Facebook Blender Chatbot Does Open Source

Anything that can make a chatbot more conversational, and have the ability to use and update context
continually is good. 

Facebook releases its 'Blender' chatbot as an open-source project
It could help tomorrow's AI converse more naturally with people

Andrew Tarantola, @terrortola in Engadget

The virtual assistants that inhabit our smartphones are helpful, sure, but they’re not going to pass the Turing test any time soon. They’re designed for understanding specific commands and actions like checking on restaurant reservations or getting updates on the weather, rather than, say, carrying on an in-depth conversation with a human. But chatbots could soon become far more loquacious thanks to Facebook, which this morning released a startlingly lifelike chatbot that it’s been developing, dubbed Blender, as an open-source resource for AI research.

Facebook has been pouring money and resources into its Natural Language Processing technologies for a few years now and those efforts appear to have paid off. The company claims that Blender is the single largest open-source chatbot created to date. It’s been trained on a whopping 9.4 billion parameters -- nearly 4x as many as Google’s Meena and more than 10x as many as the previous largest OS chatbot available on the internet.

“One of the recent findings in the area of NLP, and AI in general, has been that as you scale, as these neural network models larger and larger, they tend to perform better,” Stephen Roller, a research engineer at Facebook’s AI lab (FAIR), told Engadget. “We had a number of issues when we were trying to train this thing. When you start to get that large, these things no longer are able to fit on a single GPU anymore.”  .... "

See more on their Blended Talk Ability.

Examining Seasoning Research

Was for some time involved deeply in the manufacturing of CPG food.   Blending commonly used foods from agricultural commodities.   Also have been a cook for a long time.   So attuned to the use and usage trends in ingredients and spices.  Just recently noted the existence of a  Chinese 'Tainjin Seasoning Research Institute',  which makes a tomato paste sold at the local ethnic market.

China probes 'fake seasoning producing hub' near Tianjin
and more in the China seasoning realm. ... "

Algorithmic Gastronomy and Beyond

Fascinating look about how to morph recipes.  Might this be used to rewrite other process + ingredient kinds of decisions?  We were influenced in coffee blending and roasting decisions by recipe style approaches.  There is something here.

Deep Learning Algorithm Rewrites Traditional Recipes for New Regions, Ingredients
by Michael Byrne

 Algorithmic gastronomy is here.

Imagine your favorite go-to recipe mutated to conform to the traditional methods and ingredients of any number of diverse regional food cultures. Consider, say, lasagne, but a sort of lasagne that's instead a naturally occurring part of Japanese or Ethiopian cuisine. Not "fusion," but something deeper—a whole rewriting of what a lasagne even is according to the culinary traditions of some other place.

It's not necessarily an easy or natural thing to do, but a new machine learning algorithm developed by a team of French, American, and Japanese researchers offers an automated solution based on neural networks and large amounts of food data. The result, which is described in a paper published this month to the arXiv preprint server (via I Programmer), is a system that can take a given recipe and shift it into an alternative dietary style—sushi lasagne, say—as well as parse a recipe for its underlying style components.  .... " 

Electronic Tongue Tastes Wine

An area we explored on the coffee blending business.  Was still very primitive in capabilities at the time.  Some advances are seen here.  How can resulting data be connected to blending or manufacturing data?   In New Scientist. 

Scanning the Brains of Dogs

In Wired:   An article about how DARPA is planning to scan the brains of dogs to make better choices of military pairings.  An interesting idea that could look at things like task optimization for applications.  The research might also ultimately also lead to new learning for human neuroscience as it can be practically applied.

Deep Learning and Artificial Intelligence

Mindhacks mentions an article in the NYTimes that I have not yet read, about advances in artificial intelligence.   Recall we did early work in using AI applied to research and manufacturing decision making in areas like blending, maintenance, diagnostics and other related human intelligence areas.   Mentioned is Geoffrey Hinton, a long time icon in the interaction of data and intelligence.   Hinton was particularly known for the linking of neural nets and learning systems.  We used some of his concepts and directions for application to learning systems that could be taught to do what humans achieve in the enterprise.   The brief outline of the work that he did was for integrating pharma research.  We did similar work with R&D concepts to develop new products for the enterprise.

The NYT article about deep learning is here.  Reading.  See Hinton's background at the U of Toronto with some descriptive information about his Pharma work.

Pack Mule

DARPA's pack mule for the soldier is getting closer to reality.  Saw some early tests in robotics exhibitions but wondered if it would ever be used.  Expect the appearance of other forms of walking drones as well.

Cybernetic Binoculars

I have been following a number of DARPA developments.  Here a project which seeks to link brain waves and sensor interactions.  A kind of brain augmentation that we will see more of.    " ... DARPA's Cybernetic Binoculars Tap Soldiers' Brains To Spot Threats  ... The U.S. Army and DARPA have concluded field tests on next-generation binocular replacements that read human brain signals and have a 91% threat detection success rate. They might just help you control your car with your thoughts too ... " 

DARPA sees Analog

My introduction to digital computing came just as the last courses were taught using analog methods.   The convenience of digital devices was so overwhelming that you could easily see analog methods disappearing completely.  Yet there were then and are now examples where analog works better.   DARPA looks at this future.  " ... the DoD is funding a new program called UPSIDE, short for Unconventional Processing of Signals for Intelligent Data Exploitation. Basically, the program will investigate a brand-new way of doing computing without the digital processors that have come to define computing as we know it ... " 

Deep Natural Language Analysis

Of Interest:  Another DARPA example I have just examined.  Automated, deep natural-language understanding technology may hold a solution for more efficiently processing text information. When processed at its most basic level without ingrained cultural filters, language offers the key to understanding connections in text that might not be readily apparent to humans. DARPA created the Deep Exploration and Filtering of Text (DEFT) program to harness the power of language. Sophisticated artificial intelligence of this nature has the potential to enable defense analysts to efficiently investigate orders of magnitude more documents so they can discover implicitly expressed, actionable information contained within them.

DARPA Unshredding Challenge

In CACM:  Solved it says.   I encountered a related problem very early in my enterprise career.  Not involved with anything secret.   We partially solved the problem and got real value from it.  These additional learnings would have helped considerably.  It has been known for a long time that paper shredding is insecure.   We used a 'pulping' method that mixed paper with water and created a sludge.   The background of the challenge

" ... To seek "unexpected advances" that could be applied to such situations, the U.S. Defense Advanced Research Projects Agency (DARPA) issued its Shredder Challenge in October 2011, daring the public to unshred five documents and extract their hidden messages. The puzzles were designed to be progressively difficult on multiple axes, ranging from about 200 chads to more than 6,000 each. Text files accompanying the scanned chads provided questions, with points reflecting each question's difficulty. For example, puzzle #3 asked "What is the indicated location?" while the reassembled document showed a set of geographic coordinates and a drawing of Cuba. (Naming the country was worth two points; the city of Cienfuegos was worth an additional six.) Solvers needed to both answer the questions and show how their reconstruction of the document led to that answer. More than 9,000 teams applied for the challenge, but only 69 of them answered one or more questions correctly.... "

DARPA Seeks Radical Data Analysis

I had long ago involvement with DARPA, they are building up their large scale data involvement. An example:  Military Transformers: 20 Innovative Defense Technologies (click image for larger view and for slideshow) The Defense Advanced Research Projects Agency is planning an unusual eight-week residency program in which teams of researchers brainstorm over ways to use visual and geospatial images to identify people, places, objects, and activities .... " 

Robotics Challenge for Disaster Response

A very detailed  DARPA challenge for robotics in the area of disaster response.  In IEEE spectrum.   I have been surprised that more has not been done in this context.  It is an obvious application.   At the same time, disasters are by their nature chaotic, and the advanced pattern recognition needed to address chaotic environments is more difficult.

The Pentagon and Augmented Reality

And as you might expect, the US DOD is very interested in the idea of hands-free augmented reality displays.  They have played in this space in a number of ways in the past, including early heads-up displays in aircraft.   In the enterprise we tested the idea for maintenance applications, but at the time they were not appropriate.  Now their time may have come.  Integrating them in contact lenses could remove some of associated issues we encountered:

"  ... DARPA orders prototypes of Innovega's iOptik displays that use special contact lenses so a person can focus both on images shown on the lenses and far-away objects for augmented reality.... " 

Data Blending in Tableau

An academic perspective about data blending.  Aimed at the non-expert in the use of data and modeling.

Cognitive Computing Page

A  reconnect with some IBM researchers led me to revisit their cognitive computing presence:

" ... Researchers at IBM have been working on a cognitive computing project called Systems of Neuromorphic Adaptive Plastic Scalable Electronics (SyNAPSE). By reproducing the structure and architecture of the brain—the way its elements receive sensory input, connect to each other, adapt these connections, and transmit motor output—the SyNAPSE project models computing systems that emulate the brain's computing efficiency, size and power usage without being programmed.

IBM is combining principles from nanoscience, neuroscience and supercomputing as part of a multi-year cognitive computing initiative. The Defense Advanced Research Projects Agency (DARPA) has awarded approximately US$21 million in new funding for phase 2 of the SyNAPSE project. For this project, a world-class, multi-dimensional team has been assembled, consisting of IBM researchers and collaborators from Columbia University; Cornell University; University of California, Merced; and University of Wisconsin-Madison.... "

Air Force Social Radar

The idea of air force social radar sounds much like neuromarketing efforts underway today, but some more seriousness involved.   Reading minds all the way down to the intent level is still some time away.  Thoughts like those expressed could result in DARPA like developments that could be spread back into the marketing context.

Future Cola

We took at look at Coca-Cola's futuristic drink blending and delivery dispenser for some time.  It is a classic case of late differentiation and tailoring a product to the local consumer.  And the dispenser can gather valuable data about what consumers want.  For a long time it seemed like this would be one of those concept ideas that would never see general installation.  A new report indicates that the idea is moving forward.

 " ... The fountain serves up 125 different flavors of soft drinks, flavored waters, sports drinks and lemonades and sends usage data, such as what flavors are most popular at what times of the day, to Coca-Cola HQ. Already the beverage giant is analyzing data pouring in from more than 1,500 machines in restaurants including Wendy's, Burger King, Taco Mac and Five Guys.

While individual restaurant customers have done marketing campaigns using Coca-Cola assets, the beverage giant is now working with Ogilvy & Mather to develop creative concepts for a 2012 marketing campaign. Sydney Taylor, group director of Coca-Cola Freestyle consumer marketing, said that radio, outdoor and digital would likely be the initial focus of any measured media. "We needed to be at a certain threshold to make [a marketing campaign] pay off for us, so that once we advertise Freestyle, consumers can find it at most places," she said ... "

Benddesk

Reported in Engadget: the Benddesk, a workspace of the future idea that includes a flat table like structure and an integrated background area.  All accessible with  multi touch interaction.  I have seen a number of demonstrations of multitouch interaction, but  they are rarely broadly useful.  Working with a business furniture manufacturer years ago there was talk of blending multiple spaces together to make them as interoperable and flexible as possible.   For specialty uses perhaps, but better mobility would be useful.  Includes a video.

Nanotech Electronic Noses

New work in building electronic noses. We experimented with the idea for blending beverages. In general the electronic sensors were not accurate enough.