Tag Archives: carnegie

#434837 In Defense of Black Box AI

Deep learning is powering some amazing new capabilities, but we find it hard to scrutinize the workings of these algorithms. Lack of interpretability in AI is a common concern and many are trying to fix it, but is it really always necessary to know what’s going on inside these “black boxes”?

In a recent perspective piece for Science, Elizabeth Holm, a professor of materials science and engineering at Carnegie Mellon University, argued in defense of the black box algorithm. I caught up with her last week to find out more.

Edd Gent: What’s your experience with black box algorithms?

Elizabeth Holm: I got a dual PhD in materials science and engineering and scientific computing. I came to academia about six years ago and part of what I wanted to do in making this career change was to refresh and revitalize my computer science side.

I realized that computer science had changed completely. It used to be about algorithms and making codes run fast, but now it’s about data and artificial intelligence. There are the interpretable methods like random forest algorithms, where we can tell how the machine is making its decisions. And then there are the black box methods, like convolutional neural networks.

Once in a while we can find some information about their inner workings, but most of the time we have to accept their answers and kind of probe around the edges to figure out the space in which we can use them and how reliable and accurate they are.

EG: What made you feel like you had to mount a defense of these black box algorithms?

EH: When I started talking with my colleagues, I found that the black box nature of many of these algorithms was a real problem for them. I could understand that because we’re scientists, we always want to know why and how.

It got me thinking as a bit of a contrarian, “Are black boxes all bad? Must we reject them?” Surely not, because human thought processes are fairly black box. We often rely on human thought processes that the thinker can’t necessarily explain.

It’s looking like we’re going to be stuck with these methods for a while, because they’re really helpful. They do amazing things. And so there’s a very pragmatic realization that these are the best methods we’ve got to do some really important problems, and we’re not right now seeing alternatives that are interpretable. We’re going to have to use them, so we better figure out how.

EG: In what situations do you think we should be using black box algorithms?

EH: I came up with three rules. The simplest rule is: when the cost of a bad decision is small and the value of a good decision is high, it’s worth it. The example I gave in the paper is targeted advertising. If you send an ad no one wants it doesn’t cost a lot. If you’re the receiver it doesn’t cost a lot to get rid of it.

There are cases where the cost is high, and that’s then we choose the black box if it’s the best option to do the job. Things get a little trickier here because we have to ask “what are the costs of bad decisions, and do we really have them fully characterized?” We also have to be very careful knowing that our systems may have biases, they may have limitations in where you can apply them, they may be breakable.

But at the same time, there are certainly domains where we’re going to test these systems so extensively that we know their performance in virtually every situation. And if their performance is better than the other methods, we need to do it. Self driving vehicles are a significant example—it’s almost certain they’re going to have to use black box methods, and that they’re going to end up being better drivers than humans.

The third rule is the more fun one for me as a scientist, and that’s the case where the black box really enlightens us as to a new way to look at something. We have trained a black box to recognize the fracture energy of breaking a piece of metal from a picture of the broken surface. It did a really good job, and humans can’t do this and we don’t know why.

What the computer seems to be seeing is noise. There’s a signal in that noise, and finding it is very difficult, but if we do we may find something significant to the fracture process, and that would be an awesome scientific discovery.

EG: Do you think there’s been too much emphasis on interpretability?

EH: I think the interpretability problem is a fundamental, fascinating computer science grand challenge and there are significant issues where we need to have an interpretable model. But how I would frame it is not that there’s too much emphasis on interpretability, but rather that there’s too much dismissiveness of uninterpretable models.

I think that some of the current social and political issues surrounding some very bad black box outcomes have convinced people that all machine learning and AI should be interpretable because that will somehow solve those problems.

Asking humans to explain their rationale has not eliminated bias, or stereotyping, or bad decision-making in humans. Relying too much on interpreted ability perhaps puts the responsibility in the wrong place for getting better results. I can make a better black box without knowing exactly in what way the first one was bad.

EG: Looking further into the future, do you think there will be situations where humans will have to rely on black box algorithms to solve problems we can’t get our heads around?

EH: I do think so, and it’s not as much of a stretch as we think it is. For example, humans don’t design the circuit map of computer chips anymore. We haven’t for years. It’s not a black box algorithm that designs those circuit boards, but we’ve long since given up trying to understand a particular computer chip’s design.

With the billions of circuits in every computer chip, the human mind can’t encompass it, either in scope or just the pure time that it would take to trace every circuit. There are going to be cases where we want a system so complex that only the patience that computers have and their ability to work in very high-dimensional spaces is going to be able to do it.

So we can continue to argue about interpretability, but we need to acknowledge that we’re going to need to use black boxes. And this is our opportunity to do our due diligence to understand how to use them responsibly, ethically, and with benefits rather than harm. And that’s going to be a social conversation as well as as a scientific one.

*Responses have been edited for length and style

Image Credit: Chingraph / Shutterstock.com Continue reading

Posted in Human Robots

#434827 AI and Robotics Are Transforming ...

During the past 50 years, the frequency of recorded natural disasters has surged nearly five-fold.

In this blog, I’ll be exploring how converging exponential technologies (AI, robotics, drones, sensors, networks) are transforming the future of disaster relief—how we can prevent them in the first place and get help to victims during that first golden hour wherein immediate relief can save lives.

Here are the three areas of greatest impact:

AI, predictive mapping, and the power of the crowd
Next-gen robotics and swarm solutions
Aerial drones and immediate aid supply

Let’s dive in!

Artificial Intelligence and Predictive Mapping
When it comes to immediate and high-precision emergency response, data is gold.

Already, the meteoric rise of space-based networks, stratosphere-hovering balloons, and 5G telecommunications infrastructure is in the process of connecting every last individual on the planet.

Aside from democratizing the world’s information, however, this upsurge in connectivity will soon grant anyone the ability to broadcast detailed geo-tagged data, particularly those most vulnerable to natural disasters.

Armed with the power of data broadcasting and the force of the crowd, disaster victims now play a vital role in emergency response, turning a historically one-way blind rescue operation into a two-way dialogue between connected crowds and smart response systems.

With a skyrocketing abundance of data, however, comes a new paradigm: one in which we no longer face a scarcity of answers. Instead, it will be the quality of our questions that matters most.

This is where AI comes in: our mining mechanism.

In the case of emergency response, what if we could strategically map an almost endless amount of incoming data points? Or predict the dynamics of a flood and identify a tsunami’s most vulnerable targets before it even strikes? Or even amplify critical signals to trigger automatic aid by surveillance drones and immediately alert crowdsourced volunteers?

Already, a number of key players are leveraging AI, crowdsourced intelligence, and cutting-edge visualizations to optimize crisis response and multiply relief speeds.

Take One Concern, for instance. Born out of Stanford under the mentorship of leading AI expert Andrew Ng, One Concern leverages AI through analytical disaster assessment and calculated damage estimates.

Partnering with the cities of Los Angeles, San Francisco, and numerous cities in San Mateo County, the platform assigns verified, unique ‘digital fingerprints’ to every element in a city. Building robust models of each system, One Concern’s AI platform can then monitor site-specific impacts of not only climate change but each individual natural disaster, from sweeping thermal shifts to seismic movement.

This data, combined with that of city infrastructure and former disasters, are then used to predict future damage under a range of disaster scenarios, informing prevention methods and structures in need of reinforcement.

Within just four years, One Concern can now make precise predictions with an 85 percent accuracy rate in under 15 minutes.

And as IoT-connected devices and intelligent hardware continue to boom, a blooming trillion-sensor economy will only serve to amplify AI’s predictive capacity, offering us immediate, preventive strategies long before disaster strikes.

Beyond natural disasters, however, crowdsourced intelligence, predictive crisis mapping, and AI-powered responses are just as formidable a triage in humanitarian disasters.

One extraordinary story is that of Ushahidi. When violence broke out after the 2007 Kenyan elections, one local blogger proposed a simple yet powerful question to the web: “Any techies out there willing to do a mashup of where the violence and destruction is occurring and put it on a map?”

Within days, four ‘techies’ heeded the call, building a platform that crowdsourced first-hand reports via SMS, mined the web for answers, and—with over 40,000 verified reports—sent alerts back to locals on the ground and viewers across the world.

Today, Ushahidi has been used in over 150 countries, reaching a total of 20 million people across 100,000+ deployments. Now an open-source crisis-mapping software, its V3 (or “Ushahidi in the Cloud”) is accessible to anyone, mining millions of Tweets, hundreds of thousands of news articles, and geo-tagged, time-stamped data from countless sources.

Aggregating one of the longest-running crisis maps to date, Ushahidi’s Syria Tracker has proved invaluable in the crowdsourcing of witness reports. Providing real-time geographic visualizations of all verified data, Syria Tracker has enabled civilians to report everything from missing people and relief supply needs to civilian casualties and disease outbreaks— all while evading the government’s cell network, keeping identities private, and verifying reports prior to publication.

As mobile connectivity and abundant sensors converge with AI-mined crowd intelligence, real-time awareness will only multiply in speed and scale.

Imagining the Future….

Within the next 10 years, spatial web technology might even allow us to tap into mesh networks.

As I’ve explored in a previous blog on the implications of the spatial web, while traditional networks rely on a limited set of wired access points (or wireless hotspots), a wireless mesh network can connect entire cities via hundreds of dispersed nodes that communicate with each other and share a network connection non-hierarchically.

In short, this means that individual mobile users can together establish a local mesh network using nothing but the computing power in their own devices.

Take this a step further, and a local population of strangers could collectively broadcast countless 360-degree feeds across a local mesh network.

Imagine a scenario in which armed attacks break out across disjointed urban districts, each cluster of eye witnesses and at-risk civilians broadcasting an aggregate of 360-degree videos, all fed through photogrammetry AIs that build out a live hologram in real time, giving family members and first responders complete information.

Or take a coastal community in the throes of torrential rainfall and failing infrastructure. Now empowered by a collective live feed, verification of data reports takes a matter of seconds, and richly-layered data informs first responders and AI platforms with unbelievable accuracy and specificity of relief needs.

By linking all the right technological pieces, we might even see the rise of automated drone deliveries. Imagine: crowdsourced intelligence is first cross-referenced with sensor data and verified algorithmically. AI is then leveraged to determine the specific needs and degree of urgency at ultra-precise coordinates. Within minutes, once approved by personnel, swarm robots rush to collect the requisite supplies, equipping size-appropriate drones with the right aid for rapid-fire delivery.

This brings us to a second critical convergence: robots and drones.

While cutting-edge drone technology revolutionizes the way we deliver aid, new breakthroughs in AI-geared robotics are paving the way for superhuman emergency responses in some of today’s most dangerous environments.

Let’s explore a few of the most disruptive examples to reach the testing phase.

First up….

Autonomous Robots and Swarm Solutions
As hardware advancements converge with exploding AI capabilities, disaster relief robots are graduating from assistance roles to fully autonomous responders at a breakneck pace.

Born out of MIT’s Biomimetic Robotics Lab, the Cheetah III is but one of many robots that may form our first line of defense in everything from earthquake search-and-rescue missions to high-risk ops in dangerous radiation zones.

Now capable of running at 6.4 meters per second, Cheetah III can even leap up to a height of 60 centimeters, autonomously determining how to avoid obstacles and jump over hurdles as they arise.

Initially designed to perform spectral inspection tasks in hazardous settings (think: nuclear plants or chemical factories), the Cheetah’s various iterations have focused on increasing its payload capacity, range of motion, and even a gripping function with enhanced dexterity.

Cheetah III and future versions are aimed at saving lives in almost any environment.

And the Cheetah III is not alone. Just this February, Tokyo’s Electric Power Company (TEPCO) has put one of its own robots to the test. For the first time since Japan’s devastating 2011 tsunami, which led to three nuclear meltdowns in the nation’s Fukushima nuclear power plant, a robot has successfully examined the reactor’s fuel.

Broadcasting the process with its built-in camera, the robot was able to retrieve small chunks of radioactive fuel at five of the six test sites, offering tremendous promise for long-term plans to clean up the still-deadly interior.

Also out of Japan, Mitsubishi Heavy Industries (MHi) is even using robots to fight fires with full autonomy. In a remarkable new feat, MHi’s Water Cannon Bot can now put out blazes in difficult-to-access or highly dangerous fire sites.

Delivering foam or water at 4,000 liters per minute and 1 megapascal (MPa) of pressure, the Cannon Bot and its accompanying Hose Extension Bot even form part of a greater AI-geared system to conduct reconnaissance and surveillance on larger transport vehicles.

As wildfires grow ever more untameable, high-volume production of such bots could prove a true lifesaver. Paired with predictive AI forest fire mapping and autonomous hauling vehicles, not only will solutions like MHi’s Cannon Bot save numerous lives, but avoid population displacement and paralyzing damage to our natural environment before disaster has the chance to spread.

But even in cases where emergency shelter is needed, groundbreaking (literally) robotics solutions are fast to the rescue.

After multiple iterations by Fastbrick Robotics, the Hadrian X end-to-end bricklaying robot can now autonomously build a fully livable, 180-square-meter home in under three days. Using a laser-guided robotic attachment, the all-in-one brick-loaded truck simply drives to a construction site and directs blocks through its robotic arm in accordance with a 3D model.

Meeting verified building standards, Hadrian and similar solutions hold massive promise in the long-term, deployable across post-conflict refugee sites and regions recovering from natural catastrophes.

But what if we need to build emergency shelters from local soil at hand? Marking an extraordinary convergence between robotics and 3D printing, the Institute for Advanced Architecture of Catalonia (IAAC) is already working on a solution.

In a major feat for low-cost construction in remote zones, IAAC has found a way to convert almost any soil into a building material with three times the tensile strength of industrial clay. Offering myriad benefits, including natural insulation, low GHG emissions, fire protection, air circulation, and thermal mediation, IAAC’s new 3D printed native soil can build houses on-site for as little as $1,000.

But while cutting-edge robotics unlock extraordinary new frontiers for low-cost, large-scale emergency construction, novel hardware and computing breakthroughs are also enabling robotic scale at the other extreme of the spectrum.

Again, inspired by biological phenomena, robotics specialists across the US have begun to pilot tiny robotic prototypes for locating trapped individuals and assessing infrastructural damage.

Take RoboBees, tiny Harvard-developed bots that use electrostatic adhesion to ‘perch’ on walls and even ceilings, evaluating structural damage in the aftermath of an earthquake.

Or Carnegie Mellon’s prototyped Snakebot, capable of navigating through entry points that would otherwise be completely inaccessible to human responders. Driven by AI, the Snakebot can maneuver through even the most densely-packed rubble to locate survivors, using cameras and microphones for communication.

But when it comes to fast-paced reconnaissance in inaccessible regions, miniature robot swarms have good company.

Next-Generation Drones for Instantaneous Relief Supplies
Particularly in the case of wildfires and conflict zones, autonomous drone technology is fundamentally revolutionizing the way we identify survivors in need and automate relief supply.

Not only are drones enabling high-resolution imagery for real-time mapping and damage assessment, but preliminary research shows that UAVs far outpace ground-based rescue teams in locating isolated survivors.

As presented by a team of electrical engineers from the University of Science and Technology of China, drones could even build out a mobile wireless broadband network in record time using a “drone-assisted multi-hop device-to-device” program.

And as shown during Houston’s Hurricane Harvey, drones can provide scores of predictive intel on everything from future flooding to damage estimates.

Among multiple others, a team led by Texas A&M computer science professor and director of the university’s Center for Robot-Assisted Search and Rescue Dr. Robin Murphy flew a total of 119 drone missions over the city, from small-scale quadcopters to military-grade unmanned planes. Not only were these critical for monitoring levee infrastructure, but also for identifying those left behind by human rescue teams.

But beyond surveillance, UAVs have begun to provide lifesaving supplies across some of the most remote regions of the globe. One of the most inspiring examples to date is Zipline.

Created in 2014, Zipline has completed 12,352 life-saving drone deliveries to date. While drones are designed, tested, and assembled in California, Zipline primarily operates in Rwanda and Tanzania, hiring local operators and providing over 11 million people with instant access to medical supplies.

Providing everything from vaccines and HIV medications to blood and IV tubes, Zipline’s drones far outpace ground-based supply transport, in many instances providing life-critical blood cells, plasma, and platelets in under an hour.

But drone technology is even beginning to transcend the limited scale of medical supplies and food.

Now developing its drones under contracts with DARPA and the US Marine Corps, Logistic Gliders, Inc. has built autonomously-navigating drones capable of carrying 1,800 pounds of cargo over unprecedented long distances.

Built from plywood, Logistic’s gliders are projected to cost as little as a few hundred dollars each, making them perfect candidates for high-volume remote aid deliveries, whether navigated by a pilot or self-flown in accordance with real-time disaster zone mapping.

As hardware continues to advance, autonomous drone technology coupled with real-time mapping algorithms pose no end of abundant opportunities for aid supply, disaster monitoring, and richly layered intel previously unimaginable for humanitarian relief.

Concluding Thoughts
Perhaps one of the most consequential and impactful applications of converging technologies is their transformation of disaster relief methods.

While AI-driven intel platforms crowdsource firsthand experiential data from those on the ground, mobile connectivity and drone-supplied networks are granting newfound narrative power to those most in need.

And as a wave of new hardware advancements gives rise to robotic responders, swarm technology, and aerial drones, we are fast approaching an age of instantaneous and efficiently-distributed responses in the midst of conflict and natural catastrophes alike.

Empowered by these new tools, what might we create when everyone on the planet has the same access to relief supplies and immediate resources? In a new age of prevention and fast recovery, what futures can you envision?

Join Me
Abundance-Digital Online Community: I’ve created a Digital/Online community of bold, abundance-minded entrepreneurs called Abundance-Digital. Abundance-Digital is my ‘onramp’ for exponential entrepreneurs – those who want to get involved and play at a higher level. Click here to learn more.

Image Credit: Arcansel / Shutterstock.com Continue reading

Posted in Human Robots

#434616 What Games Are Humans Still Better at ...

Artificial intelligence (AI) systems’ rapid advances are continually crossing rows off the list of things humans do better than our computer compatriots.

AI has bested us at board games like chess and Go, and set astronomically high scores in classic computer games like Ms. Pacman. More complex games form part of AI’s next frontier.

While a team of AI bots developed by OpenAI, known as the OpenAI Five, ultimately lost to a team of professional players last year, they have since been running rampant against human opponents in Dota 2. Not to be outdone, Google’s DeepMind AI recently took on—and beat—several professional players at StarCraft II.

These victories beg the questions: what games are humans still better at than AI? And for how long?

The Making Of AlphaStar
DeepMind’s results provide a good starting point in a search for answers. The version of its AI for StarCraft II, dubbed AlphaStar, learned to play the games through supervised learning and reinforcement learning.

First, AI agents were trained by analyzing and copying human players, learning basic strategies. The initial agents then played each other in a sort of virtual death match where the strongest agents stayed on. New iterations of the agents were developed and entered the competition. Over time, the agents became better and better at the game, learning new strategies and tactics along the way.

One of the advantages of AI is that it can go through this kind of process at superspeed and quickly develop better agents. DeepMind researchers estimate that the AlphaStar agents went through the equivalent of roughly 200 years of game time in about 14 days.

Cheating or One Hand Behind the Back?
The AlphaStar AI agents faced off against human professional players in a series of games streamed on YouTube and Twitch. The AIs trounced their human opponents, winning ten games on the trot, before pro player Grzegorz “MaNa” Komincz managed to salvage some pride for humanity by winning the final game. Experts commenting on AlphaStar’s performance used words like “phenomenal” and “superhuman”—which was, to a degree, where things got a bit problematic.

AlphaStar proved particularly skilled at controlling and directing units in battle, known as micromanagement. One reason was that it viewed the whole game map at once—something a human player is not able to do—which made it seemingly able to control units in different areas at the same time. DeepMind researchers said the AIs only focused on a single part of the map at any given time, but interestingly, AlphaStar’s AI agent was limited to a more restricted camera view during the match “MaNA” won.

Potentially offsetting some of this advantage was the fact that AlphaStar was also restricted in certain ways. For example, it was prevented from performing more clicks per minute than a human player would be able to.

Where AIs Struggle
Games like StarCraft II and Dota 2 throw a lot of challenges at AIs. Complex game theory/ strategies, operating with imperfect/incomplete information, undertaking multi-variable and long-term planning, real-time decision-making, navigating a large action space, and making a multitude of possible decisions at every point in time are just the tip of the iceberg. The AIs’ performance in both games was impressive, but also highlighted some of the areas where they could be said to struggle.

In Dota 2 and StarCraft II, AI bots have seemed more vulnerable in longer games, or when confronted with surprising, unfamiliar strategies. They seem to struggle with complexity over time and improvisation/adapting to quick changes. This could be tied to how AIs learn. Even within the first few hours of performing a task, humans tend to gain a sense of familiarity and skill that takes an AI much longer. We are also better at transferring skill from one area to another. In other words, experience playing Dota 2 can help us become good at StarCraft II relatively quickly. This is not the case for AI—yet.

Dwindling Superiority
While the battle between AI and humans for absolute superiority is still on in Dota 2 and StarCraft II, it looks likely that AI will soon reign supreme. Similar things are happening to other types of games.

In 2017, a team from Carnegie Mellon University pitted its Libratus AI against four professionals. After 20 days of No Limit Texas Hold’em, Libratus was up by $1.7 million. Another likely candidate is the destroyer of family harmony at Christmas: Monopoly.

Poker involves bluffing, while Monopoly involves negotiation—skills you might not think AI would be particularly suited to handle. However, an AI experiment at Facebook showed that AI bots are more than capable of undertaking such tasks. The bots proved skilled negotiators, and developed negotiating strategies like pretending interest in one object while they were interested in another altogether—bluffing.

So, what games are we still better at than AI? There is no precise answer, but the list is getting shorter at a rapid pace.

The Aim Of the Game
While AI’s mastery of games might at first glance seem an odd area to focus research on, the belief is that the way AI learn to master a game is transferrable to other areas.

For example, the Libratus poker-playing AI employed strategies that could work in financial trading or political negotiations. The same applies to AlphaStar. As Oriol Vinyals, co-leader of the AlphaStar project, told The Verge:

“First and foremost, the mission at DeepMind is to build an artificial general intelligence. […] To do so, it’s important to benchmark how our agents perform on a wide variety of tasks.”

A 2017 survey of more than 350 AI researchers predicts AI could be a better driver than humans within ten years. By the middle of the century, AI will be able to write a best-selling novel, and a few years later, it will be better than humans at surgery. By the year 2060, AI may do everything better than us.

Whether you think this is a good or a bad thing, it’s worth noting that AI has an often overlooked ability to help us see things differently. When DeepMind’s AlphaGo beat human Go champion Lee Sedol, the Go community learned from it, too. Lee himself went on a win streak after the match with AlphaGo. The same is now happening within the Dota 2 and StarCraft II communities that are studying the human vs. AI games intensely.

More than anything, AI’s recent gaming triumphs illustrate how quickly artificial intelligence is developing. In 1997, Dr. Piet Hut, an astrophysicist at the Institute for Advanced Study at Princeton and a GO enthusiast, told the New York Times that:

”It may be a hundred years before a computer beats humans at Go—maybe even longer.”

Image Credit: Roman Kosolapov / Shutterstock.com Continue reading

Posted in Human Robots

#433276 Army researchers teaching robots to be ...

Researchers at the U.S. Army Research Laboratory and the Robotics Institute at Carnegie Mellon University developed a new technique to quickly teach robots novel traversal behaviors with minimal human oversight. Continue reading

Posted in Human Robots

#432333 Pipe-crawling robot will help ...

A pair of autonomous robots developed by Carnegie Mellon University's Robotics Institute will soon be driving through miles of pipes at the U.S. Department of Energy's former uranium enrichment plant in Piketon, Ohio, to identify uranium deposits on pipe walls. Continue reading

Posted in Human Robots