Curiosity and AI

Curiosity is often key to solution.

MIT News | Massachusetts Institute of Technology

 SUBSCRIBE TO MIT NEWS NEWSLETTER

Ensuring AI works with the right dose of curiosity

Researchers make headway in solving a longstanding problem of balancing curious “exploration” versus “exploitation” of known pathways in reinforcement learning.

Rachel Gordon | MIT CSAIL, Publication Date:November 10, 2022

It’s a dilemma as old as time. Friday night has rolled around, and you’re trying to pick a restaurant for dinner. Should you visit your most beloved watering hole or try a new establishment, in the hopes of discovering something superior? Potentially, but that curiosity comes with a risk: If you explore the new option, the food could be worse. On the flip side, if you stick with what you know works well, you won't grow out of your narrow pathway. 

Curiosity drives artificial intelligence to explore the world, now in boundless use cases — autonomous navigation, robotic decision-making, optimizing health outcomes, and more. Machines, in some cases, use “reinforcement learning” to accomplish a goal, where an AI agent iteratively learns from being rewarded for good behavior and punished for bad. Just like the dilemma faced by humans in selecting a restaurant, these agents also struggle with balancing the time spent discovering better actions (exploration) and the time spent taking actions that led to high rewards in the past (exploitation). Too much curiosity can distract the agent from making good decisions, while too little means the agent will never discover good decisions.

In the pursuit of making AI agents with just the right dose of curiosity, researchers from MIT’s Improbable AI Laboratory and Computer Science and Artificial Intelligence Laboratory (CSAIL) created an algorithm that overcomes the problem of AI being too “curious” and getting distracted by a given task. Their algorithm automatically increases curiosity when it's needed, and suppresses it if the agent gets enough supervision from the environment to know what to do.  ... ' 

Considering the forms of Curiosity

Intriguing idea, is creativity just adding a goal to curiosity?

Automating the Search for New 'Curiosity' Algorithms
MIT News
Kim Martineau

Massachusetts Institute of Technology (MIT) researchers have used machine learning to identify new algorithms for encoding forms of curiosity. MIT's Ferran Alet said, "We were inspired to use [artificial intelligence] to find algorithms with curiosity strategies that can adapt to a range of environments." The researchers chose a set of basic modules to define their exploration algorithms, selecting about 36 high-level operations to guide the agent to perform tasks like remembering previous inputs, comparing current and past inputs, and using learning methods to change its own modules. The computer integrated up to seven operations at a time to generate computation graphs describing 52,000 algorithms. Two were entirely new, or apparently too obvious or counterintuitive to be of human design, and outperformed human-designed algorithms on a wide range of simulated tasks and environments.  .... " 

Curiosity Driven Data Science

Any time you examine data analytically you exercise your curiosity.    And that curiosity is naturally driven by your own or business goals.  So it makes sense to attach the goals more systematically to the analysis, whether it is as simple as a regression,  or complex as a trained neural net.  Just make sure you do exploratory examination of your results.

Curiosity-Driven Data Science      By Eric Colson in HBR

Data science can enable wholly new and innovative capabilities that can completely differentiate a company. But those innovative capabilities aren’t so much designed or envisioned as they are discovered and revealed through curiosity-driven tinkering by the data scientists. So, before you jump on the data science bandwagon, think less about how data science will support and execute your plans and think more about how to create an environment to empower your data scientists to come up with things you never dreamed of.

First, some context. I am the Chief Algorithms Officer at Stitch Fix, an online personalized styling service with 2.7 million clients in the U.S. and plans to enter the U.K. next year. The novelty of our service affords us exclusive and unprecedented data with nearly ideal conditions to learn from it. We have more than 100 data scientists that power algorithmic capabilities used throughout the company. We have algorithms for recommender systems, merchandise buying, inventory management, relationship management, logistics, operations — we even have algorithms for designing clothes! Each provides material and measurable returns, enabling us to better serve our clients, while providing a protective barrier against competition. Yet, virtually none of these capabilities were asked for by executives, product managers, or domain experts — and not even by a data science manager (and certainly not by me). Instead, they were born out of curiosity and extracurricular tinkering by data scientists.

Data scientists are a curious bunch, especially the good ones. They work towards clear goals, and they are focused on and accountable for achieving certain performance metrics. But they are also easily distracted, in a good way. In the course of doing their work they stumble on various patterns, phenomenon, and anomalies that are unearthed during their data sleuthing. This goads the data scientist’s curiosity: “Is there a better way that we can characterize a client’s style?” “If we modeled clothing fit as a distance measure could we improve client feedback?” “Can successful features from existing styles be re-combined to create better ones?” To answer these questions, the data scientist turns to the historical data and starts tinkering. They don’t ask permission. In some cases, explanations can be found quickly, in only a few hours or so. Other times, it takes longer because each answer evokes new questions and hypotheses, leading to more testing and learning. .... " 

Machines Learn to be Curious

Clever Machines Learn How to Be Curious 
Quanta Magazine  by John Pavlus

In Quanta.  Machines Learn to be Curious.
The University of California, Berkeley's Pulkit Agrawal is embedding curiosity, or intrinsic motivation, into artificial intelligence (AI) so machines can learn unfamiliar tasks more efficiently. "You can think of curiosity as a kind of reward, which the agent generates internally on its own, so that it can go explore more about its world," Agrawal says. His team at the Berkeley Artificial Intelligence Research Lab have developed a learning agent with an intrinsic curiosity module (ICM) so it can learn to play a video game. Their model is young children's innate curiosity in objects that surprise them, mimicked by the ICM's generation of an intrinsic reward signal defined by how mistaken the prediction model is. Agrawal says this reward for being surprised creates a feedback loop enabling the AI to correct its ignorance. The AI also translates visual input from raw pixels into an abstract version of reality, highlighting environmental features with the potential to affect the agent.   .... " 

Science of Creativity

AI and Machine learning find us answers, but do they make us curious about deeper values? Discussion podcast:

" ...  Astrophysicist and author Mario Livio discusses his new book on curiosity.

 Curiosity is a fundamental human trait. Everyone is curious, but the object and degree of that curiosity is different depending on the person and the situation. Astrophysicist and author Mario Livio was so curious about curiosity that he wrote a book about it. He recently appeared on the Knowledge@Wharton show on SiriusXM channel 111 to talk about what he learned in the course of writing his book, Why? What Makes Us Curious.

An edited transcript of the conversation follows.

Knowledge@Wharton: What is it that really drives our curiosity?  ... " 

Curiosity and AI

How do you build curiosity in a system.  And the ability to act on curiosity?  In a simplistic sense, mining for data and then ingesting it?   If you are a robot or drone with sensors your have an element of being able to reach out for it as well.  And measure the value of what you find.  Back to the 'Explore/Exploit' dilemma.  Mentioned by Brian Christian in his recent talk.   Good piece in AAAS:

Scientists imbue robots with curiosity  By Matthew Hutson 

In a twist on artificial intelligence (AI), computer scientists have programmed machines to be curious—to explore their surroundings on their own and learn for the sake of learning. The new approach could allow robots to learn even faster than they can now. Someday they might even surpass human scientists in forming hypotheses and pushing the frontiers of what’s known.

“Developing curiosity is a problem that’s core to intelligence,” says George Konidaris, a computer scientist who runs the Intelligent Robot Lab at Brown University and was not involved in the research. “It’s going to be most useful when you’re not sure what your robot is going to have to do in the future.”  .... " 

Over the years, scientists have worked on algorithms for curiosity, but copying human inquisitiveness has been tricky. For example, most methods aren’t capable of assessing artificial agents’ gaps in knowledge to predict what will be interesting before they see it. (Humans can sometimes judge how interesting a book will be by its cover.)

Todd Hester, a computer scientist currently at Google DeepMind in London hoped to do better. “I was looking for ways to make computers learn more intelligently, and explore as a human would,” he says. “Don’t explore everything, and don’t explore randomly, but try to do something a little smarter.”

Curiosity and Creativity in an AI

Something we worried about when we built logic based AIs.   One executive rejected the effort, saying:  "  .. I want something new, fresh ... not what comes from old expertise, or old data ... ".   In the HBR:

Can AI Ever Be as Curious as Humans?
by Tomas Chamorro-Premuzic and Ben Taylor

 Curiosity has been hailed as one of the most critical competencies for the modern workplace. It’s been shown to boost people’s employability. Countries with higher curiosity enjoy more economic and political freedom, as well as higher GDPs. It is therefore not surprising that, as future jobs become less predictable, a growing number of organizations will hire individuals based on what they could learn, rather than on what they already know.   ... " 

From Natural Language to AI

Reminds me of some of our own investigations of building more perceptive AIs and bots.  We still don't have a workable language model of our world.  Limited, focused models do exist,  and are called Ontologies.

The Two Paths from Natural Language Processing to Artificial Intelligence  By Jonathan Mugan
Robot psychologist. Co-founder and CEO at DeepGrammar, a startup specializing in NLP and deep learning. Author of The Curiosity Cycle.

Why isn’t Siri smarter? AI has accelerated in recent years, especially with deep learning, but current chatbots are an embarrassment. Computers still can’t read or converse intelligently. Their deficiency is disappointing because we want to interact with our world using natural language, and we want computers to read all of those documents out there so they can retrieve the best ones, answer our questions, and summarize what is new.

To understand our language, computers need to know our world. They need to be able to answer questions like “Why does it only rain outside?” and “If a book is on a table, and you push the table, what happens?” .... " 

Autonomous Selection of Mars Laser Targets

An example of autonomy for a complex system.  Here the Curiosity Rover on Mars.   At one level this is similar to closed loop process control, but with more complex sensor analysis being done in that loop.  

Assume this increases accuracy, speed in going through analysis goals ... and even decreases targeting labor required to enact,  thus decreasing cost.     So is  closed loop process control we did in manufacturing, though the adjustments here appear to start to arise to the strategic.   Article and image examples:

From the Jet Propulsion Lab: 
  " .... NASA's Mars rover Curiosity is now selecting rock targets for its laser spectrometer -- the first time autonomous target selection is available for an instrument of this kind on any robotic planetary mission.

Using software developed at NASA's Jet Propulsion Laboratory, Pasadena, California, Curiosity is now frequently choosing multiple targets per week for a laser and a telescopic camera that are parts of the rover's Chemistry and Camera (ChemCam) instrument. Most ChemCam targets are still selected by scientists discussing rocks or soil seen in images the rover has sent to Earth, but the autonomous targeting adds a new capability. ... "

Empirical Creativity

Imagination Man ...

Scott Barry Kaufman has been called “the leading empirical creativity researcher of his generation.” Now he wants to use the tools he’s developed to unleash the “quiet potential” of vulnerable people—including kids like him—and help them flourish. .... 

Earlier this year Kaufman published Wired to Create: Unraveling the Mysteries of the Creative Mind, co-authored with journalist Carolyn Gregoire, in which he argues that creativity—with all of its questioning, close observation, and thinking differently—can be another (or even better) way to evaluate smarts than traditional measures. “You can have a very high IQ and still be ‘dumb’ in my opinion,” he says, “if you score low on things like intellectual curiosity and openness to experience.”

He wrote the book “to bring together the latest science on creativity and get it out there to a popular audience as a platform for talking about the importance of creativity and imagination as a way of living. A lot of it comes out of research I’ve been conducting [on ideas] like the importance of ‘openness to experience.’”

By  Joann Greco

Big Data Isn't Enough

Colleague Martin Lindstrom's recent book:  Small Data: The Tiny Clues that Uncover Huge Trends, was recently covered in a Fortune article.    " .... It can’t give you the whole picture .... Here are 5 reasons why Big Data no longer can stand alone, and why the future is likely to always include Small Data .... "   

The Master storyteller looks at the small indicators:  

" ... Hired by the world's leading brands to find out what makes their customers tick, Martin Lindstrom spends 300 nights a year in strangers’ homes, carefully observing every detail in order to uncover their hidden desires, and, ultimately, the clues to a multi-million dollar product.

Lindstrom connects the dots in this globetrotting narrative that will enthrall enterprising marketers, as well as anyone with a curiosity about the endless variations of human behavior.  ... " 

Playful City Designs

Not only practical design, but how about playful design?  Here a UK effort on the idea.  Not gamification dynamics in the sense of competition, but promoting curiosity and even innovation. Too often forgotten.   Also with a strong element of cognitive language interaction. " ... Hello Lamp Post and the idea of playful cities ...  A new project that allows city dwellers to converse with urban objects begins in Bristol this July. We speak to the creators about this fascinating collision of Smart Cities and location-based gaming  ... " 

Become a Watson Analytics Expert

Gives you a good introduction about how this all works.  Non technical.  By Francis Hall:

" ... Watson Analytics is a tool that’s been designed for business users. It’s not technical; you don’t need any prior knowledge of data science or statistics to start gaining actionable insights from your data, all that’s required is a sound knowledge of your business, and a certain level of curiosity. By the way, it’s free to sign up for and use (not just a timed trial) – you can sign up and get started today here. 

There’s more to Watson Analytics than may appear at first glance. I hope the following tips will help take you to the next level in your analysis. ... " 

Try it for free.

PureMatter Marketing

Newly discovered:  PureMatter.

" ... Since PureMatter was founded in 2002, the world has changed. The economy tanked. Technology exploded. Expectations keep increasing. Consumers have never been so connected – and there’s no end in sight. Today, we’re excited to be leading the charge toward humanizing business, being led by Bryan Kramer, the instigator of the #H2H movement. (Kramer's Shareology book) 

As marketers, we use our smartest thinking to stay ahead of these shifts. Our process helps our clients find the very best energy in their marketing that attracts the right customers, and then keeps them there over time. Energy creates momentum. Extreme heat. And nuclear results.

Our process is simple. We leverage technology, structure and methodology to run our business. Smart thinking to drive our strategy. And curiosity and humor to manifest creativity. It keeps us efficient, focused and disciplined, and keeps clients coming back for more. Simplifying the complex is our specialty.

We’re pattern and trend-spotters. Scientists of life. And purveyors of cracking what makes marketing work today. Our formula’s working, and have picked up numerous local, regional and National awards along the way.  ... "   (See also their blog)  @PureMatter 

Analytics Magazine Addresses Human Curiosity in AI

Analytics Magazine, Jul-Aug 2015

Building Human Curiosity into A.I.
by Scott Zoldi
Self-learning models: How neuro-dynamic programming enables smart machines to think ahead.

Predicting Patient Experience  
by Sagar Anisingaraju and Mo Kaushal
Why narrative data is a healthcare goldmine: Four reasons to feel confident about the "consumerization" wave.

Curing What Ails the Healthcare Industry
by Charlie Bitzis
Interaction analytics is transforming health insurance: Four key areas where analytics is driving change.

Network Analytics for Everyone
by Will Towler
From social media to healthcare, more and more fields are turning to network science for greater insight.

Networks vs. Fraud: Connecting the Dots
by Bart Baesens, Véronique Van Vlasselaer and Wouter Verbeke
Well-constructed analytical models useful in thwarting fraudsters and their complex but revealing patterns.

Data Science and Journalism

Good CACM overview of the increasing use of advanced data science in journalism

" .. The key attributes journalists must have—the ability to separate fact from opinion, a willingness to find and develop strong sources, and the curiosity to ask probing, intelligent questions—are still relevant in today's 140-character-or-less, ADHD-esque society. Yet increasingly, journalists dealing with technical topics often found in science or technology are turning to tools that were once solely the province of data analysts and computer scientists. ,,, " 

Fidelity Labs Develops StockCity

Have examined a number of methods that use realistic spaces to represent abstract data. Never found these methods that useful beyond attracting curiosity.  Here another example by Fidelity. With the latest technology:

" ... Fidelity isn't the first company to launch an interactive virtual city app, but it's the first to create one in which the city is built entirely out of stocks. StockCity, developed by Fidelity's research and development think tank Fidelity Labs, is a data visualization app that recreates investors' portfolios as a 3-D city.

StockCity was developed for virtual reality headset Oculus Rift, and it's also compatible with the Google Chrome Web browser. In StockCity, each stock in a portfolio is a skyscraper, and the height of each tall building is determined by its closing stock price. Buildings are clustered together in neighborhoods that represent market sectors.  .... " 

See also, Fidelity Labs.

Is Your Content Compelling?

Could be quite valuable for a number of areas.  This is being experimented with every day in advertising and online.  The data to confirm or deny it is considerable.  Have not read as yet.

  In Fastcocreate:   " ... Riveted, A book from cognitive science professor Jim Davies, presents a unified theory of compellingness. ....  What makes for compelling art? Any creator who has given half a thought to paying the rent, or achieving immortality, has considered what makes art sell. We know that the notion of quality--the idea that "the best" art and marketing and media reaches the most people--is insufficient to explain what gives some creations mass appeal. So why do people--large number of people--find books, ads, movies and art works compelling? How can we know, ahead of time, what will pique our curiosity and sustain our interest?   ... " 

Linkedin Retiring InMaps

We were exposed to this tool some time ago as a means to gain insight into corporate structure.  Now Linkedin is retiring it.  Would like to hear of cases where it was used practically by organizations, beyond simple curiosity.  The service is now already unsupported.  Linkedin says they plan to replace it with an alternative network visualization method.  Previously here on Inmaps.

Interdisciplinary Computing and Computer Science

Had an interesting conversation with Lisa C. Kaczmarczyk See her Interdisciplinary Computing Blog :" ... Computing and people who work with computers are not the nerdy and negative images often portrayed in the media. As a computer scientist, educator and author with my hands and feet in many fields I live these realities every day. I am like the kid who never stops asking “why?” In this blog, I share my questions and curiosity about the interdisciplinary role of computing with a special concern for how computing can make the world a better place. ... "