Blog

Advances in AI

A number of recent stories illustrate the rapid progress that’s being made in many different fields linked to artificial intelligence:

Europe’s four-year AMARSi project (Adaptive Modular Architecture for Rich Motor Skills) could lead to humanoid robots that quickly acquire skills from human co-workers as well as their environments.
AMARSi project could see robots learn from co-workers

Also from Europe is HUMAVIPS (Humanoids with Auditory and Visual Abilities in Populated Spaces) which seeks to improve the ways humans and robots communicate and interact.
Humanoid robots to gain advanced social skills

A research team at the University of Michigan at Ann Arbor has demonstrated how memristors can mimic the behavior of brain synapses.
Electronics ‘missing link’ brings neural computing closer

The Intelligence Revolution (Part 1)

In the intelligent future, positions, roles, entire fields will disappear from the jobs market. In many cases, the tasks themselves will still exist but increasingly they’ll be done by evermore intelligent technologies instead of by humans. As in the past, these changes will be motivated by market forces and the ongoing incentive to improve efficiency and productivity. And like the Digital Revolution and the Industrial Revolution before it, the Intelligence Revolution will transform the jobs landscape for still another reason. Because it will lead to increasingly rapid innovation and product creation.

In the past, machines have eliminated jobs because they could be designed to be faster, stronger and more tireless than their flesh and blood counterparts. During the Industrial Revolution, this meant lower skilled, often highly-repetitive labor was displaced. Spinning, weaving and sewing machines transformed the work environment. Vast numbers of trades people saw their work taken away by what would eventually become a worldwide clothing industry. Which would itself one day create millions of jobs. Likewise, unskilled manual labor. Before the beginning of the Industrial Revolution, roads and railways were built on the backs of thousands of workers. Today with the aid of machines, dozens can do the work previously performed by hundreds, if not thousands. These technologies (along with many others that came after them) built the cities, highways and factories that led to still more jobs.

The Digital Revolution had similar impacts. Office functions have seen such increases in efficiency that a single person can perform the work once performed by many. At the same time, that worker has seen a huge increase in available skills and technologies. These changes also took away a lot of jobs but led to the creation of many new ones as well. Desktop publishing, digital photography and digital video production utterly transformed their fields. At the same time these new technologies eliminated jobs, they put the ability to create mass media into the hands of an unprecedented number of people.

This antenna was created for a NASA communication satellite using evolutionary algorithms.  It was designed to be efficient, not pretty.

But I suspect these examples will pale beside the changes that will be brought about by the Intelligence Revolution. Two fields that are poised to transform entire industries are those of automated invention and rapid prototyping. Using techniques such as genetic programming (algorithms that utilize fitness selection), computers are increasingly able to invent devices that are far more efficient than anything that could be created by a person. For instance, Evolutionary Antenna Synthesis is leading to better, more cost-effectively designed antennas. In another field, evolvable hardware has improved microprocessor design, creating circuits that are far denser, increasingly complex and fault-tolerant than anything that could be created by a team of unaided humans.

As they develop, 3-D printer rapid-prototyping technologies will allow us to implement these inventions and designs far faster than we ever could before. The creation of open source rapid-prototypers, such as the RepRap and MakerBot projects, will make these technologies widely available, leading to a democratization of this process as well. 3-D designs for these machines are already widely shared and distributed via the internet and the concept will likely lead to downward price pressure as these technologies become increasingly sophisticated.

All of this is leading to a sea change for job markets and how we approach our careers and education. As always, such change will have consequences both good and bad. I’ll elaborate on this further in Part 2 of this series.

Natural responses to technological change

This music was composed by a prolific, young composer named Emi. During a brief career, Emi created thousands of works, many of them in the style of famous composers, such as Bach, Beethoven, and Chopin. These compositions have been described as moving, soulful, emotional. In many ways, Emi may prove to be one of the most important composers of the age. Emi is also a computer program.

When many listeners first learn the nature of these compositions, particularly after believing they were created by a person, their opinion of the work changes dramatically. Music they previously called soulful or inspiring becomes calculating, stilted, even mechanistic. This points to a human chauvinism about the growing intelligence of our machines which we would be wise to examine. As with so many developments in artificial intelligence, there seems to be resistance against any encroachment into what we regard as the realm of human intellect. Just as chess masters once denigrated the abilities of chess-playing computers, each new hurdle in AI will probably be met with similar resistance.

Though it may sound a little strange, I’m going to suggest that this response follows a pattern similar to Kubler-Ross’s five stages of grief: denial, anger, bargaining, depression and acceptance. Grief is a normal response to loss and it may be that many people feel they are losing something crucial when technology begins to take on functions previously exclusive to humans. Obviously, disbelief is a common enough reaction to new and developing technology, but it’s hard to maintain for long in the face of direct evidence. Emi’s creator, University of California, Santa Cruz professor, David Cope, has reported considerable anger from listeners in response to Emi’s work, as well as his later program, Emily Howell.

It’s almost as if the listener felt tricked or cheated once they discovered the music was created by an artificial intelligence rather than a natural one. As for bargaining and depression, I suspect these are not unfamiliar to people who have seen their career plans radically altered by the continually shifting technological landscape.

Then there’s acceptance. So much AI has been accepted that most of us no longer categorize it as AI at all. When was the last time you used spell check, or a voice menu or played a video game and acknowledged the leaps in artificial intelligence that made them possible? Typically, we don’t. They simply become another piece of the background of our technologically enhanced lives.

All of these are probably very natural responses. We humans have considered ourselves to stand at the pinnacle of intelligence for so very long. Nothing in the animal kingdom even comes close to the height of human achievement and for the most part, our machines are even further down the scale. Many will argue that the types of intelligences exhibited by these machines is very limited and domain-specific. That they are examples of weak AI – sets of rules and knowledge bases and Bayesian pattern recognition algorithms. There’s no way any of these could ever develop into something that would rival our marvelous minds.

But more and more evidence is indicating that our brains are themselves composed of myriad subsystems which together make up the sum total of our intellect. Marvin Minsky’sSociety of Mind” describes the concept very well. Why shouldn’t an artificial general intelligence be composed of modules, agents and subsystems too?

With each new leap in machine intelligence, we come a little closer to slipping from our pinnacle on Mount Intellect. Lightning-fast calculation, expert diagnostic systems, face and image recognition, real-time voice translation. Exponential improvements in both hardware and software are rapidly driving us into a new era. It will be an era in which we’ll likely share the stage of superior intelligence with many other players. That is, we will if we’re lucky.

Smart dust for making smarter decisions

Imagine if you didn’t have a nervous system.  Your body would have no way of regulating any of its other systems.  It wouldn’t know if it was too hot or too cold.  It couldn’t register dangers and harmful conditions.  Every aspect of the environment would be shut off to it.  In short, without a nervous system, you wouldn’t survive very long.

The fact is, in order to know how to respond to conditions you need to know what those conditions are.  This is some of the thinking behind “smart dust” – very small, very cheap networked sensors for measuring all kinds of different aspects of our environment. It’s also the idea behind HP’s “Central Nervous System for the Earth” project or CeNSE.  By developing sensors that can detect motion, vibration, light, temperature, air pressure, air flow and humidity, HP hopes to see them deployed throughout the environment.  These will be able to keep watch over the structural integrity of buildings, bridges and other infrastructure.  Chemical sensors will be able to detect dangerous conditions in our air, food and water.  They’ll eventually be capable of alerting us in the event of a terrorist attack using biological agents.  In short, they’ll be our eyes, ears, noses and much more.  They’ll become a new kind of nervous system.

This would be tremendously useful for monitoring manmade structures.  A US DOT 2008 survey of over 600,00 bridges found nearly 27% to be structurally deficient or functionally obsolete.  There is simply no way we have a sufficient number of trained people to adequately monitor all of these.  And that’s only the bridges.  A system of inexpensive sensors that can watch for excessive vibration levels and structural deformity in order to avoid catastrophic collapse will be money very well spent.

Obviously, the environment already has its own kind of feedback loops through which it adapts to changing conditions.  But humanity has imposed itself so thoroughly onto the environment that we need better ways of gauging our effect on it.  Hopefully, this kind of data will allow us to make better, more informed decisions about mitigating environmental impact.  Certainly, this would be preferable to making knee-jerk, expensive, politically feel-good decisions that often do more harm than good.  (e.g., subsidizing the conversion of food crops to ethanol crops and thereby exacerbating food shortages in parts of the world.)

Obviously, there will be downsides to this kind of technology, most notably in terms of its potential use in surveillance.  As with most technological developments, the answer is not in trying to prohibit it but to adapt our laws and institutions to deal with our changing world.  A world we will be knowing much more about, very shortly.

Printable electronics

Xerox silver ink is a breakthrough for printable electronics

Xerox announced this week they’ve developed several breakthroughs in the printing of low-cost electronic circuits. Most notably, this includes their development of a silver-based conductor ink and significant improvements to their previously-developed semi-conductor ink. These advances will make circuits possible on a wide range of materials, including plastics, film and textiles. The process is scalable so it could be used to create everything from low-cost miniature circuits to large video displays. Because the new silver ink can be delivered at sufficiently low temperatures (140°C), it won’t melt plastics, which was a previous obstacle. Finally, the process uses ink-jet technology and doesn’t require clean room facilities such as those needed for silicon chip manufacturing.

Xerox has already begun aggressively marketing the new technology which I believe will make a significant impact, both in established industries and in entirely new uses. The ability to print on fabrics should make wearable electronics really start to take off. Custom large-scale, flexible video screens and signage will become feasible. Sensors capable of detecting all manner of airborne pathogens and toxins will eventually become widespread.

Given the relatively inexpensive setup costs this will bring about in time, I’d expect a significant open source movement to build around printable electronics. Not only will this yield all sorts of innovations, but it will eventually make many consumer devices more affordable. Think RepRap for electronics.

But to me, by far and away the most exciting aspect of this breakthrough is the promise of low-cost RFID tags and sensors. Once these are in widespread use, every item in a warehouse, every book in a store, every piece of furniture in a home will be able to convey information about its location and state. Your refrigerator will be able to poll its contents, generating a shopping list or identifying items past their use-by-date. The possessions in your house could be easily inventoried and recorded for insurance purposes, in case of theft or fire. Misplaced items could be located in an instant.

In a word, everything in our environment would become more intelligent. And that’s only the beginning. Wireless communication to a central server or internet services could allow a failing device to arrange it’s own repair. (Obviously, with over-ride options by the owner.) Sensors along roads and buildings could assist in everything from collision avoidance to giving directions to tourists. Gradually, additional features and processing would be added to each device, creating an ecosystem of interacting, inter-communicating electronics.

Of course, there will no doubt be considerable privacy and security issues to contend with, not to mention all kinds of new scams involving counterfeit RFID tags, false information, misdirection and so forth. But to be honest, I don’t see us foregoing the functional and marketable benefits because of such concerns. We’ll just have to build methods and measures and legislation to deal with the possible downsides of the technology. Because in the end, smarter really is better.

Science literacy

I’m not the first to say it, but we could certainly stand to have a little more science literacy in this country.  In almost all countries, actually.  It’s not that everyone needs to understand thermodynamics or be able to calculate a Fibonacci sequence.  But as our world and our devices grow in complexity, there are certain basic tools we could all use to make our lives better, safer and more financially secure.

Vaccinations save lives

The recent death of a teenage girl in Coventry, England following her vaccination for human papillomavirus (HPV) is an prime example of this.  Immediately following the girl’s death, the media jumped on the story, reporting it in such a way that it created a considerable and unnecessary scare.  Though the cause of death was in fact an unrelated tumor, it was days before this information was available to the public.  While the young girl’s death was tragic, it would be so much more tragic if hundreds of others lost their lives because they didn’t receive this vaccination.  HPV is recognized as the major cause of cervical cancer.  Worldwide, there are over half a million new cases of cervical cancer annually resulting in over a quarter million deaths each year.  Yet it only takes a small number of serious adverse reactions to sway a large portion of the public because they don’t understand the difference between correlation and causation.

In the US, 24 million doses of HPV vaccine have been given in the last four years with a little under a thousand reports of serious adverse reactions.  While some people may see this as a large number of reactions, this amounts to four thousandths of a percent!  Given any sufficiently large group, it’s inevitable that something will occur to a small subset – they’ll get a cold or be in an auto accident or win the lottery that day.  It’s only correlation.  The two events are not linked.

Our minds are structured to want to spot patterns in the world around us.  But as the world grows more complex, it’s getting harder to simply intuit those patterns correctly.  That’s one reason why people believe in wild conspiracies or that a particular number or color or talisman is lucky for them.  But with a little better grounding in science, people could make much better decisions.  A little introductory physics will convince you that tailgating is a really bad idea, no matter how advanced your braking system.  A quick cost-benefit analysis will show that three seconds of seatbelt-fastening every single car trip beats taking the chance of being one of the twenty thousand that die each year because they didn’t.  Exposure to basic probability will show why those trips to the casino are a lousy investment.  Knowing something about how polls, surveys and studies are performed will show that the methodology used greatly affects their accuracy and usefulness.  (Unless your idea of usefulness is misinformation.)

As technological progress accelerates, it’s only going to become increasingly difficult to navigate through life successfully without these basic tools of science.  Without them, many of us will increase our chances of being exposed to preventable health risks, of being needlessly scammed and taken advantage of, and even of dying an unnecessary death.  And that’s a tragedy.

Page 10 of 12« First...89101112
Like us on Facebook