AI Reduces 100,000-Equation Quantum Physics Problem to Four Equations

The power of leveraging patterns in the right context

ACM NEWS

Uncovering Hidden Patterns: AI Reduces 100,000-Equation Quantum Physics Problem to Four Equations

By SciTechDaily, September 29, 2022

A daunting quantum problem that until now required 100,000 equations has been compressed into a bite-size task of as few as four equations by physicists using artificial intelligence. All of this was accomplished without sacrificing accuracy. The work could revolutionize how scientists investigate systems containing many interacting electrons. Furthermore, if scalable to other problems, the approach could potentially aid in the design of materials with extremely valuable properties such as superconductivity or utility for clean energy generation. .... 

From SciTechDaily    

Programmable Matter for Product Design

With a zap of light, system switches objects’ colors and patterns

“Programmable matter” technique could enable product designers to churn out prototypes with ease.

Watch Video  https://news.mit.edu/2021/light-colors-patterns-surface-0504#article-video-inline

Daniel Ackerman | MIT News Office

With Zap of Light, System Switches Objects' Colors, Patterns

MIT News, May 4, 2021

A programmable matter system developed by researchers at the Massachusetts Institute of Technology (MIT) and Russia's Skolkovo Institute of Science and Technology can update imagery on object surfaces rapidly by projecting ultraviolet (UV) light onto items coated with light-activated dye. The ChromoUpdate system's UV light pulse changes the dye's reflective properties, creating colorful new images in minutes. The system’s UV projector can vary light levels across the surface, granting the operator pixel-level control over saturation levels. MIT's Michael Wessley said the researchers are investigating the technology's application to flexible, programmable textiles, "So we could have clothing—t-shirts and shoes and all that stuff—that can reprogram itself."  ..' 

Early Warning Systems for Self Driving

Recognizing situations for warnings. Classic AI application.

Early Warning System for Self-Driving Cars

Technical University of Munich (Germany),  March 30, 2021

An early warning system for self-driving vehicles developed by researchers at Germany's Technical University of Munich (TUM) leverages artificial intelligence (AI) to learn from real traffic situations. During tests on public roads, the researchers identified about 2,500 situations that required driver intervention; they found the system issued warnings about potentially critical situations seven seconds in advance with more than 85% accuracy. Using a recurrent neural network, the system recognizes patterns in the data collected by sensors and cameras and will warn drivers if it identifies a situation the control system has had difficulty handling in the past. TUM's Eckehard Steinbach said, "We limit ourselves to the data based on what actually happens and look for patterns. In this way, the AI discovers potentially critical situations that models may not be capable of recognizing, or have yet to discover."  .. '

Detecting Earthquakes and Waves

A favorite topic, detecting pattern.  Its a big part of our intelligence.   Here using new kinds of sensor data. 

Google Uses Underwater Fiber-Optic Cable to Detect Earthquakes   By New Scientist

A submarine fiber-optic cable owned by Google was used by researchers at the search engine giant and California Institute of Technology (Caltech) to detect earthquakes and ocean waves generated by storms.

The investigators measured changes in pressure and strain using traffic data from the 10,000-kilometer (6,213-mile)-long cable on the floor of the Pacific Ocean, recording about 30 ocean storm swell events and roughly 20 quakes exceeding magnitude 5 over nine months. Caltech's Zhongwen Zhan described this approach as more flexible and scalable than other attempts to deploy fiber-optic sensors, as new infrastructure is unnecessary.   Anthony Sladen at the University Côte d’Azur in France says the study constitutes “a major step in exploiting the benefits of existing cables.”

From New Scientist

Spotting Risky Behavior

 I was struck by this.   What is high risk behavior?  How is it defined, tested, updated?   OK, depression, anxiety or suicide. But at any one time it might be one or another set of beliefs.  Where (obviously) the other is wrong?  Novel, creative, unique can be risky.  Will it get me cancelled?

Spotting High-Risk Behavior Online, By Sandrine Ceurstemont,  Commissioned by CACM Staff, January 21, 2021

Posts on social media could conceal clues about mental health problems or high-risk behaviors.

Social media is used widely to share experiences with friends, or to join like-minded communities to discuss common interests. Yet people's posts also could conceal clues about mental health problems or high-risk behaviors that, if recognized early enough, could help save lives.

"Depression, anxiety, suicidal behaviors and some disordered eating behaviors are difficult to detect in person and it's unlikely that people are going to go to a clinic because of how stigmatized these conditions are," says Stevie Chancellor, a researcher at Northwestern University in Evanston, IL. "If we could use social media data as a way to understand these behaviors, perhaps we could use that information to assist them."

Chancellor and other researchers are investigating how machine learning could be harnessed to identify signs of dangerous behavior on social media. Around half of the world's population, roughly three billion people, now use social media platforms including Facebook, Twitter, and Reddit, so there is lots of data available. "It allows us to target a lot more people at a greater level than we've ever been able to before to understand these populations," says Benjamin Ricard, a Ph.D. student at the Geisel School of Medicine at Dartmouth College in Hanover, NH.

Some research is using social media to examine what people write, attempting to predict risky behavior from language used. Another approach involves looking at information related to posts, such as how much information a person shares, at what time of day, and whether that individual's  posting habits have changed. "A common symptom of depression is insomnia, so if your posting history over time starts shifting later, that might indicate that you are struggling with insomnia, which could relate to depression," says Chancellor.  ... "

Benford's Law in the Universe

 We used Benford's law to detect potential fraud.  I have mentioned it here a number of times. Turns out its much more broadly observed.   Which is fairly obvious, math is also observed throughout the universe.   Fits my interests well.  Here is a good overview.  Worth understanding. 

Benford’s law and distances to stars  

Mysterious digit-law embedded in the universe?

By Jurjen de Jong in TowardsDataScience

One day Simon Newcomb (1835–1909) was looking at the pages of logarithmic tables when he saw that the first pages were more worn out than further pages. This simple observation and the way logarithmic tables are structured, meant that numbers starting with digit 1 were more common in nature than numbers starting with digit 2 and digit 2 more common than digit 3 and so on. This strange pattern in the first-digit frequency might sound counterintuitive at first, but it turned out to be true for many numerical datasets.

To make more clear what we mean with the first digit: this means that 1, 1213123, and 0.00153 all have digit 1 as the first digit, while 312, 0.3, and π share 3 as the first digit.

Newcomb published a paper about it in 1881, but his discovery wasn’t really picked up by the scientific community. So, when Frank Benford (1883–1948) rediscovered this phenomenon in 1938, he didn’t know about Newcomb’s discovery. Benford looked at the first digits of many different datasets, such as lengths of rivers, addresses, atomic weights, random newspaper numbers, and so on. Every time, he saw a similar pattern, which supported the idea that numbers could be sorted on their first digit. This digit-law is now called Benford’s law but could also have been called Newcomb’s law. .... '

Limitations of Deep Learning

About limitations of deep learning problems with higher dimensional data. Technical.

Geometric Deep Learning Advances Data Science   By Samuel Greengard

Communications of the ACM, January 2021, Vol. 64 No. 1, Pages 13-15

10.1145/3433951

Deep learning has transformed numerous fields. In tackling complex tasks such as speech recognition, computer vision, predictive analytics, and even medical diagnostics, these systems consistently achieve—and even exceed—human-level performance. Yet deep learning, an umbrella term for machine learning systems based primarily on artificial neural networks, is not without its limitations. As data becomes non-planar and more complex, the ability of the machine to identify patterns declines markedly.

At the heart of the issue are the basic mechanics of deep learning frameworks. "With just two layers, a simple perceptron-type network can approximate any smooth function to any desired accuracy, a property called 'universal approximation'," points out Michael Bronstein, a professor in the Department of Computing at Imperial College London in the U.K. "Yet, multilayer perceptrons display very weak inductive bias, in the sense that they assume very little about the structure of the problem at hand and fail miserably if applied to high-dimensional data."  .... ' 

Healthcare AI Looks for Patterns

The easiest thing to do with AI (or other kinds of analytics) is to pattern match.  Not the only thing,  but if you have enough data to train it, you can find a likelihood that you have detected the same thing.    You also have to consider the risk of false positives and negatives in context.  There is no downside in testing the idea with just the data you have.  I am then often asked then, how much data?

  The answer is really any amount of data,  say hundreds of examples, to get started to test some ideas, set up the methods.    But then to deliver real results your are talking thousands to many thousands of examples.    And then also consider re-testing of other contexts, and testing over time.   And reworking the risk analysis as well.   Below a real example, note the variants in setting up the idea, and the context of alerts.  Note that the data is plentiful, generated by reliable sensors, a good sign,

Hospital System Uses AI to Predict Deadly Condition
By The Wall Street Journal 

The app predicts the likelihood of acute respiratory failure.  ... 
The Montefiore Health System and its affiliated medical school in New York are employing artificial intelligence to predict the likelihood of a patient suffering from a common respiratory malady.  ... 

The Montefiore Health System and its affiliated medical school in New York are employing artificial intelligence (AI) to predict a common respiratory malady among patients.

Parsa Mirhaji at Montefiore's Center for Health Data Innovations said his team uses three live AI applications, with four more in development; three hospitals use the app to predict acute respiratory failure, which can require intubation.

The app analyzes more than 40 data points for each patient every several hours, including drug intake and blood pressure, with patients at risk of intubation or death assigned a risk score.

An intervention alert is activated, and sent to doctors, when patients' scores surpass a specific threshold.

Tests indicated the AI app yielded lower false-positive rates than traditional hospital alert systems.

From The Wall Street Journal
View Full Article - May Require Paid Subscription  ... "

Predicting Patterns in Large Data Streams

Its all about patterns in context, whether to gain insight, predict or prescribe.  So its analytics too.  But from my perspective,  AI occurs when we insert the analysis it into a useful process.  And 'intelligence' when it interacts to provide value in a selection of domains that add up to context that provides significant value to humans or business.

MIT CSAIL’s machine learning algorithm helps predict patterns in large data streams   By Kyle Wiggers

Ever heard of the “Britney Spears problem“? Contrary to what it sounds like, it’s got nothing to do with the dalliances of the rich and famous. Rather, it’s a computing puzzle related to data tracking: Precisely tailoring a data-rich service, like a search engine or fiber internet connection, to individual users hypothetically requires tracking every packet sent to and from the service provider, which needless to say isn’t practical. To get around this, most companies leverage algorithms that make guesses about the frequency of data exchanged by hashing it (i.e., divvying it up into pieces). But this necessarily sacrifices nuance — telling patterns that emerge naturally in large data volumes fly under the radar.

Luckily, researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) believe they’ve devised a viable alternative that relies on machine learning. In a newly published paper (“Learning-Based Frequency Estimation Algorithms“), they describe a system — dubbed LearnedSketch, because of the way it “sketches” data in a data stream — that predicts if specific data elements will appear more frequently than others and, if they in fact do, autonomously separates them from the rest of the hashed portions. ... "

Looking for Patterns in Neurons

Fascinating after-training analysis of how neurons capture specifics.   Deep research  and technical,  but perhaps a way to squeeze new kinds of information from Deep Learning.  Beyond analytics to embedded knowledge?   Good detailed links at the below article

Putting neural networks under the microscope

Researchers pinpoint the “neurons” in machine-learning systems that capture specific linguistic features during language-processing tasks.

By Rob Matheson | MIT News Office 

Researchers from MIT and the Qatar Computing Research Institute (QCRI) are putting the machine-learning systems known as neural networks under the microscope.

In a study that sheds light on how these systems manage to translate text from one language to another, the researchers developed a method that pinpoints individual nodes, or “neurons,” in the networks that capture specific linguistic features.

Neural networks learn to perform computational tasks by processing huge sets of training data. In machine translation, a network crunches language data annotated by humans, and presumably “learns” linguistic features, such as word morphology, sentence structure, and word meaning. Given new text, these networks match these learned features from one language to another, and produce a translation. ... " 

Pattern Based Thinking

Podcast and Transcript regarding a new book.   Knowledge@Wharton:

Wharton's Eric K. Clemons explains why even the newest business models echo a pattern from successful companies in the past.

 Today’s technology giants, such as Uber and Google, are successful because they introduced something new and innovative to the market, according to conventional wisdom. But Wharton professor of operations, information and decisions Eric K. Clemons thinks that’s too simplistic. Patterns repeat throughout history, and one can find glimpses of today’s new business models in the most successful companies of yore, he says.

 Mastering “pattern-based thinking” will help today’s companies get ahead, Clemons argues. He joined the Knowledge@Wharton radio show on SiriusXM to talk about this mindset, which he encapsulates in his book, New Patterns of Power and Profit: A Strategist’s Guide to Competitive Advantage in the Age of Digital Transformation.


Knowledge@Wharton: Why did you decide to write a book about this topic?

Eric K. Clemons: It’s actually a memoir. It’s the history of a great adventure. About mid-1980s, I realized that economics really understood big industry and [Harvard professor] Michael Porter had said just about everything that needed to be said about strategy for traditional manufacturing, transportation and retailing companies. But traditional industry wasn’t where things were happening. Economics had started to look at the power, the value of information.  ... " 

AI Will Make You Smarter Through Augmentation

Course that is what we are aiming at,  it augments us.   Conversation is a great start.

Artificial intelligence will make you smarter
People plus machines will surpass the capabilities of either element alone. 

By Terrence Sejnowski

Francis Crick Professor and Director of the Computational Neurobiology Laboratory at Salk Institute for Biological Studies, and Distinguished Professor of Neurobiology, University of California San Diego

MIT Press provides funding as a member of The Conversation US.

University of California provides funding as a founding partner of The Conversation US.
Under Creative Commons license.

The future won’t be made by either humans or machines alone – but by both, working together. Technologies modeled on how human brains work are already augmenting people’s abilities, and will only get more influential as society gets used to these increasingly capable machines.

Technology optimists have envisioned a world with rising human productivity and quality of life as artificial intelligence systems take over life’s drudgery and administrivia, benefiting everyone. Pessimists, on the other hand, have warned that these advances could come at great cost in lost jobs and disrupted lives. And fearmongers worry that AI might eventually make human beings obsolete.

However, people are not very good at imagining the future. Neither utopia nor doomsday is likely. In my new book, “The Deep Learning Revolution,” my goal was to explain the past, present and future of this rapidly growing area of science and technology. My conclusion is that AI will make you smarter, but in ways that will surprise you.

Recognizing patterns

Deep learning is the part of AI that has made the most progress in solving complex problems like identifying objects in images, recognizing speech from multiple speakers and processing text the way people speak or write it. Deep learning has also proven useful for identifying patterns in the increasingly large data sets that are being generated from sensors, medical devices and scientific instruments.

The goal of this approach is to find ways a computer can represent the complexity of the world and generalize from previous experience – even if what’s happening next isn’t exactly the same as what happened before. Just as a person can identify that a specific animal she has never seen before is in fact a cat, deep learning algorithms can identify aspects of what might be called “cat-ness” and extract those attributes from new images of cats. .... " 

Imaging Large Human Spaces by Satellite

Impressive patterns revealed.

Imagining gigantic places with satellite imagery   Data Visualization and Statistics by Nathan Yan

Artist Marcus Lyon imagines worlds where there are so many people that the only thing left to do is to make gigantic places to fit everyone. The patterns repeat themselves over and over, and it’s no longer about the individual exploring an entire place. [via kottke] .... " 

AI Aiding Retail

Its all about finding and leveraging the patterns from the data.

How AI Helped One Retailer Reach New Customers  By Dave Sutton in HBR

When Naomi Simson founded RedBalloon, an online gift retailer that sells personal experiences, she was pioneering the category in Australia. With a $25,000 personal investment and a small office in her home, she began aggregating sales leads and aggressively acquiring customers through very traditional marketing means — like yellow page advertisements. It was 2001, and  online advertising was at its nascent stage. Internet Explorer was the leading Internet browser and Google AdWords had only just recently launched. With a cost of customer acquisition of just 5 cents, Simson’s traditional approach to advertising was generating an impressive return on investment. RedBalloon was setting the pace for gifting experiences like outdoor adventures, wine tastings, concert tickets, and spa treatments. ... 

 .... Enter “Albert”, a digital marketing platform powered by artificial intelligence (AI). Working across Facebook, Google, YouTube and other paid and earned media channels, Albert autonomously targets audiences, mixes and matches creative assets, buys media, runs campaigns, measures performance, applies insights from one channel to another, and then makes adjustments based on what “he” learns to optimize the return on marketing investment. I met the team at Albert (formerly known as Adgorithms) while I was doing research into artificial intelligence, machine learning, and data-driven marketing technologies for my book, Marketing, Interrupted. As part of my research, I interviewed several of Albert’s customers, including Simson. .... "

Conversation Patterns

A scarier yet example of pattern recognition, here in conversation.

AI Can Predict Lasting Relationships Based on How You Speak to Your Partner   The Conversation  by  Ian McLoughlin

Researchers at the University of Southern California trained a machine-learning algorithm to understand the relationship between vocal characteristics of couples in therapy and the eventual outcome of therapy, and the program predicted those outcomes with more accuracy than human psychologists. "The significance [of this experiment] is revealing how much information about our underlying feelings is encoded in the way we speak--some of it completely unknown to us," writes University of Kent professor Ian McLoughlin in the U.K. He notes although the therapy participants were speaking naturally, the algorithmic analysis uncovered insights into their mutual feelings they were inadvertently "leaking" into their speech. "This may be one of the first steps in using computers to determine what we are really thinking or feeling," McLoughlin says. He notes potential applications include computers counseling humans about potential partners, or detecting leanings toward antisocial behavior and other negative tendencies or psychological conditions. .... " 

Fract for Geospatial Patterns

Intriguing concept have not heard of this being used.  More at the link.

What is Fract?

We provide actionable prescriptive geospatial intelligence to businesses and help them make all the right choices based on their data.

Fract is inspired by fractals – countless geometric figures that have the exact same characteristics as the whole. Fractals can be found everywhere in nature – snowflakes, clouds and even our own hearts – and are used to describe various complex, recurring natural events like crystal growth and galaxy formation.

This is what we do for you. We predict and help businesses to maximize their potential based on the tiny patterns found inside the data. We believe in continuous data analysis and uncovering patterns to make use of the infinite amounts of data that increases every single hour, of every single day ... " 

Caution with False Patterns

A favorite topic of mine, and a nice piece on caution about finding patterns in data.  Our brains are wired to find patterns, for survival and efficiency.  Worth repeating ...

The inherent clumpiness of randomness
Finding patterns isn't really a question about random processes; it's a question about the human brain. By Mike Loukides  

I've always been interested in random processes. I steered away from writing a Math Honors essay in High School about randomness: that route certainly lead to madness. But the fascination with randomness has persisted, and particularly with what I call the "inherent clumpiness of randomness."  ... "

Rule of Three for Engagement

In Copyblogger:  Using rule of three to create engaging content.

" .... It’s no accident that the number three is pervasive throughout some of our greatest stories, fairy tales, and myths.

It’s also no coincidence that some of the most famous quotes from throughout history are structured in three parts, nor is it surprising that the Rule of Three also works wonders in the world of comedy. It all comes down to the way we humans process information. We have become proficient at pattern recognition by necessity, and three is the smallest number of elements required to create a pattern.

This combination of pattern and brevity results in memorable content, and that’s why the Rule of Three will make you a more engaging writer.  .... "