AI Curbing Traffic Accidents in Cities

 Analysis of patterns from city images recommends changes to decrease accidents.  

How AI Can Help Curb Traffic Accidents in Cities

By Universitat Oberta de Catalunya (Spain), March 19, 2021

A research project at the Universitat Oberta de Catalunya in Spain is harnessing the capabilities of artificial intelligence to make decisions that will make urban traffic safer.

The researchers trained neural networks to detect probable hazards in an area and found that certain patterns in urban scene layout—such as the arrangement of street furniture and the location of parked cars, advertisements, and facades—may impact accident rates.

The researchers developed a heuristic method to improve urban scenes, but it requires urban planners, architects, or engineers to implement changes based on the algorithm-driven data.

UOC’s Javier Borge said, “Our biggest hurdle is data availability: the analysis requires a rich collection of street view images and open, geolocalized data on accident rates with details of those involved, which are not currently easy to obtain."

From Universitat Oberta de Catalunya (Spain)

Detecting Likelihood of Suicide

Have consulted and spoke at the Veterans Administration.  Here is another pattern recognition type problem of interest.  

Can an Algorithm Prevent Suicide?

The New York Times, Benedict Carey

The U.S. Department of Veterans Affairs (VA) is using machine learning algorithms to help identify veterans at risk of suicide. The VA's Reach Vet algorithm is the first such program used in daily clinical practice, and is designed to generate a new list of high-risk veterans every month. When someone is flagged at risk, their name appears on the computer dashboard of the local clinic's Reach Vet coordinator, who contacts them to set up a meeting. The algorithm is based on analysis of thousands of veteran suicides; it considers numerous factors in veteran medical records to focus on those with the strongest cumulative association with suicide risk. Initial results indicate that over six months, high-risk veterans more than doubled their use of VA services—and had a lower mortality rate—with Reach Net installed, compared to a control group. ... 

Coffee Analysis Using Pattern Recognition

Used to be in the coffee blend analysis and optimization space, so this was interesting.  Nicely done.  We did some similar things with classical optimization and statistical systems.  Below an intro, link takes you to more detail exploration and to the data and technical details.  Also a good example of this kind of pattern recognition problem.   See my 'Coffee' tag below for many mentions of our work and related research.

Comparing Coffee using Pattern Recognition

Exploring coffee similarities using cupping grades and flavors

By Robert McKeon Aloe   in Towards Data Science

After trying a variety of coffees from around the world, I have often wondered if coffee grades or flavors could tell me which coffees were similar to each other and which were different. I am particularly interested if such comparisons could be used to help determine the best blends by increasing my understanding of how coffees relate to each other. Luckily, I found some data to investigate just that question.

I have looked at two databases with coffee grades, and previously, I’ve used the CQI database to investigate similarities between coffees. I’m applying the same techniques here.

I amended one of the databases from Sweet Maria’s. I previously pulled their Q-scores, but they also had flavor ratings for each coffee. So I went back, and I pulled flavor ratings for all the beans. I ended up with a slightly larger database than previously at 407 coffees. ... "

Capturing Words from Silent Speech

An Indication of how AI can become very powerful, even replacing the need for actual speech with mouthed muscle activity.  A kind of pattern recognition.

UC Berkeley researchers detect ‘silent speech’ with electrodes and AI

Khari Johnson  @kharijohnson in VentureBeat

UC Berkeley researchers say they are the first to train AI using using silently mouthed words and sensors that collect muscle activity. Silent speech is detected using electromyography (EMG), with electrodes placed on the face and throat. The model focuses on what researchers call digital voicing to predict words and generate synthetic speech.

Researchers believe their method can enable a number of applications for people who are unable to produce audible speech and could support speech detection for AI assistants or other devices that respond to voice commands.  ... " 

Finding the New from the Old

Makes the case that you can use older data to make useful conclusions.  Though in our experience the required meta data was often not available.    Physics/astronomy is at least often based on some strong science foundations.

AI Identifies 50 New Planets From Old NASA Data
Jessie Yeung
August 26, 2020

Machine learning artificial intelligence (AI) developed by astronomers and computer scientists at the U.K.'s University of Warwick found 50 new planets by mining old data from the U.S. National Aeronautics and Space Administration (NASA). The researchers educated the algorithm on data collected by the Kepler Space Telescope, teaching it to differentiate real planets from false positives. The AI was then tasked to analyze old datasets of planetary candidates, in which it discovered the 50 previously unknown exoplanets. Warwick's David Armstrong said this is the first time machine learning has been used to rank planetary candidates in a probabilistic framework, and the research suggests the AI could "validate thousands of unseen candidates in seconds."... '

Algorithm Predicts Corn Yields

What seems to be some direct uses of CNN for yield prediction.

AI Algorithm Better Predicts Corn Yield
By Illinois ACES
February 24, 2020
  
An interdisciplinary research team at the University of Illinois at Urbana-Champaign has developed a convolutional neural network that generates crop yield predictions.

An interdisciplinary research team at the University of Illinois at Urbana-Champaign has developed a convolutional neural network (CNN) that generates crop yield predictions, incorporating information from topographic variables such as soil electroconductivity, nitrogen levels, and seed rate treatments.

The team worked with data captured in 2017 and 2018 from the Data Intensive Farm Management project, in which seeds and nitrogen fertilizer were applied at varying rates across 226 fields in the Midwest U.S., Brazil, Argentina, and South Africa.

In addition, on-ground measurements were combined with high-resolution satellite images from PlanetLab to predict crop yields.

Said Illinois's Nicolas Martin, while "we don’t really know what is causing differences in yield responses to inputs across a field … the CNN can pick up on hidden patterns that may be causing a response.”    ... '

Patterns in Recycling Waste Streams

Consider the complexity of recycling waste, a major reason it is not re-used efficiently enough is the cost of sorting it.  Training data would be interesting to see.

Recycling Centers Roll Out AI-Trained Robots
The Wall Street Journal
By John Murawski

Recycling facilities across the U.S. are starting to use robots guided by artificial intelligence (AI) to sort waste streams more quickly and accurately than human workers. For example, Sarasota, FL-based Single Stream Recyclers is planning to add eight robots from AMP Robotics to the six it already operates to help it process up to 350 tons of recycling waste a day. The robots can sort 80 items per minute, about twice as much as the average human worker. AMP uses robots from third-party vendors, to which it adds machine learning software trained on databases of images of water bottles, beer cans, milk jugs, food cartons, and other items. Meanwhile, a Berkeley County, SC, recycling facility is using nine robots from AI developer Bulk Handling Systems, which also uses off-the-shelf robots to execute its proprietary AI algorithms. ... "

Glass Recognizes Numbers with AI

We continue to discover new ways we can link sensors and pattern recognition to create knowledge.   Note the glass does not need to be plugged into power.

AI made from a sheet of glass can recognise numbers just by looking
A sheet of glass could serve as an AI      By Donna Lu in New Scientist

It’s the smartest piece of glass in the world. Zongfu Yu at the University of Wisconsin–Madison and his colleagues have created a glass artificial intelligence that uses light to recognise and distinguish between images. What’s more, the glass AI doesn’t need to be powered to operate.

The proof-of-principle glass AI that Yu’s team created can distinguish between handwritten single digit numbers. It can also recognise, in real time, when a number it is presented with changes – for instance, when a 3 is altered to an 8.

The AI works by taking advantage of small bubbles deliberately incorporated into the glass, as well as embedded impurities such as graphene. As the light waves reflected off the handwritten numbers pass through the glass AI, they are bent in specific ways by the bubbles and impurities. This bending process focuses the light onto one of ten specific spots on the other side of the glass, depending on the number presented. The ten spots then correspond to the digits 0 to 9. “It is like a key and a lock,” says Yu.  .... " 

Pattern Recognition in Road Tracing

We had a similar challenge when we managed roads and their quality in Alberta.   We were mostly interested in if the roads had degraded, and needed to manage resources to improve them to a level where they would satisfy predicted transport needs.   We had Aerial and Satellite images.   Now you can selectively deploy drones.   And we needed to include predicted changes like tree growth, storms  and forest fires.  This does some of the things we could have used then.  Perhaps also some applications to supply chain needs?

A new way to automatically build road maps from aerial images
“RoadTracer” system from the Computer Science and Artificial Intelligence Laboratory could reduce workload for developers of apps like Google Maps. ....  By Adam Conner-Simons | MIT, CSAIL 

A New Way to Automatically Build Road Maps From Aerial Images 

" ... Massachusetts Institute of Technology (MIT) Computer Science and Artificial Intelligence Laboratory (CSAIL) researchers have developed a system for automatically constructing road maps from aerial images with 45-percent more accuracy than existing methods. The RoadTracer system “is well-suited to map areas of the world where maps are frequently out of date, which includes both places with lower population and areas where there's frequent construction," says MIT professor Mohammad Alizadeh. RoadTracer begins with a known location on a road network, using a neural network to examine the surroundings and determine which point is most likely to be the next part of the road. It adds that point and repeats the process to plot out the network of the road step by step..... " 
(Updated) See also 'Automated site Mapping'  Used in Archaeology,  related method beyond examining just roads.

Spotting Viruses with Machine Learning

A further indication of the value of pattern recognition ...

Machine Learning Spots Treasure Trove of Elusive Viruses 
In Nature  by Amy Maxmen

Researchers have applied artificial intelligence to the discovery of nearly 6,000 previously unknown species of virus. Simon Roux at the U.S. Department of Energy's Joint Genome Institute trained computers to identify viral genetic sequences from the Inoviridae family. He gave a machine-learning algorithm two sets of data, including 805 genomic sequences from known Inoviridae viruses, and about 2,000 sequences from bacteria and other types of virus, so the program could learn to differentiate between them. The model was then fed metagenomic datasets, and the computer retrieved more than 10,000 Inoviridae genomes, clustering them into groups indicative of different species. A separate study conducted at the University of Sao Paulo in Brazil used machine learning to identify viruses in compost piles at the city's zoo, by programming an algorithm to look for a few distinguishing characteristics of virus genomes. Following training, the computer recovered several genomes that appeared to be new. .... " 

Humans as Neurons Strategy

Brought to my attention.  Takes me back to our early days of neural nets, and thinking about how error minimizing reinforcement worked in alliance with an architecture of neurons.  We thought of them as very restricted societies, or companies aiming at some goal.  Like making money or recognizing patterns. Some objective function linked to an outcome.  - FAD

The Human Strategy
A Conversation With Alex "Sandy" Pentland   In The Edge,   Podcast

The idea of a credit assignment function, reinforcing “neurons” that work, is the core of current AI. And if you make those little neurons that get reinforced smarter, the AI gets smarter. So, what would happen if the neurons were people? People have lots of capabilities; they know lots of things about the world; they can perceive things in a human way. What would happen if you had a network of people where you could reinforce the ones that were helping and maybe discourage the ones that weren't?

That begins to sound like a society or a company. We all live in a human social network. We're reinforced for things that seem to help everybody and discouraged from things that are not appreciated. Culture is something that comes from a sort of human AI, the function of reinforcing the good and penalizing the bad, but applied to humans and human problems. Once you realize that you can take this general framework of AI and create a human AI, the question becomes, what's the right way to do that? Is it a safe idea? Is it completely crazy?  .... "

ALEX "SANDY" PENTLAND is a professor at MIT, and director of the MIT Connection Science and Human Dynamics labs. He is a founding member of advisory boards for Google, AT&T, Nissan, and the UN Secretary General. He is the author of Social Physics, and Honest Signal. Sandy Pentland's Edge Bio page  ...... " 

Grocery Startup Uses AI

Artificial Intelligence is drawing attention for its potential and its pitfalls. Tesla founder Elon Musk recently called for greater regulation of AI robotics due to its “public risk.” ...

CGTN’s Mark Niu reports from San Francisco about a startup using AI to better serve their business.
Going head to head with giants like AmazonFresh Grocery and big name supermarkets, Farmstead is a new Silicon Valley startup that deliveries groceries. From what it calls micro hubs, Farmstead tracks locally sourced products every step of the way as they’re delivered to homes in San Francisco within one hour of ordering. Farmstead’s secret recipe is AI.

“Artificial intelligence plays a massive role. We take into account all of our customers buying patterns to date. To help us predict exactly how much product to order for every particular item that we carry. We waste a lot less, and as a result we can pass those savings along to our customers,” Farmstead Co-founder & CEO Pradeep Elankumaran said.  .... "

Technologies for Good and Bad

Just as new machine learning techniques can be used for the good, they can also be used for bad purposes.  The below research is been being reported as an example, ways of cracking encryption is another. ...

AI Model Fundamentally Cracks CAPTCHAs
National Public Radio   By Merrit Kennedy

Researchers at the artificial intelligence company Vicarious say they have developed a computer model that can solve Completely Automated Public Turing Tests to Tell Computers and Humans Apart (CAPTCHAs). Vicarious co-founder Dileep George says the model cracks a CAPTCHA's defenses by parsing the text the test presents more effectively than earlier deep-learning models and with less training. George cites the model's use of a Recursive Cortical Network, which is better able to reason about what it is seeing with less training than previous models. He says the network's training entails constructing internal models of the letters it is exposed to, and when a new image is presented it attempts to explain it, trying to define all of the pixels of that new image in terms of the characters it has previously seen. George notes the long-term objective of such experiments "is to build intelligence that works like the human brain. CAPTCHAs were just a natural test for us."

Identifiying Plant Disease

Out of the fall garden and back to the possibilities of pattern recognition.  Back to my alternate botanic tech universe.  Penn State work using deep learning to identifying plant disease.  Great example:

 Phone Powered AI Spots Sock Plants with Remarkable Accuracy,   by Matt Simon, Science

Researchers at Pennsylvania State University (PSU) say they have designed a smartphone-based neural-network program that can automatically identify diseases in the cassava plant with near-flawless accuracy. The network is based on Google's open source TensorFlow machine-learning library, and Google's Pete Warden notes the TensorFlow Mobile app requires only about 25 million parameters, versus the hundreds of millions some networks need. "It only requires about 11 billion floating point operations to actually calculate its result, and some other networks require hundreds of billions of operations to do a similar job," Warden notes. He also says thanks to transfer learning, the network was trained to recognize cassava leaves on much less data. "It really comes down to the data, because garbage in, garbage out," says PSU's Amanda Ramcharan. She believes the growing affordability of smartphones and the continuing simplification of algorithms will combine to make such tools more widely available to farmers. ... " 

Fingerprint Analysis Scoring

Recall seeing this problem as a classic pattern recognition issue,  closer to being solved.  Note this addresses knowledge content in a pattern, where else might this be used? Anywhere you might want to prioritize an action.

Scientists Automate Key Step in Forensic Fingerprint Analysis
By U.S. National Institute of Standards and Technology 

Researchers at Michigan State University and the U.S. National Institute of Standards and Technology have developed an algorithm that automates the fingerprint-analysis process.

The researchers trained the machine-learning algorithm on data from 31 fingerprint experts who had analyzed 100 latent prints each, scoring the quality of each on a 1-to-5 scale. The prints and their scores were used to train the algorithm to determine how much information a latent print contains.

The researchers tested the algorithm by having it score a group of new latent prints. The team submitted those scored prints to Automated Fingerprint Identification System software connected to a database of more than 250,000 rolled prints. The researchers found the scoring algorithm performed slightly better than the average of human examiners involved in the study.

They next want to test the system on a larger dataset, which will enable them to improve the algorithm's performance and more accurately measure its error rate.

From U.S. National Institute of Standards and Technology

(More details at the link)

Detecting and Identifying Images

Home/News/Air Force Tests IBM's Brain-Inspired Chip as an Aerial Tank Spotter

ACM Tech News
Air Force Tests IBM's Brain-Inspired Chip as an Aerial Tank Spotter.   Neuromorphic chips.
In Technology Review

Detecting Review and Follower Fraud

From CMU.  Tested against some very large set of Twitter data.   Uses a form of Graph Analytics.  Python code is available as Open Source.  Uses a form of  Faloutsos' NetProbe  (technical paper) for the analysis.

New algorithm detects online fraudsters: Method sees through camouflage to reveal fake followers, reviewers.

The method, called FRAUDAR, marks the latest escalation in the cat-and-mouse game played by online fraudsters and the social media platforms that try to out them. In particular, the new algorithm makes it possible to see through camouflage that fraudsters use to make themselves look legitimate, said Christos Faloutsos, professor of machine learning and computer science.

In real-world experiments using Twitter data for 41.7 million users and 1.47 billion followers, FRAUDAR fingered more than 4,000 accounts not previously identified as fraudulent, including many that used known follower-buying services such as TweepMe and TweeterGetter.

"We're not identifying anything criminal here, but these sorts of frauds can undermine people's faith in online reviews and behaviors," Faloutsos said. He noted most social media platforms try to flush out such fakery, and FRAUDAR's approach could be useful in keeping up with the latest practices of fraudsters.

The CMU algorithm is available as open-source (Python) code at http://www.andrew.cmu.edu/user/bhooi/camo.zip. A research paper describing the algorithm won the Best Paper Award last month at the Association for Computing Machinery's Conference on Knowledge Discovery and Data Mining (KDD2016) in San Francisco. ..... " 

Plotting Analytics Terminology

An interesting article in the SAS blog on the evolution of 'data mining', actually a term less used these days.   They quote the image below about how a number of capabilities overlap, which was published by SAS in 1998.  

The terms all still exist today,  are understood.  But how they overlap and refer to specific available methods is still very debatable.    Some are narrow, some very broad. I am involved in designing a course that addresses just this confusion today.     Worth thinking about each of the terms.

Driving Light Show with Pattern Recognition AI

An example of using sensors to drive performance light position.  In the BBC.  Ultimately a good example.  The sensors gather data.  Patterns can be matched to scripts (goals).   Adaptations to performance are made to achieve results.  Potential to learn by feeding back results to performers ( or script writers) . Whole thing recorded for later strategy adaptation.  So what is AI?  The advanced machine learning / pattern recognition. Not dissimilar from more complex versions of adaptive process control.

Microsoft to buy Swiftkey

In Fortune:  When I first read this, I remembered seeing Swiftkey demonstrated.  Though I never had it on one of my own devices, the aha came quickly.  Quickly matching key patterns to retrieve used words.  Now where else could that be used?  One of those embedded AI era developed things we forget about.   Instructive to see where this will go,