Reynolds advocates faster nano/AI R&D for safety reasons

In Popular Mechanics, longtime Foresight friend Prof. Glenn Reynolds looks at the future of nanotech and artificial intelligence, among other things looking at safety issues, including one call that potentially dangerous technologies be relinquished.  He takes a counterintuitive stance, which we’ve discussed here at Foresight over the years:

But I wonder if that’s such a good idea. Destructive technologies generally seem to come along sooner than constructive ones—we got war rockets before missile interceptors, and biological warfare before antibiotics. This suggests that there will be a window of vulnerability between the time when we develop technologies that can do dangerous things, and the time when we can protect against those dangers. The slower we move, the longer that window may remain open, leaving more time for the evil, the unscrupulous or the careless to wreak havoc. My conclusion? Faster, please.

OK, it’s counterintuitive, but it may be right.  —Chris Peterson

Nano PVs: cheaper or better?

Over at Nanoclast, Dexter Johnson writes:

It seems when nanotech is applied to photovoltaics it can either boost their efficiency to new heights or it can cheapen their manufacturing process. But it never seems to provide a solution to both of these. It’s always a tradeoff: increased efficiency but difficult manufacturing processes or a cheaper production process but less efficiency.

The solution to this, of course, is that the efforts in nanotech research should be going toward developing atomically-precise machinery that can do the manufacturing.  Like any form of research and capital formation – vs – consumption question, there is a balance between this and direct application-oriented work, but the more spent on the former, the better in the long run. And the use of the word “nanotechnology” to characterize the latter has confused the issue.

nanoarrows

"Nanotechnology" hunting arrows

As I wrote in Nanofuture:

… the stuff that’s going on in most labs today under the name of nanotechnology may make smaller computer chips, or stronger aerospace materials, or whatever, but it’s really more of the same old conventional technology by another name. You don’t need to read a whole new book to learn that people are trying to make more stain-resistant (and expensive) pants, or stronger (and more expensive) tennis racquets, or smaller, faster computers. Nor do you need to worry over the fact that marketing departments will be calling these things, and lots of other things over the coming years, “nanotechnology”–it’s just a word.

… So “nanotechnology” really does have two different meanings. One is the broad, “stretched” version meaning any technology dealing with something less than 100 nanometers in size. The other is the original meaning: designing and building machines in which every atom and chemical bond is specified precisely. I’ll refer to the former as “nanoscale technology” when I need to; but I won’t refer to it much. The capabilities and dangers of nanoscale technology are simple and straightforward extensions of current trends in the capabilities and dangers of chemistry, materials science, and microfabrication. The majority of new techniques being discovered and trumpeted as the latest thing in “nanotechnology” today will be obsolete in ten years.

… Nanotechnology has the potential for increasing our physical capabilities more than did the industrial revolution; expanding our ability to learn and communicate more did than the printing press; accelerating our ability to travel more than did the boat or the wheel; and enlarging the range of places we can live more than clothing did. It could induce greater biological changes in the human organism than the difference between humans and chimpanzees; indeed, greater than the difference between humans and horseshoe crabs. It is coming, possibly in the next decade, probably in the next two-and-a-half, almost certainly in the twenty-first century.

Meanwhile, solar power continues to fall in a Moore’s Law – like fashion; but it won’t really be mature until we get real nanotechnology.

Gallery – A joyride through the nanoscale – Image 1 – New Scientist

Gallery – A joyride through the nanoscale – Image 1 – New Scientist.

This New Scientist article has some nice images from Whitesides recent book, sort of a retake on the “Secret House” idea.

 

Technology Review: Self-Cleaning, Super-Absorbant Solar Cells

Technology Review: Self-Cleaning, Super-Absorbant Solar Cells.

Amorphous-silicon solar cells patterned with nanoscale domes absorb more light–and shed water and dust.

 

Moore’s Law Marches On

According to the loose length-scale based definition, nanotechnology has long since conquered the world: feature sizes in microprocessors have been below the 100 nanometer mark for some time, qualifying them, if anyone wanted to, to be called nanoprocessors. The latest reports and plans are mentioning 22-nanometer parts just 2 years from now:

DailyTech – AMD Desktop Roadmap Features Bulldozer Architecture, New Chipsets.

Next Big Future: AMD has an Ambitous Roadmap for 2010 and 2011

More Merkle at Singularity University

Ted Greenwald continues his Singularity University executive program coverage over at Wired:

These days, though, Merkle is setting his sights much higher. Over the past few years he has put together a theoretical system for building diamond, atom by atom. It involves nine molecular tools and methane/hydrogen feedstock on a diamond substrate. He has analyzed all the side reactions, he says, and shown why they won’t throw the process out of kilter. “This is the first effort to define a minimal tool set for positional diamond mechanosynthesis,” he says. “It’s hard,” he says — an understatement — “but it ought to work.”

It’s great to see so many Foresight members helping teach over at SU.  —Chris Peterson

 

Self-assembly of carbon nanotubes into two-dimensional geometries using DNA origami templates

Self-assembly of carbon nanotubes into two-dimensional geometries using DNA origami templates.

Harnessing DNA origami to arrange CNTs.

Nanotechnology researchers find reliable, mess-free way to grow graphene

Nanotechnology researchers find reliable, mess-free way to grow graphene. from nanowerk

 

single atom-thick graphene transistors

“You can imagine trying to peel a piece of shrink wrap off a dish to put it on a new dish — it’s going to be messy,” said lead researcher Jiwoong Park, Cornell assistant professor of chemistry and chemical biology.
Inspired by previous work in which scientists grew graphene on copper foil, the team grew the graphene directly onto silicon wafers coated with a special evaporated copper film. They then cut the graphene films into their desired shapes using such standard methods as photolithography, and removed the underlying copper with a chemical solution. What was left was a graphene film that draped down over the silicon wafer with little defect.
“Once the graphene is made on top of this wafer, you can apply any thin-film processing technique,” Park said.
The team is now experimenting with growing full-scale, four-inch graphene wafers, which would further demonstrate the manufacturing potential of graphene-based electronics.

One wonders if this could be combined with the recently invented surface-plasmon fluidic logic.

New Digital 'Electronics' Concept May Continue Moore's Law

Merkle on nanotech at Singularity University

Ted Greenwald posted yesterday at Wired about Foresight member Ralph Merkle’s presentation on nanotechnology at the Singularity University’s first Executive Program, which has just convened over at NASA Ames here in Silicon Valley:

From there he skims through a catalog of progress — familiar example of pushing atoms into IBM logos and such on a 2D grid — to the goal of 3D shapes, and ultimately nanoscale machines. It doesn’t always work. “You’re not seeing the failures,” he allows, and describes a planetary gear he built that was just too slippery to hold together. “There’s no friction at that scale.” Moreover, that style of assembly is one atom at a time — very resource-intensive. A better solution is self assembly, along the line of, say, a redwood tree — a huge structure self-assembled by nanomachinery. If we can accomplish that, “manufacturing costs will go through the floor.” Products of nanomachinery will be as cheap as potatoes.

The notion that nanotech will provide new materials with superior strength-to-weight characteristics or other cool properties is familiar. Eye-opening proposals: Respirocytes (carry oxygen in the bloodstream so you can hold your breath for an hour), microbivores (eliminate diseases more rapidly than they body’s own system), chromallocytes (removes chromosomes in a cell and replaces them with a new set). Finally, Merkle sketches out a single-stage-to-orbit vehicle made of specific (theoretical) nanomaterials that apparently has been designed by someone in a published paper, name and title I didn’t catch. Bottom line: It could transport four passengers into space for a few thousand dollars.

Other topics include artificial intelligence, robotics, networking, computing, and quantum computing.  —Chris Peterson

The bad robot takeover

From the Albany (OR) Democrat Herald:

Phone robots: Let’s all rebel

By Hasso Hering, Columnist | Posted: Saturday, November 7, 2009 11:45 pm

What this country needs – even more than a shorter baseball season so the World Series doesn’t go into November – is a popular uprising against the tyranny of telephone robots.

This is how those talking machines drive you up the wall.

You want some information from a company, but there is no local number. So, dreading what comes next, you dial the toll-free number in the book.

After the greeting and a burst of Spanish – which presumably means that if you prefer that language you should push numero uno or something – a machine asks you for your account number.

You don’t have one, of course. And while you’re thinking of what you might say to get to the next step, the machine gets impatient:
“I’m sorry, I didn’t get what you said. In order to proceed with this call, I need your account number.”

You sputter something in response, but it’s not an account number.

The robot comes back wanting to know your phone number. This is something you can provide, and you do, grudgingly, knowing that it really won’t help.

Sure enough, the robot asks: “I don’t recognize this number in our records. Is this the phone number for the account you have with our company?”
No, you dummy, it’s not. It’s my own phone number.

“I don’t have an account,” you say.

Robot: “I’m sorry, I didn’t understand. Is this the telephone number on the account? In order to proceed with your request, I need an account number or the telephone number for the account. If you do not have an account number or do not know it, say: I don’t know it.”

“I don’t know it,” you mumble, obediently.

Robot: “I’m sorry, I did not understand. …

This is, unfortunately, the kind of “robotic” robots that actually are taking over the world.  And the problem is not that they’re too good, or too intelligent, or anything like that.  Indeed, it’s just the opposite: the problem is that they’re incompetent.  If Hering had gotten a polite, friendly, knowledgeable, and helpful agent on the phone, there wouldn’t have been much of a column.

On the other hand, it should be pretty clear to any business that they would be better off with polite, friendly, knowledgeable, and helpful robots.  There’s a strong market pressure and money available for development (to the extent that there’s money available for the development of anything).

A call-center help-desk was one of the possibilities mentioned at the AGI roadmap for an intelligence test.  The idea is that the system would be given a manual and some software (or other system) and a week (or whatever) to read and learn, and then be put on the phone and judged on how well it managed to help people who were having problems with the system.

The state of the art in phone-answering systems isn’t quite as bad as the humorous editorial above makes out, but it’s still not good enough to carry on a reasonable conversation even on the simple, constrained subjects that an automated receptionist should handle. I confidently expect this to change over the coming decade — but it remains a toss-up, in my opinion, whether we’ll have a system that can learn to be a  competent receptionist, as opposed to having been laboriously hand-coded and trained to be one.  And if we do, it’ll most likely have major chunks of general skills coded in — things like speaking and reading, for example.

But to the company that wants a roboreceptionist, it doesn’t matter where the skills came from — the company will decide between learned and coded skills on the basis of cost. So if I had a system that could do the learning, it would be worth as much as the development and training team.  I would want to sell trained systems with skills, not learning systems — that would be like giving away my factory.  (It will be interesting to see what happens when open-source IDEs get good enough to be said to be learning the program rather than being a pile of tools for a programmer.)  And it seems unreasonable to think that at any level of technology, learning a skill would be as cheap as simply doing it once learned.

So it seems very likely that the technology of learning AI will develop, in early days at least, in a form of learning machines that create separate narrow AIs, instead of a more human-like learning paradigm.  And it seems likely that a common origin of these learning systems will be AI development envirionments, which today are intended for very heavy human involvement and should simply become more and more automated over time.  And of course these will be self-improving — the first thing everyone with a development environment does is use it to work on its own code — but again with lots of human input.

Let’s just see if we can’t just get to the point where I, as a software architect, can simply talk and wave my hands to my development system, which does all the low-level design and coding. Competently.

 

 

Robots

There was some objection to my post Is Robo Habilis a gateway to Intelligence? to the effect that it might take a lot of extra time to build the robots, and that would lengthen the time necessary to develop AI. That might certainly be true of the garage experimenter, but in the world at large, the robots are already here. The kind of robots I’m thinking of are bolted-to-the-table torsobots in the tradition of Cog:

Cog picture

The reason is that as of even date, you just can’t put enough processing in a mobile robot to be doing the kind of processing a human is doing with its sensory and motor streams.

There’s also of course Kismet:

kismet

and various other research robots ranging from the simple –

Babybot

babybot

Ludwig

Ludwig

They get more complex, e.g. Berti:

Berti,Berti

to the extremely anthrobiomimetic CRONOS2:

Cronos2

to some serious engineering:

Domo:

Domo

iCub:

iCub

Saika:

Saika

Nasa’s Robonaut:

Robonaut

These use commercial arms and manipulators:

The UMASS Torso robot:

UMASS torso

The Iowa torsobot:

Iowa bot

And finally, these appear to be commercially available:

Hiro:

Hiro

Motoman SDA10:

Motoman

and the Meka (a descendent of Domo in commercial form):

headarm

so … the bodies are there.

Brain mapping and the connectome

I’m at the AAAI Fall Symposium session on Biologically Inspired Cognitive Architectures, and there was a really interesting talk by Walter Schneider of Pitt about progress in mapping the nerve bundles that are the “information superhighways” between the various parts of the brain.  You’ll find his slides from last year’s talk on his home page, and there has apparently been progress amounting to a breakthrough in the interim.

This and fMRI together are giving us an understanding of what’s going on in the brain that’s advancing faster than anybody (with the possible exception of Ray Kurzweil) thought it would.

Schneider claims that the techniques now being worked on could be pushed to a resolution of 20 microns, with appropriate resources, by 2014 or thereabouts. That’s essentially good enough to have a complete wiring diagram of the brain.

Is Robo Habilis a gateway to Intelligence?

In response to my Robo Habilis post, Tim Tyler replied:

An intelligence challenge should not involve building mechanical robot controllers – IMO. That’s a bit of a different problem – and a rather difficult one – because of the long build-test cycle involved in such
projects.
There are plenty of purer tests of intelligence that use more abstract ideas – games, puzzles, and other classical intelligence test fodder.
If you want to measure the abilities of mechanical robots, then fine, but let’s not pretend that it’s the same thing as measuring intelligence.

This is a fairly widely held view — there were a couple of researchers at the AGI Roadmap meeting expressing the same idea.  If I understand him correctly, Minsky feels the same way.  I believe, however, that it is not true.

To begin with, that was the reigning paradigm of the entire “golden age” of AI from the 50s through the 70s. Even Shakey the Robot had a bicameral control architecture: a body control program written in SAIL, and a cognitive engine written in LISP.  It was strongly believed that the parts of thought that were hard for humans would be the hard ones to program, and that once we got those licked, building the lower-level body-controller stuff (or vision, or speech-to-text for the input) would be an afterthought, or at most a clean-up engineering exercise.

Over the course of the 60s, classic AI had a tremendous run of success, which is pretty neatly summed up by the work in Minsky’s “Semantic Information Processing.”  They had programs that did games, puzzles, intelligence tests, arithmetic word problems, freshman calculus. The hard stuff. They were full of optimism, and predicted that AI would run to a successful conclusion, creating an artificial mind, in another decade or two. They had done the college student; how much more effort should it take to do a toddler?

They were wrong.  The greatest lesson that came out of the Golden Age was that “the hard stuff is easy, and the easy stuff is hard.”  Any toddler could recognize a dog in a picture; it would be three more decades before AI could get even close (and it’s still not really there yet).

The mind, it turns out, is like an iceberg — most of it is unseen to consciousness, below the waterline.  Perhaps a better analogy would be that consciousness is like the legislature of a country, or the head office of a company. What they perceive is in reality only an executive summary of what’s really happening. What the early AI researchers had done was to build a “company” consisting only of the board of directors and secretaries, but no factories, no sales force, no middle managers, no shop foremen, and no labor force.

The brain was evolved as a body controller. Evolution typically takes a structure that works and copies and adapts it to the next task. Consider the increasing intelligence of animals as we work ourselves up the evolutionary tree towards the human: insects, reptiles, mammals, primates.  At every level new and improved kinds of control, feedback, discrimination, planning, and learning are built into the structure — and it’s all still there forming the part below the iceberg, the real company outside the boardroom, of human intelligence.

The classic AIers at the Roadmap asked me, “Isn’t a blind paraplegic still intelligent?” and of course he is — but only because his brain still contains all the mechanism that was evolved to to the control and interpretation he now lacks.

The buzzword in current AI for the reason bodies are important is “symbol grounding.”  This refers to philosophical theories of meaning among symbols in symbol-processing machinery, and a simplistic reading of it is that whereas SHRDLU doesn’t “really know” what a red block is, a physical robot that plays with them really does.  Unfortunately, the term in common use is often taken as implying that there is some magical transubstantiation of meaning into symbols by virtue of having a physical body, and this isn’t right and obscures the real issue. The paraplegic still has meaning in his mind.

What has to be there is not the actual body, but the mental mechanism for controlling it — that allows the mind to imagine, predict, describe, and relate other concepts to the one said to be understood. Most of our higher-level concepts are drawn from, by analogy and blending, the basic (very large) set of concepts we have learned, by experience, on the shop floors of our minds as we interact with the real world over the course of our lives.

Could that interpretive, predictive, concept-building, etc, cognitive machinery be built another way than working up a controller for a humanoid robot body? Certainly. But there are two reasons to do it with a body: first, it’s most likely easiest that way.  There are a lot of things we don’t know yet about how the mind works.  There’s no reason to think that we have no more blind spots like the classic AIers did. Working with real robots will show us the gaps fastest.

The second reason is that once we get the brain built, if we’ve put it together in a rough semblance of the phylogenetic/ontogenetic sequence that the human mind is built, there’ll be a much better chance that its meanings will match ours. It will understand things the way we do (of course humans vary a lot in the way we understand things), and do things the way we do, and thus appreciate the way we do them, and vice versa.  For example, the parts of the brain that control language and manual manipulation are strongly overlapped. Try to teach your robot sign language without a similar structure and it will never get the “accent” right.  Nor, unless it has the same kind of manipulation control to borrow, will it ever be as fluent in English as a human.

Separating “intelligence” from the rest of cognitive function is a false dichotomy, and one that has led AI astray — in a big way — before.

Nanotechnology devices: Molecular machines shift into gear

Nanotechnology devices: Molecular machines shift into gear.

An atomically precise gear, rotated by pushing the teeth one at a time with a STM tip.

More on the AI takeover

There are at least 4 stages of intelligence levels that AI will have to get through to get to the take-over-the-world level. In Beyond AI I refered to them as hypohuman, diahuman, epihuman, and hyperhuman; but just for fun let’s use fake species names:

  • Robo insectis: rote, mechanical gadgets (or thinkers) with hand-coded skills, such as Roomba or industrial robots or automated call-center systems or dictation programs.
  • Robo habilis: Rosie the housemaid robot level intelligence, able to handle service level jobs in the real world but not a rocket scientist.
  • Robo sapiens: up to and including rocket scientists, AI researchers, corporate executives, any human capability.
  • Robo googolis: a collection of top R. sapiens wired together in a box running at accelerated speed, equivalent to, say, Google (the company and the search engine together).

First point: One R. googolis can’t take over the world, any more than Google could. You’d have to get to the next stage (R. unclesammus).  Any AI in the earlier stages of development that acted antisocial gets stomped on fast (and in early days, they’ll have no rights — so they’ll basically be exterminated).

Second point: As Robin Hanson and many economists point out, the complementary effect of machines up through the R. insectis stage has generally been much stronger than the substitution effect, so that improving technology has had a general beneficial effect on incomes even though it put specific people, buggy-whip makers for example, out of work. Complementarity is seen when comparative advantage holds, substitution when it doesn’t:

So far, machines have displaced relatively few human workers, and when they have done so, they have in most cases greatly raised the incomes of other workers. That is, the complementary effect has outweighed the substitution effect–but this trend need not continue.
In our graph of machines and humans, imagine that the ocean of machine tasks reached a wide plateau. This would happen if, for instance, machines were almost capable enough to take on a vast array of human jobs. For example, it might occur if machines were on the very cusp of human-level cognition. In this situation, a small additional rise in sea level would flood that plateau and push the shoreline so far inland that a huge number of important tasks formerly in the human realm were now achievable with machines. We’d expect such a wide plateau if the cheapest smart machines were whole-brain emulations whose relative abilities on most tasks should be close to those of human beings.

I don’t think that the “plateau” is really flat, though. There are two reasons. The first is that human capability is a range, with R. habilis at one end and R. sapiens at the other. It’ll take some time to get through — at least a decade, maybe two.

The other reason is that the comparative advantage we saw in the Industrial Revolution may just get turned on its head.  Right now we have a Moore’s Law for the robot’s brain but not for its body.  In other words, we may enter a strange period where white-collar workers are replaced by beige boxes but blue-collar ones are still cheaper — for a little while — than a fully-capable humanoid robot body.  (That will disappear soon enough after nanotech manufacturing takes hold, but at the moment, it looks like AI may be a decade earlier than real nanotech.)

The key thing to remember when thinking about the economic AI takeover is that it is not something we should be trying to prevent. Why shouldn’t we, the human race as a whole, build machines to do the hard work we need done, and spend our time enjoying the resulting wealth?  Why shouldn’t we spend our efforts deciding what needs to be done, and let the machines do it?

Questions like unemployment are the result of taking a system that is well-adapted for one economic situation and applying it to a totally different one. What should the economic system look like when robots do all the work? And once we get that figured out, how do we get there from here?

Rice scientists point out that nanotubes are polymers


(http://www.youtube.com/watch?v=PSxihhBzCjk)

From NanoWerk: Rice scientists argue nanotubes can be treated like polymers

Wade Adams, Matteo Pasquali, Micah Green and Natnael Behabtu at Rice pick up that thread in their discussion of what we know — or think we know — about carbon nanotubes.
Their review in the journal Polymer (“Nanotubes as polymers”) makes the argument that single-walled carbon nanotubes (SWNTs) are polymers and should be treated as such.
The point is to remind the nano community that decades of research into polymers can be applied to their work and hasten the development of novel materials for all kinds of uses.
“In one of his earliest lectures about nanotubes, (late Rice professor and Nobel laureate) Rick Smalley said they’re the ultimate polymer molecule, with every atom in its place, just like a polymer chain would have,” said Adams, director of the Richard E. Smalley Institute for Nanoscale Science and Technology, who focused on polymers for many years at the Air Force Research Laboratory. “I really didn’t believe him initially.”

Adams said the goal is to change the mindset of a generation of scientists who have come to think of carbon nanotubes as special when, in a very important way, they’re not special at all.
“We were seeing a lot of literature out there about nanocomposites that were totally ignorant of the 15-, 20- and 30-year-old literature that explored a lot of these areas and had already clarified some of the things you need to think about if you’re going to use these materials,” he said.

The article coincides with an announcement of a new development in nanotube processing that does, in fact, treat nanotubes as polymers and thus allows for considerably greater industrial use:

Rice University scientists today unveiled a method for the industrial-scale processing of pure carbon-nanotube fibers that could lead to revolutionary advances in materials science, power distribution and nanoelectronics. The result of a nine-year program, the method builds upon tried-and-true processes that chemical firms have used for decades to produce plastics. The research is available online in the journal Nature Nanotechnology.
“Plastics is a $300 billion U.S. industry because of the massive throughput that’s possible with fluid processing,” said Rice’s Matteo Pasquali, a paper co-author and professor in chemical and biomolecular engineering and in chemistry. “The reason grocery stores use plastic bags instead of paper and the reason polyester shirts are cheaper than cotton is that polymers can be melted or dissolved and processed as fluids by the train-car load. Processing nanotubes as fluids opens up all of the fluid-processing technology that has been developed for polymers.”

This is something of a halfway-point to true industrial-scale nanotube use, though, since nanotubes still can’t be made with purity of the types that have the kinds of properties (e.g. conductivity) one would like:

But a final breakthrough remains before the true potential of high-quality carbon nanotubes can be realized. That’s because HiPco and all other methods of making high-end, “single-walled” nanotubes generate a hodgepodge of nanotubes with different diameters, lengths and molecular structures. Scientists worldwide are scrambling to find a process that will generate just one kind of nanotube in bulk, like the best-conducting metallic varieties, for instance.
“One good thing about the process that we have right now is that if anybody could give us one gram of pure metallic nanotubes, we could give them one gram of fiber within a few days,” Pasquali said.

Do we need Friendly AI?

My Robo Habilis post was picked up on by Michael Anissimov who wrote:

(me:) It seems to me that one obvious way to ameliorate the impact of the AI/robotics revolution in the economic world, then, is simple: build robots whose cognitive architectures are enough different from humans that their relative skillfullness at various tasks will differ from ours. Then, even after they are actually better at everything than we are, the law of comparative advantage will still hold.

Boom, friendliness problem solved. Build robots with different cognitive architectures than us, and they will be forced to keep us around, due to Ricardo’s law of comparative advantage. Sounds wildly naive to me.

All I can say is thanks for noticing I’ve solved the most important problem of the 21st century with a single paragraph! I’m confidently expecting my Nobel Peace Prize.

But seriously, I would like to argue that the concept of the “friendliness problem” is a dangerous misreading of the real difficulties and problems we will face as a result of the development of artificial intellegence over the next few decades. It seems to me that one could characterize the people working on “Friendly AI” as essentially trying to redo moral philosophy, from scratch, and get it right this time.  There’s nothing wrong with this; moral philosophy is a valuable intellectual tradition and worthwhile human activity.  But the notion that the whole business, with the addition of the new insight that there can be intelligent machines as well as humans among the class of moral agents, could be solved in any useful sense, just strikes me as silly. Indeed, the new insight makes moral philosophy a lot harder, rather than bringing it any closer to any kind of closure.

Instead let’s look at the kind of problems we’re really going to face. There is not — I guarantee it — going to be any single overarching solution to them; there will be a host of minor things we can do to ameliorate the problems as they arise, and we’ll just have to keep coming up with them as problems arise.

We know what it will be like should we manage to invent and implement a giant, powerful decision-making system that takes over the world. We know because we’ve already done it. Some people have observed this system in action and seem to think that it has a “friendliness problem”:

We’re Governed by Callous Children

When I see those in government, both locally and in Washington, spend and tax and come up each day with new ways to spend and tax—health care, cap and trade, etc.—I think: Why aren’t they worried about the impact of what they’re doing? Why do they think America is so strong it can take endless abuse?
I think I know part of the answer. It is that they’ve never seen things go dark. They came of age during the great abundance, circa 1980-2008 (or 1950-2008, take your pick), and they don’t have the habit of worry. They talk about their “concerns”—they’re big on that word. But they’re not really concerned. They think America is the goose that lays the golden egg. Why not? She laid it in their laps. She laid it in grandpa’s lap.
They don’t feel anxious, because they never had anything to be anxious about.

Peggy Noonan thinks the government is screwing us up because it’s made of people who don’t care. But I beg to differ.  There’s a classic fallacy in the philosophy of mind that shows up in places ranging from Leibniz’ story of the “magnified mill” to Searle’s Chinese Room, which is that for a system to have some property, the property must be present among the parts. This is just as false for caring as it is for understanding or consciousness. In fact the existing system is a perfect example, although in reverse — it’s composed of people who do care, but they interact in a structure that results in an evil bureaucracy.

Instead, what’s happened is that we made a blunder in designing the system that is exactly equivalent to a favorite example of Eliezer Yudkowsky: instead of building a paperclip-maximizing machine, we built a vote-maximizing machine.

I claim that the problem is much more productively looked at from another point of view: the system as a whole is incompetent. It doesn’t do what it was built to do (“… promote the general welfare, secure the blessings of liberty …”). The designers simply assumed a vote-maximizer would do the things they wanted, but they were wrong.  Similarly, no human wants the universe to be converted into paperclips, so if he built a machine with that goal, he would have designed incompetently.  I claim we should be spending our time on is figuring out how to build competent AI.

First principle of competent AI design: Build a machine that understands what you want.  The paperclip maximizer is a study in amazing contrasts — presumably an intelligence powerful enough to take over the world would be capable of understanding human motivations even better than we do, so as to manipulate us effectively. Yet it’s built with a complete cognitive deficit of appropriate motivations, goals, and values for itself. Incompetent.

Second principle: build machines that know their limitations.  This basically means that it should confine its activities to those areas where it does understand the effects of its actions.

But in order to do that, we first have to be able to build a machine that can actually understand something — anything — in the full human-level meaning of understanding. And that is the necessary first step to a future of useful and beneficial AI, and it’s what anyone concerned about such things should be working on.

Robo Habilis

One of the species of early hominids is named Homo habilis, meaning “handy man,” after their significant advancement in tool use over previous hominids. One of the goals of the AGI Roadmap is to chart paths to full human intelligence, and one of the paths might follow the one that evolution took. The Wozniak Test, i.e. being able to make coffee in any randomly-chosen home, is a case of tool use competence. It is a special case of what we might call the Nilsson Test, as outlined in a paper in 2005 by Nils Nilsson, one of the leading figures in AI:

Machines exhibiting true human-level intelligence should be able to do many of the things humans are able to do. Among these activities are the tasks or “jobs” at which people are employed. I suggest we replace the Turing test by something I will call the “employment test.” To pass the employment test, AI programs must be able to perform the jobs ordinarily performed by humans. Progress toward human-level AI could then be measured by the fraction of these jobs that can be acceptably performed by machines.
Let me be explicit about the kinds of jobs I have in mind. Consider, for example, a list of job classifications from “America’s Job Bank.” A
sample of some of them is given in figure 1:

Meeting and Convention Planner
Maid and Housekeeping Cleaner
Receptionist
Financial Examiner
Computer Programmer
Roofer’s Helper
Library Assistant
Procurement and Sales Engineer
Farm, Greenhouse, Nursery Worker
Dishwasher
Home Health Aide
Small Engine Repairer
Paralegal
Lodging Manager
Proofreader
Tour Guide and Escort
Geographer
Engine and Other Machine Assembler
Security Guard
Retail Salesperson
Marriage and Family Counselor
Hand Packer and Packager

Just as objections have been raised to the Turing test, I can anticipate objections to this new, perhaps more stringent, test. Some of my AI colleagues, even those who strive for human-level AI, might say “the employment test is far too difficult—we’ll never be able to automate all of
those jobs!” To them, I can only reply “Just what do you think human-level AI means? After all, humans do all of those things.”

Now some of those jobs require specialized training and years of experience, while some of them are entry-level, accessible immediately to the average human. Most are somewhere in between. Note that “Maid and housekeeping cleaner” is in itself a superset of the Wozniak Test.

The ability of an AGI (= human-level AI) to do most or all of the jobs humans do is cause for a certain amount of concern. This brings us to a recent post by Robin Hanson:

Yes, techies agree on the long term plausibility of machines doing almost all jobs at a cost below human subsistence wages, thereby gaining almost all income, while economists ignore this scenario. …

Economists should listen more to techies on what techs will be feasible at what costs, but techies should also listen more to economists on the social implications of tech costs.  Alas, just as economists prefer to rely on their intuitive folk tech forecasts, techies prefer to rely instead on their intuitive folk economics. …

The standard views of techies about what techs will be feasible might be wrong, and the standard views of economists of how to forecast tech consequences might be wrong.  And it is fine for contrarians to try to persuade specialists they are in error, though contrarians would be wise to at least understand the standard view before trying to overturn it.  But surely what the world needs first and foremost is to see and take seriously the simple combination of the standard views on such important topics.

One of the standard economic laws that applies in this case is Ricardo’s Law of Comparative Advantage. It states basically that it is generally to the advantage of parties of differing productivities to trade. In particular, the counter-intuitive part, it is to the advantage of the more productive party (e.g. the machines) to trade with the less productive (us, in the robot economy scenario).  The exception is where the abilities (productivities across goods) are in the same exact proportions, leaving the parties nothing to specialize in.

It seems to me that one obvious way to ameliorate the impact of the AI/robotics revolution in the economic world, then, is simple: build robots whose cognitive architectures are enough different from humans that their relative skillfullness at various tasks will differ from ours. Then, even after they are actually better at everything than we are, the law of comparative advantage will still hold.

AGI Roadmap meeting

Foresight’s mission is essentially an educational one.  In simplest terms we are here to point out foreseeable technological developments that not only will make the future different from the past, but make it different in ways that aren’t obvious and which everyone isn’t already planning for. Nanotechnology — true nanotech in Drexler’s original sense of having a thorough control over the structure of matter at the atomic scale and thus being able to build productive machinery — is such a development, even though the word “nanotechnology” is widely used for much more mundane, predictable, linear, and non-revolutionary progress.

Similarly, the term “Artificial Intelligence” is widely used for predictable, linear progress in software engineering. The field has come a long way, so that it is getting close to the point that any well-specified human skill, such as driving a car, can be implemented given an appropriate application of talent and resources. Just like “nanotechnology,” though, it originally meant something more revolutionary:

Some years ago, Ben Goertzel coined the term “AGI” — artificial general intelligence — to distinguish the original, revolutionary goal of AI as originally seen by such pioneers as McCarthy and Minsky, from the more mundane, incremental work that the term AI had come to cover. This was very similar in spirit to the term MNT — molecular nanotechnology — coined by Drexler and Foresight for essentially the same reason.

Within the past couple of years, the Productive Nanosystems Roadmap was organized and published, under the names of a wide sampling of people from academia, industry, and the national laboratories.  This had the effect of making it clear that the ultimate goal of nanotechnology research is indeed “MNT”-style capabilities, and is one that is ultimately feasible and worth working toward.

While the “diaspora” in AI may have been deeper than the one in nanotech, it was also longer ago — there was no need for the AGI Roadmap to re-establish the possibility of an artificial intelligence in the full sense, but to try and make some sense of the state of the art with respect to it, figure out some milestones and metrics that might be used to judge progress, and so forth.

The meeting last weekend at the University of Tennessee, organized by Ben Goertzel and Itamar Arel, served to bootstrap the process and begin to work out what kind of roadmap might be possible.  The main problem, of course, is that we don’t really know how intelligence works, which pieces are essential and which ancillary, or indeed whether there are a few powerful underlying principles or a huge kludge of random techniques.

To that end we began by trying to define the kind of tasks that we felt a general intelligence could do but that no hand-coded “narrow AI” could do. The classic such task, or course, is the Turing Test, which has many points in its favor but is also considered (a) too high a bar, and (b) a test of the wrong thing, since it requires fooling a judge as well as exhibiting basic intelligence.

To give some of the flavor of the scenarios, here’s the one I proposed:

The Wozniak Test

In an interview a few years ago, Steve Wozniak of Apple fame opined that there would never be a robot that could walk into an unfamiliar house and make a cup of coffee. I feel that the task is demanding enough to stand as a pons asinorum for embodied AGI.

A robot is placed at the door of a typical house or apartment. It must find a doorbell or knocker, or simply knock on the door. When the door is answered, it must explain itself to the householder and enter once it has been invited in. (We will assume that the householder has agreed to allow the test in her house, but is otherwise completely unconnected with the team doing the experiment, and indeed has no special knowledge of AI or robotics at all.) The robot must enter the house, find the kitchen, locate coffee-making supplies and equipment, make coffee to the householder’s taste, and serve it in some other room. It is allowed, indeed required by some of the specifics, for the robot to ask questions of the householder, but it may not be physically assisted in any way.

The state of the robotics art falls short of this capability in a number of ways. The robot will need to use vision to navigate, identify objects, possibly identify gestures (“the coffee’s in that cabinet over there”), and to coordinate complex manipulations. Manipulation and physical modelling in a tight feedback learning loop may be necessary, for example, to pour coffee from an unfamiliar pot into an unfamiliar cup. Speech recognition and natural language understanding and generation will be necessary. Planning must be done at a host of levels ranging from manipulator paths to coffee-brewing sequences.

But the major advance for a coffee-making robot is that all of these capabilities must be coordinated and used appropriately and coherently in aid of the overall goal. The usual set-up, task definition, and so forth are gone from standard narrow AI formulations of problems in all these areas; the robot has to find the problems as well as to solve them. That makes coffee-making a strenuous test of a system’s adaptiveness and ability to deploy common sense.

I claim that this test addresses the bulk of the aspects of general intelligence that are missing from AI today. Although standard shortcuts might be used, such as having a database of every manufactured coffeemaker built in, it would be prohibitive to have the actual manipulation sequences for each one pre-programmed, especially given the variability in workspace geometry, dispensers and containers of coffee grounds, and so forth. Transfer learning, generalization, reasoning by analogy, and in particular learning from example and practice are almost certain to be necessary for the system to be practical.

Coffee-making is a good test of generality because, although it would be possible to hand-code most of the skills needed, it would be much cheaper simply to build a coffeemaker into the robot! Thus the only economical way to approach the task is to build general learning skills and have a robot that is capable of learning not only to make coffee but any similar domestic chore.

Coffee-making is a task that most 10-year-old humans can do reliably with a modicum of experience. I would guess that a week’s worth of being shown and practicing coffeemaking in a variety of homes with a variety of methods would provide the grounding for enough generality that a 10-year-old could make coffee in the vast majority of homes in a Wozniak test.

IEEE Spectrum: German Environmental Agency Miffed at Exploitation of Position Paper on Nanotechnology

IEEE Spectrum: German Environmental Agency Miffed at Exploitation of Position Paper on Nanotechnology.

From Dexter Johnson at nanoclast:

Germany’s Federal Environment Agency (UBA) last week made a background paper available on their website, which they now concede contained no new research and none that their organization had actually performed, entitled “Nanotechnology for Humans and the Environment: Increasing Chances, Minimizing Risks,” that got the German and international press to generate frightening headlines like “Germany warns over dangers of nanotechnology”.

This wasn’t the reaction they were expecting so the the UBA authorities wanted to make clear in a new article that they don’t think nanotechnology is all bad. …