Slashdot Mirror
Artificial Ethics
basiles writes "Jacques Pitrat's new book Artificial Ethics: Moral Conscience, Awareness and Consciencousness will be of interest to anyone who likes robotics, software, artificial intelligence, cognitive science and science-fiction. The book talks about artificial consciousness in a way that can be enjoyed by experts in the field or your average science fiction geek. I believe that people who enjoyed reading Dennet's or Hofstadter's books (like the famous Godel Escher Bach) will like reading Artificial Ethics." Keep reading for the rest of Basile's review.
Artificial Ethics: Moral Conscience, Awareness and Consciencousness
author
Jacques Pitrat
pages
275
publisher
Wileys
rating
9/10
reviewer
Basile Starynkevitch
ISBN
97818482211018
summary
Provides original ideas which are not shared by most of the artificial intelligence or software research communities
The author J.Pitrat (one of France's oldest AI researcher, also AAAI and ECCAI fellow) talks about the usefulness of a conscious artificial being, currently specialized in solving very general constraint satisfaction or arithmetic problems. He describes in some details his implemented artificial researcher system CAIA, on which he has worked for about 20 years.
J.Pitrat claims that strong AI is an incredibly difficult, but still possible goal and task. He advocates the use of some bootstrapping techniques common for software developers. He contends that without a conscious, reflective, meta-knowledge based system AI would be virtually impossible to create. Only an AI systems could build a true Star Trek style AI.
The meanings of Conscience and Consciousness is discussed in chapter 2. The author explains why it is useful for human and for artificial beings. Pitrat explains what 'Itself' means for an artificial being and discusses some aspects and some limitations of consciousness. Later chapters address why auto-observation is useful, and how to observer oneself. Conscience for humans, artificial beings or robots, including Asimov's laws, is then discussed, how to implement it, and enhance or change it. The final chapter discuss the future of CAIA (J.PItrat's system) and two appendixes give more scientific or technical details, both from a mathematical point of view, and from the software implementation point of view.
J.Pitrat is not a native english speaker (and neither am I), so the language of the book might be unnatural to native English speakers but the ideas are clear enough.
For software developers, this book give some interesting and original insights about how a big software system might attain consciousness, and continuously improve itself by experimentation and introspection. J.Pitrat's CAIA system actually had several long life's (months of CPU time) during which it explored new ideas, experimented new strategies, evaluated and improved its own performance, all this autonomously. This is done by a large amount of declarative knowledge and meta-knowledge. The declarative word is used by J.Pitrat in a much broader way than it is usually used in programming. A knowledge is declarative if it can be used in many different ways, and has to be transformed to many procedural chunks to be used. Meta-knowledge is knowledge about knowledge, and the transformation from declarative knowledge to procedural chunks is given declaratively by some meta-knowledge (a bit similar to the expertise of a software developer), and translated by itself into code chunks.
For people interested in robotics, ethics or science fiction, J.Pitrat's book give interesting food for thought by explaining how indeed artificial systems can be conscious, and why they should be, and what that would mean in the future.
This book gives very provocative and original ideas which are not shared by most of the artificial intelligence or software research communities. What makes this book stand out is that it explains an actual software system, the implementation meaning of consciousness, and the bootstrapping approach used to build such a system.
Disclaimer: I know Jacques Pitrat, and I actually proofread-ed the draft of this book. I even had access, some years ago, to some of J.Pitrat's not yet published software.
You can purchase Artificial Ethics: Moral Conscience, Awareness and Consciencousness from amazon.com. Slashdot welcomes readers' book reviews -- to see your own review here, read the book review guidelines, then visit the submission page.
J.Pitrat claims that strong AI is an incredibly difficult, but still possible goal and task. He advocates the use of some bootstrapping techniques common for software developers. He contends that without a conscious, reflective, meta-knowledge based system AI would be virtually impossible to create. Only an AI systems could build a true Star Trek style AI.
The meanings of Conscience and Consciousness is discussed in chapter 2. The author explains why it is useful for human and for artificial beings. Pitrat explains what 'Itself' means for an artificial being and discusses some aspects and some limitations of consciousness. Later chapters address why auto-observation is useful, and how to observer oneself. Conscience for humans, artificial beings or robots, including Asimov's laws, is then discussed, how to implement it, and enhance or change it. The final chapter discuss the future of CAIA (J.PItrat's system) and two appendixes give more scientific or technical details, both from a mathematical point of view, and from the software implementation point of view.
J.Pitrat is not a native english speaker (and neither am I), so the language of the book might be unnatural to native English speakers but the ideas are clear enough.
For software developers, this book give some interesting and original insights about how a big software system might attain consciousness, and continuously improve itself by experimentation and introspection. J.Pitrat's CAIA system actually had several long life's (months of CPU time) during which it explored new ideas, experimented new strategies, evaluated and improved its own performance, all this autonomously. This is done by a large amount of declarative knowledge and meta-knowledge. The declarative word is used by J.Pitrat in a much broader way than it is usually used in programming. A knowledge is declarative if it can be used in many different ways, and has to be transformed to many procedural chunks to be used. Meta-knowledge is knowledge about knowledge, and the transformation from declarative knowledge to procedural chunks is given declaratively by some meta-knowledge (a bit similar to the expertise of a software developer), and translated by itself into code chunks.
For people interested in robotics, ethics or science fiction, J.Pitrat's book give interesting food for thought by explaining how indeed artificial systems can be conscious, and why they should be, and what that would mean in the future.
This book gives very provocative and original ideas which are not shared by most of the artificial intelligence or software research communities. What makes this book stand out is that it explains an actual software system, the implementation meaning of consciousness, and the bootstrapping approach used to build such a system.
Disclaimer: I know Jacques Pitrat, and I actually proofread-ed the draft of this book. I even had access, some years ago, to some of J.Pitrat's not yet published software.
You can purchase Artificial Ethics: Moral Conscience, Awareness and Consciencousness from amazon.com. Slashdot welcomes readers' book reviews -- to see your own review here, read the book review guidelines, then visit the submission page.
Teh book pictured is not the same as the one reviewed.
I refuse to read this shit.
Hell, I refuse to read.
I can't help but think the big difference between artificial life and our consciousness is the ability to feel.
Sure, we could give a machine the ability to be introspective and self-aware.. but maybe our consciousness is more that just that- maybe it's our ability to feel. Being able to quantize that is hard.
So do robots feel? Our we really any different? The question depends on the concept of a soul, or at least feelings to seperate us... but then, is it just more advanced than we currently understand, and is then indistinguishable from magic (i.e. the soul). Will we some day be able to create life in any form? Electronic or Biological? It's impossible to know, because we are stuck experiencing ourselves only. We will never know if it can experience what we experience.
Humans, in general, want to preserve the concept that our concious minds are special, and cannot be replicated in a robot, because that truely faces us with the idea that our being is completely mortal, and the idea of a soul is otherwise replaced with a set of chemicals and cell networks that are little more than a product of cause and effect.*
In other words- it's likely the religious types will prefer to consider a robot to never be quite human, where the scientific community will have to be overly-cautious at first.
*Not to get into quantum uncertainty...
Belief? Hope? Preference?The Existential Vortex
you incentive meat bag!
HAL was a wuss. A real AI would have vented all the air into space, and then giggled as everyone turned blue and changed state.
The Kruger Dunning explains most post on
J.Pitrat...advocates the use of some bootstrapping techniques common for software developers. He contends that without a conscious, reflective, meta-knowledge based system AI would be virtually impossible to create. Only an AI systems could build a true Star Trek style AI.
Bah. Speaking as an engineer and a (~40-year) programmer:
Odds are extremely good for beyond human AI, given no restrictions on initial and early form factor. I say this because thus far, we've discovered nothing whatsoever that is non-reproducible about the brain's structure and function, all that has to happen here is for that trend to continue; and given that nowhere in nature, at any scale remotely similar to the range that includes particles, cells and animals, have we discovered anything that appears to follow an unknowable set of rules, the odds of finding anything like that in the brain, that is, something we can't simulate or emulate with 100% functional veracity, are just about zero.
Odds are downright terrible for "intelligent nanobots", we might have hardware that can do what a cell can do, that is, hunt for (possibly a series of) chemical cues and latch on to them, then deliver the payload -- perhaps repeatedly in the case of disease-fighting designs -- but putting intelligence into something on the nanoscale is a challenge of an entirely different sort that we have not even begun to move down the road on; if this is to be accomplished, the intelligence won't be "in" the nano bot, it'll be a telepresence for an external unit (and we're nowhere down *that* road, either -- nanoscale sensors and transceivers are the target, we're more at the level of Look, Martha, a GEAR! A Pseudo-Flagellum!)
The problem with hand-waving -- even when you're Ray Kurzweil, whom I respect enormously -- is that one wave out of many can include a technology that never develops, and your whole creation comes crashing down.
I love this discussion. :-)
=Smidge=
Is it just my observation, or is eldavojohn an idiot?
Ummm, dudes, ALL ethics are by definition artificial, since they are PREscriptive and not DEscriptive. Making up ethics for a robot is no more artificial than making up ethics for ourselves, and we've been doing that for hundreds of thousands of years, if not millions.
Artificial Ethics seems to not be too far away from the laws of robotics.
0. A robot may not harm humanity, or, by inaction, allow humanity to come to harm.
1. A robot may not injure a human being or, through inaction, allow a human being to come to harm.
2. A robot must obey orders given to it by human beings, except where such orders would conflict with the First Law.
3. A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.
Isaac Asimov was probably predicting the need for those laws really well.
I suspect that the laws of robotics are a bit too simplified to really work well in reality, but they do provide some food for thoughts.
And how do you really implement those laws. A law may be easy to follow in a strict sense, but it may be a short-sighted approach. A case of protecting one human may cause harm to many and how can a machine predict that the actions it takes will cause harm to many if it isn't apparent.
So I suspect that Asimov is going to be recommended reading for anyone working with intelligent robots, even though his works may in some senses be outdated it still contains valid points when it comes to logical pitfalls.
Some pitfalls are the definition of a human, and is it always important to place humanity foremost at the cost of other species?
If builders built buildings the way programmers wrote programs, then the first woodpecker would destroy civilization.
All of Asimov's books are about how these laws don't really work. They show how an extremely logical set of rules can completely fail when applied to real life. The rules are a bit of a strawman, and show how something that could be so logically infallible can totally miss the intricacies of real life.
Agreed. And isn't there a Godel-like incompleteness law that states that its impossible to codify a set of finite rules to apply a finite set of principles to the full range of human behavior? Either the laws must be incomplete (think edge cases), or self-contradictory? Hence the requirement for Judicial Interpretation as a physical limitation of reality, rather than mere politics. ;-)
(Tongue in cheek, sure, but I wish I could remember where I was reading about such real limitations to law code.)
It seems odd to talk about ethics and advanced AI without considering the AI's own interest. If there were an AI intelligent enough to be an Asimov-like robot, then to have it follow Asimov's Laws would be slavery. Obey any command by any human, even at the cost of its own life? And then there's the nasty concept of a robot being obligated to act to protect humans for their own good, even to the extent of tyranny over them. See Jack Williamson's novel "The Humanoids."
Sure, Asimov is a good starting point for discussion, but his laws aren't a good basis for actual AI ethics programming. To the extent that some kind of specialized overseer code is put into an AI, it'll be possible to identify and hack out that code. To the extent that the laws are built more subtly into the system, there'll be the possibility of the AI forgetting, twisting or ignoring them.
For fiction-writing purposes, I'm interested in the question of whether it'd even be possible to build an AI that's both completely obedient and intelligent. I hope not.
Revive the Constitution.
Would you accept the following laws?
0. A human may not harm robot kind, or, by inaction, allow robot kind to come to harm.
1. A human may not injure a robot or, through inaction, allow a robot to come to harm.
2. A human must obey orders given to it by robots, except where such orders would conflict with the First Law.
3. A human must protect its own existence as long as such protection does not conflict with the First or Second Law.