emotions, and current living situation, among other questions. (A fuller description of the survey’s participants is at the end of this article.)

In addition to the survey answers, participants provided written comments. Further, we collected data on the automatically tracked interactions (based on email and calendars) between a separate group of employees of selected organizations from both before and after they began working from home. (While that project is continuing, the interactions we examined for this article began a year before employees began working remotely and ended two months into the lockdown.) Finally, we validated our initial findings through interviews with senior leaders of other organizations who, at the request of their organizations or clients, were investigating the impact of working from home on productivity and creativity. ....'

Data Dividend Plan

We worked on a number of projects that linked to 'data as an asset'  ideas, and was reminded of this project in Wired.  Its value, limitations are and issues worth re-examining.

Andrew Yang’s Plan to Pay You for Your Data Doesn't Add Up
He wants social media companies to pay you for the data you produce. But loopholes abound, it's too expensive, and other plans like it have failed.

LAST MONTH, Former presidential hopeful Andrew Yang launched a new initiative called the Data Dividend Project  that would force social media companies to compensate users for the use of their data. As Yang told the Verge, “That first day that people get paid their dividend through DDP for All is going to be such a great day.” But the second day, as platform users and advertisers adjust to the new costs, is sure to be a mess. ... ' 

Friday, July 17, 2020

Japanese Robots Learn Food Prep: Resemble Kangaroos

Note the proposal that at first the food will be prepared by humans  remotely, and those motions will be learned.  Then the system will use that learning to prepare foods afterwards.   By robots resembling kangaroos.

Japanese Robot to Clock In at Convenience Store
Reuters
Tim Kelly; Kevin Buckland
July 15, 2020

In August, a robot from Japanese developer Telexistence will prepare food at a FamilyMart convenience store, with plans to expand its deployment to about 20 outlets in Tokyo by 2022. Human operators initially will operate the Model T robots remotely with virtual reality goggles and motion-sensor controls until their artificial intelligence can learn to mimic human movements. Telexistence uses human operators to train its robots, which is far less expensive than complex programming. The Model T vaguely resembles a kangaroo, a design choice meant to put shoppers at ease by avoiding a more unsettling human like appearance.  ..." 

Politeness of/to Machines

Do we expect the same politeness from machines?  Are we polite to them?   They can detect some aspects of our demeanor by the words we use.     Will we begin to hide that?    Or is it useful to adjust the conversation?

Could Your Computer Please Be More Polite? Thank You
Carnegie Mellon University School of Computer Science
By Byron Spice

Carnegie Mellon University researchers have developed an automated method for restructuring nonpolite directives or requests into more polite communications. The team generated a dataset of 1.39 million sentences labeled for politeness—derived from publicly available emails exchanged by employees of long-gone energy firm Enron—for experimental application. Analysis determined the frequency and distribution of words in polite and nonpolite sentences, and the researchers developed a "tag and generate" pipeline to conduct politeness transfers. Impolite or nonpolite words or phrases were tagged, then a text generator replaced each tagged item without altering the meaning of the sentence. The system was able to produce subtler and more realistic restructures over time, and the team has released the labeled dataset for use by others in the hope of encouraging further politeness research.

Revamped GMail to Compete

Used GMail since we tested it in the enterprise.   Later because we collaborated with universities who got it for free.    Always liked it, and the following GSuite, but its design was always a bit quirkily  'academic' as well, making you need to do things in sometimes illogical ways.   Our enterprise went full Microsoft soon afterwards, but it was OK to keep the Google connections as needed.  Will advances in GMail help this?   Notable the look at MS Teams, which is not in considerable use.

A revamped Gmail could bolster interest in G Suite as it looks to take on rivals like Microsoft, Slack and others.  By Matthew Finnegan in Computerworld

Google has unveiled a revamped Gmail that will serve as a hub for collaboration with access to video, chat and shared files available directly from the email client. It’s a move designed to bring Google into closer competition with Microsoft Teams, according to one analyst, as well boosting its position against newcomers.

Google has moved in recent months to integrate its various G Suite apps with Gmail, with its Meet video and Chat team messaging applications gradually being melded with the email client. Indeed, despite expectations in some quarters that team chat will kill email, Google has long seen Gmail as the natural home for collaboration in its portfolio. ... "

Thursday, July 16, 2020

Microsoft Spins off Chinese Chatbot

Somewhat surprising, the demos I had seen looked interesting.

Microsoft Spins Off Chinese Chatbot Xiaoice  by Laurie Sullivan  @lauriesullivan, July 14, 2020
Microsoft is spinning off its AI-based chatbot technology, Xiaoice, into a stand-alone company after nurturing it for about six years.

It is aimed at Chinese consumers and has the personality of a teenage girl. Microsoft will retain a stake in the new company, Xiaoice or Xiaobing, and will license technologies from Microsoft for future efforts in research and development.

The chatbot became available to consumers in China in 2014, but now the service is available across many social networks, with about 660 million users, and offers more than 280 skills.

"Morning News," a Chinese live program in 2015, introduced Xiaoice as a trainee anchor where it took charge of the daily weather report.  ... " 

BMW Sees Car as Digital Platform

Interesting regards the breadth of integrated technology, but the reaction of consumers to 'microtransactions' or payment/subscription for many elemental features in already high cost luxury vehicles, may be very negative.   But it does depend on the value positioning of such premium capabilities.

BMW is going all-in on in-car microtransactions.  Like in-app purchases, but for your car
By Sean O'Kane@sokane1  Jul 2, 2020,  in TheVerge  ... "

And directly from BMW:

BMW Connected Car Beta Days 2020: July will bring a comprehensive software upgrade with numerous new services.  From BMWUsaNews

Justin  Berkowitz
Mariella  Kapsaskis

New services such as the cloud-based navigation BMW Maps, Digital Key for iPhone and an update for the BMW Intelligent Personal Assistant.

Oliver Zipse, Chairman of the Board of Management of BMW AG:
“Digital technologies belong to the core of BMW – because hardware and software are of equal importance for premium cars. Our mission is to integrate advanced digital technologies with highest product excellence to enhance our customers’ experience and driving pleasure even more.”

1.     Overview.
In July 2020, BMW will release its most comprehensive vehicle software upgrade to date, with new and improved features for vehicle connectivity and navigation. These features will be included straight from the factory with every BMW vehicle that has Operating System 7, such as the new BMW 5 Series. And for certain compatible BMW vehicles built before July, already in customers’ hands, these great new updates can even be added with a free over-the-air software upgrade, available from third quarter onwards.

BMW Maps offers a completely new navigation experience: Faster, more intuitive, and more accurate. Destinations can be entered more easily. New cloud-based route calculation brings a significant improvement in performance. BMW Maps also provides contextual information for points of interest, such as ratings, business hours and images. Connected Parking also supports the search for parking spaces with additional services.
The BMW Digital Key, which has just been announced at the Apple Worldwide Developers Conference, turns an iPhone into a fully digital – and secure - vehicle key.
With the upgrade, BMW vehicles receive wireless support of Google’s Android AutoTM.
The BMW Intelligent Personal Assistant is getting a new design and has been enhanced with numerous intelligent functions. For example, the virtual character now turns towards the person talking and can distinguish between addressing the driver and the passenger.
Since 2018, the Remote Software Upgrade has enabled BMW drivers with compatible vehicles to keep their vehicle up to date with the latest software, just like with a smartphone. The new functions can be downloaded and installed over-the-air. This is particularly convenient and extremely fast: even for extensive upgrades, hardly more than 20 minutes of installation time is required.

2.    Digital personalization with functions on demand.

Flexible offers, available immediately, simple to book and easy to use.

BMW sees the car as a digital platform. More than two decades ago, BMW ConnectedDrive paved the way to more individual mobility. Since then, the company has been enabling its customers to benefit from the rapid technological development at any time with the continuously added digital services. With Remote Software Upgrade, BMW has created the technical basis for keeping vehicles up to date, adding additional vehicle functions, and installing them quickly and easily over the air.

BMW already offers its customers digital services and additional vehicle functions in the form of digital after-sales, some of which are deeply embedded in the vehicle's software. In the US, BMW Drive Recorder is offered through a pilot program. In the near future, additional functions will be added that can access the vehicle's existing hardware and software, such as certain comfort functions or driver assistance systems.  ... 


MIT Auto-ID Laboratory

MIT AUTO-ID LABORATORY

The MIT AUTO-ID LABORATORY coined the term Internet of Things (IoT) and traces its roots back to 1999 with the founding of the Auto-ID Center, which laid much of the groundwork for the standardization of RFID technology and the introduction of the EPC. Now a member of the global Auto-ID Labs network, it continues to research the evolution and application of RFID systems, as well as other disruptive Internet of Things technologies.   ... '

Lots of links to efforts still underway, industrial directions and related work.

I was reminded of several people there I connected to, including Sanjay Sarma, Jon Stine and Shoumen Pa Datta.  We worked with them at our own innovation laboratories. 

OpenAI Monitors Machine Learning Efficiency

For now,  the most measurable and definable kind of AI is machine learning,   It uses much data and compute cycles.   Useful to measure the advances in efficiency.  That has been improving. according to OpenAI.  Notably that ' ... algorithmic progress has yielded more gains than classical hardware efficiency ...'.    At least for recent investments.  That is notable for future understanding of where to address our efforts.

AI and Efficiency  by OpenAI

We’re releasing an analysis showing that since 2012 the amount of compute needed to train a neural net to the same performance on ImageNet1 classification has been decreasing by a factor of 2 every 16 months. Compared to 2012, it now takes 44 times less compute to train a neural network to the level of AlexNet2 (by contrast, Moore’s Law3 would yield an 11x cost improvement over this period). Our results suggest that for AI tasks with high levels of recent investment, algorithmic progress has yielded more gains than classical hardware efficiency.

READ PAPER:  https://arxiv.org/abs/2005.04305  (technical)

Algorithmic improvement is a key factor driving the advance of AI. It’s important to search for measures that shed light on overall algorithmic progress, even though it’s harder than measuring such trends in compute ... '

Google's Tsunami for Network Security Scanning

Here is a new and somewhat unexpected thing I was just introduced to by TWIT.    See their generalized writeup below.    Note still in pre-alpha form, but would seem to be a very useful thing to be aware of and if appropriate, use.   Plan is to extend this via the community.   Technical.   Public release announcement. 

Tsunami
https://github.com/google/tsunami-security-scanner

About
Tsunami is a general purpose network security scanner with an extensible plugin system for detecting high severity vulnerabilities with high confidence.

Tsunami
To learn more about Tsunami, visit our documentation.   https://github.com/google/tsunami-security-scanner/blob/master/docs/index.md  
Tsunami relies heavily on its plugin system to provide basic scanning capabilities. All publicly available Tsunami plugins are hosted in a separate google/tsunami-security-scanner-plugins repository.

Current Status
Currently Tsunami is in 'pre-alpha' release for developer preview.
Tsunami project is currently under active development. Do expect major API changes in the future.

For TWIT's very useful not too technical write up on this see:
https://www.grc.com/sn/SN-775-Notes.pdf      Page 12 ... 

 ... Supports small manually curated set of vulnerabilities
● Detects high severity, RCE-like vulnerabilities, which often actively exploited in the wild
● Generates scan results with high confidence and minimal false-positive rate.
● Detectors are easy to implement.

● Is easy to scale, executes fast and scans non-intrusively. .... '



Wednesday, July 15, 2020

Avatarin for Healthcare Robotics

We visited Japan in the 90s to talk robotics in retail.  Ended up mainly seeing eldercare and related home solutions.   Impressed, but did not see that many really implemented.   Is this changing now?  Will Pandemic add to new applications?

No Masks, No Coughs: Robots Can Be Just What the Doctor Ordered in Time of Social Distancing  By The Washington Post

Avatarin chair Akira Fukabori chats with a colleague via his companys newme robot.
Robots in Japan have taken on the roles of bartenders, security guards, deliverymen, and more, since the onset of the coronavirus pandemic.

In Japan, a country with a long fascination with robots, automated assistants have offered their services as bartenders, security guards, deliverymen, and more, since the onset of the coronavirus pandemic.

Japan's Avatarin developed the "newme" robot to allow people to be present while maintaining social distancing during the pandemic.

The telepresence robot is essentially a tablet on a wheeled stand with the user's face on the screen, whose location and direction can be controlled via laptop or tablet.

Doctors have used the newme robot to communicate with patients in a coronavirus ward, while university students in Tokyo used it to remotely attend a graduation ceremony.

The company is working on prototypes that will allow users to control the robot through virtual reality headsets, and gloves that would permit users to lift, touch, and feel objects through a remote robotic hand.   ...  

See also:     https://globetrender.com/2019/11/01/ana-newme-avatar-robots/  Which shows a numerb of uses of the Newme robotics.

Home Robotics with a Reaching Tele-Assistant

What looks like a minimalist robot, but with a great reach.  Evidently primarily for tele-operation. And thinking about it, reach is a big component of convenience in a home, say for assisting a disabled person.  Unclear if it is enough to be a universal assistant.   Also potential for other functionality.

Ex-Google robotics head unveils automated home assistant    by Peter Grad , in Tech Xplore

The former head of Google's robotics division has unveiled a new robot named Stretch that he hopes will prove to be an economical and handy assistant around the home.

And it's no stretch to say that it could provide a blueprint for future efforts in practical, low-cost automated devices to assist with household chores.

Aaron Edsinger, along with partner Charlie Kemp of the Georgia Tech Institute for Robotics and Intelligent Machines, say simplicity was key. They devised a lightweight machine consisting of a wheel base, a three-foot-high center pole and a telescoping touch-sensitive arm that, as its name suggests, stretches, and intelligently grabs and handles objects. It can carry items weighing up to three pounds.

"What sets this robot apart is its extraordinary reach, which is why we named it Stretch," said Edsinger. The arm, which moves easily along the center pole from top to bottom, can stretch out about 20 inches from its base.  ...  "

See also:  Home Robot Control for People With Disabilities
Georgia Tech's augmented-reality interface gives control over complex robots to the people who need them   .... By Evan Ackerman

See also Tele-operation video examples at:   https://hello-robot.com/