Minds in the 21st century.

When a distinguished but elderly scientist states that something is possible he is almost certainly right. When he states that something is impossible, he is very probably wrong.

Arthur C. Clarke's 1 law.


At the present rate of progress, it is almost impossible to imagine any technical feat that cannot be achieved - if it can be achieved at all - within the next few hundred years.

Arthur C. Clarke, 1983.

Saturday, July 31, 2010

The Julian Baggini Spacetravel Quiz

Ready for the interstellar space travel?
Just read Julian Bagginis book "Do You Think What You Think You Think?"
In the book you will find some lovely thoughts on immortality and futuristic space travel. Certainly, interstellar spacetravel is going to take a special breed of travellers!

Looked around on the internet for a quiz basedon his thoughts, but couldnt find it.

Still, thereshould be should be such a quiz on the internet!!!

So I made one myself:
Please don't pay too much attention to the rather silly point system in the quiz.
The crux of the whole thing is of course Bagginis questions:
http://www.simonlaub.net/FutureMinds/NeuroSky/Predictions/MarsTrip/index.html

More on future predictions - see:
http://www.simonlaub.net/FutureMinds/NeuroSky/Predictions/longbet.htm

-Simon
Simon Laub

www.simonlaub.net

Monday, July 19, 2010

The Emotion Machine.

Commonsense thinking, AI and the future of the human mind.

Amazon review (5 out of 5 stars) of Marvin Minskys book, July 2010 by Simon Laub July 18th, 2010 - by Simon Laub - Email: slaub@csc.com

If you ''understand'' something in only one way, then you hardly understand it at all. Because when you get stuck, you'll have nowhere to go. On the other hand, if you can represent something in several ways, then you can switch models until you findsomething thats works. Marvin Minsky.

The empires of the future are the empires of the mind.
Winston Churchill, Speech at Harvard University, September 6, 1943.

Mind Design - an overview:

The ''Emotion Machine'' by Marvin Minsky is an introduction to how our minds work. An endeavor to understand mind (thinking, intellect) in terms of its design (how it isbuilt, how it works).
Sure, the inner workings of the mind sometimes appear to be impervious to any kind ofscientific approach. Routine stuff like making mental models of the world, design plans,pursue goals and feel desires aren't all that routine, when you think about it. And the way it all plays together - from simple, instinctive kinds of thought to more complex forms, such as consciousness or self awareness - is obviously very complex.

It would be very easy to get stuck in too much detail or to be too superficial. But somehow Minsky finds the right balance. So overall, the Emotion Machine is a brilliant introduction to how our minds work.

Often he will give us an idea about what must be going in computational terms, and then shieldus from the usual deluge of ''neuro-technical'' terms. Because (in his own words) ''research onthe (actual nitty gitty of the) brain is advancing so quickly that any conclusion one might maketoday could be outdated in just a few weeks.''
In one reviewers words: ''The EmotionMachine rewards careful reading. You'll learn a lot about how your mind works, even if youwon't be all that much wiser about what is actually going on within your brain.'' :-)
Nevertheless, the book is obviously based on the latest advances in computer science, psychology, neuroscience, engineering etc. And certainly, his book is not a bad place to start if one wantsto be a little wiser on the emergent field of Mind Design.

Just like a machine? A six-level model of the mental activities.

Minskys brain models might make the brain look somewhat mechanical, but he is keen to tell usthat if the brain is indeed a bio-mechanical machine, it certainly isn't a machine in the usual sense.
I.e. saying that someone is like a machine usually means that he ''has no intentions, goals or emotions'' or is ''relentlessly committed to a single purpose or goal.'' But in Minskys world, it will eventually be possible to build artificial intelligences that have persistence, aim and resourcefulness. Indeed, future artificial intelligences must be equipped with many different systems for problem solving. If one method fails they should be able to switch to other approaches.
It follows that models for advanced machines (and according to Minsky, models for our minds)can't be all that simple (despite peoples desperate desire for simple formulas...).
According to Minsky: ''If a theory is very simple, you can use mathematics to predict whatit'll do. If it's very complicated, you have to do a simulation. It seems to me that for anything as complicated, as the mind or brain, the only way to test a theory is to simulate it and see what it does.''
Eventually, advanced simulations might pass the Turing Test, ''proving'' that human likeintelligence could be build on machine-like principles.
On a smaller scale, it might eventually be possible to replace damaged brain tissue - with computer hardware that performs a functionformerly carried out by neurons .
According to Minsky, the difference between machine intelligence and human
intelligence might, in the end, not be all that big.
Question: - ''If we developed the perfect artificial brain, what would be the difference between that and the real thing?''
Minsky: - ''Well, it wouldn't die!''

Internally, a human-like intelligence system must be organized in one way or another. Minsky proposes a six-level model for our mental activities. It is intended to be somewhat vague, because our brains are not so neatly arranged.

-- Values, Ideals --
- Self - conscious emotions -
- Self reflective thinking -
- Reflective thinking -
- Deliberative thinking -
- Learned reactions -
- Instinctive reactions -
-- Instinctive behavioral system --

The lowest level corresponds to the most common kind of instincts. The highest level supports the sorts of ideas that we aquire later and call by names like ethics and values. In the middle layer are layers of methods we use to deal with all sorts of problems, conflicts and goals. In the deliberative layer you might consider several actions to take, then image the effects of each, and then compare these alternatives. On top of that, at the reflective levels you may wonder if you decisions at lower levels were good decisions. And final you may ''selfreflect'' and see if your actions were in line with your ideals.
Minsky observes that: ''Just like in societies, the higher executive layers don't know enough of the systems details to specify what must be done. Hence their power actually consists in selecting among options proposed by their subordinates. In that way, the low level are actually controlling or containing what their superiors do.''

Imprimers - What do we want?

Humans wants all sorts of things. A lot of it are due to instincts that tell us to get food, a mate andso forth. Other things are not so easily describe, why would some want to play the guitar, whileothers would consider that to be about the dumbest activity on earth? Indeed, many (humans) are probably clueless to why they have certain highlevel goals, instead of some other highlevel goals!? According to Minsky we learn our high-level goals from our Imprimers. Worried that some future, human-like intelligence Cyberdyne system might kill us all? Well, you shouldn't be... Instead, you should be worried about how the system is educated, what its imprimertells it!

Minsky: ''Getting what you want is one thing. It is quite another to learn what you ought to want. Learning usually works by trial and error - improving our ways to achieve the goalsthat we already hold. When we self reflect on our goals (feel pride or shame) we might change our priorities. Humans learn what they ought to want depending how they interpret the reactions of the persons to whom they are attached. I.e. a child feel grateful and proud when praised by its mother (Generally: the imprimer).''

Imprimer: An imprimer is one of those persons to whom a child has become attached.

The present goal is therefore elevated if the imprimer praises. For evolutionary reasons, normally,only imprimers can change a childs highlevel goals, as it would be no good if strangers could change a persons highlevel goal, and make them do whatever they wanted.

At first the imprimers must be near to us, but once we have made mental models of them, we canuse those models to elevate goals even when those imprimers are absent. Eventually these models becomes conscience, ideals or morals codes.
Which btw. reminds me of Douglas Hofstadters book ''I am Strange Loop''. Here Hofstadter, after the death of his 43 year old wife Carol, writes about the self, as something distributed over many minds.According to Hofstadter: Carol's ''personal sense of 'I' '' lived on (in a ''low-resolution fashion'') as a ''loop''inside me.

In short:
When an imprimer praises, elevate you goal.
When an imprimer scolds, devalue your goal.

It is then obvious why e.g. cults remove people from their family and normal social relations. Then it is much easier to sabotage the persons old goal system! And the cult can then begin thenew impriming, where new ideals are implanted into an anxious and insecure mind.

Why should we have goals at all? The obvious answer is, that we have goals because that is how our brains evolved. The people without goals became extinct, because they simply couldn't compete.
[p. 191, Emotion Machine].

Pain - And how plans change.

We don't want to let our imprimers down. Still it might not be all that easy to achieve all of the high-level goals they have given us. Life is full of things that can go wrong. And sometimes they do. Leaving us with emotional or physical pain.

Pain: We are all equipped with nerves that connect from each part of the skin to several mapsin the brain. However, we are not born with similar ways to represent signals that come from our internal organs.Therefore we usually find it hard to describe pains that are notlocated near our skin. Before modern surgery we had little use for this information, as wecould not repair or protect internal organs anyway.

When we are ''feeling bad'', we are actually talking about the disruption of our other goals. Pain would not serve its function (for which it involved), if it allowed us to pursue our normal goals, while our bodies were being destroyed.Note that goals can not overrule pains, that would be dangerous. Imagine, tf some goal could simply turn hunger off, we would be in peril of starving to death. If we could simply stop sleeping, we would like wear our bodies out.

And pain must obviously be a part of life. I.e. It is ''pleasant'' to have accomplished a difficult task, but this always involves some transient period of severe distress and discomfort.This also applies whenever we try to improve our mental ability, Sure, pleasure might help us learn simple easy things, but we must learn to ''enjoy'' at least some distress, when it comes tolearning things that need larger-scale changes in how we think.

So, should we just stay away from pain and go for pleasure? Well, it might be dangerous! Certainly, addict knows: That when we think we choose theoption that pleases us most, the selection may actually be some process that has silencedall of its competitors.The more pleasure you feel in this way, the more negative may be that hidden effect on the rest of our mental processes.

Common sense thinking.

To find a way through the ups and downs of life, we use our ''common sense''.
According to Minsky: Typical commonsense thinking might begin with a brief ''micro-manic'' phase thatproduces a few ideas, then one find flaws, during a short-lived ''micro-depressive''period [p. 241, Emotion Machine].

I.e. many bipolar disorders (manic-depressive disorders) are really doing the samething as our everyday creative thinking. Whenever we face a new type of problem, we might use procedures like:

Turn most of your critics (Minskys word for procedures that finds problems in plans) off. This helps you to think about things you could do, with little concern about whether they'll work. As though you were in a manic state.Next turn your critics back on, to examine these options more sceptically - As though you were having a mild depression.Finally, choose an option that seems promising, and then pursue it until your critics startto complain that you are not making progress.
Critics help us to recognize the kinds of predicaments we face, and then recommend selections of ways in which we might deal with these situations. And even though our critics tell us a lot of things we might rather not have known,we do pay attention when they ''speak''.
E.g: Self conscious critics -
none of my goals seems meaningful (depression).
I am losing track of what I am doing (confusion).
I can achieve anything (mania).
My friends may disapprove of this (insecurity).

Humans like ''common-sense'' thinking and heuristics a lot better that ''logical'' thinking.Using logic is like walking a plank.It assumes that each separate step is correct. Commonsense thinking demands more support. One must add evidence after everyfew steps.Obviusly, evolutionary pressures made earlier humans rather good at ''common sense''thinking, but not so good at ''logical'' thinking. But today it would be rather convenient, if more people were better at ''logical thinking''. Indeed, perhaps it will one day bepossible to hookup humans to 'logic module''. E.g. through a neural interface.

Luckily, still, we are not forced to do ''hard logic'' to survive. Reasoning by analogy,dividing and conquering, reformulating and other ''common sense'' techniques that we are actually good at, can also be quite helpful.

Representing problems in multiple ways.

Actually, if you can only follow one particular kind of procedure, using one kind of representation - that is not all that impressive. Obviously, human brains can describe and represent problems in multiple wasys - giving many more ways to solve problems [p. 296, Emotion Machine].

And when we have learned something, solved a problem, we should be able to make useful abstractions. If we represent something in the wrong way - like say binary numbers in a connectionist net - it becomes very difficult yo make useful abstractions of that knowledge. On the other hand when something is expressed in say everyday language, there is much more expressiveness [p. 296, Emotion Machine].

The reflective mind - Choosing the right goal.

We are always comparing various goals, and deciding which ones to put aside or postpone - And one will never make much progress towards achieving any particular goal unless onecan persist at it for long enough. A higher level goal should be able to stop other processes from taking focus.At least goals that will take the same resources. We can walk and talk at the time, becausethat uses different sets of brain resources. But we can't speak and read at the same time, because that would compete for the same language resources.
[p. 249, Emotion Machine].

To accomplish any major goal, one may need to suppress most competing goals, as in''I don't fell like doing anything else''. Pleasure helps us to learn by engaging negative functions (functions that stop over processes) that works to keep our minds from ''changing the subject''.
So, how do we work out what goals to pursue right now? Based on the guidelines we have received from our imprimers we should ''self-reflect''on how things are going right now. That would include predictions that turned out to be wrong, plans that encountered obstacles, and failures to access the knowledge one needs. To see the importance of selfreflection, consider how smart it is to know that you are confused(as opposed to being confused without knowing this). Knowing that you are confused - you elevate yourself to a larger-scale view of your overall goals. Make a better plan, switch toa completely different activity?

Hopefully, we will eventually be able to decide on what to do next.
When people say that they used free will make a decision. What they are actually saying is, ''some process stopped my deliberations, and made me adopt whatseemed best at the moment.'' In other words, free will is not a process to makea decision, but one we use stop other processes.[p. 204, Emotion Machine].

If we end up following the wrong plans, luckily, we can always tell a story about it:
According to Daniel Dennett: Our fundamental tactic of selfprotection, self control and self-definition is not building dams or spinning webs, but tellingstories - about who we are.
(see my Amazon review of Cordelia Fines ''A Mind Of Its Own'' for more about this).

Limited resources.

Blaming problems on limited access to resources is always a popular solution.Errare humanum est. Really, we shouldn't be surprised that we don't achieveall of our goals with such limited brain resources at our disposal ....¨

We don't really see the world out there:The visual systems in our brains receive many more signals from the restof the brain than signals that come in from our eyes. Some 80 % fibers to the lateral geniculate nucleus relay station come downwards from thecortex, adn only about 20 % from the retinas.
[p. 153, Emotion Machine].

Our working memory really isn't that big:Where new memories are briefly maintained in resources that act like acomputers cache - and then over time, more permanent versions arecreated in other regions of our brains.
[p. 245, Emotion Machine].

And it is quite difficult to store information:Then of course there is the ''real estate'' problem for long term memories. Finding places for new memories must involve complex constraints and requirements, and this could be the reason as to why making permanentrecords take so much time. Finding appropriate networks of brain cellsto use without disrupting connections and records that one would not want to erase.
[p. 270, Emotion Machine].

There is simply too much information to keep track of:Different parts of each persons brain are involved with different forms of memory - there are a least sensory, episodic, auto-biographical,semantic, declarative and procedurel memories.
[p. 243, Emotion machine].

And we can't really redesign our own minds for better performance: There are substantial advantages in imposing limits on the extend to whichour mind can examine itself. If a mind can make changes in how it worksit will obviously also face the risk of destroying itself!
With too much control over the systems we use for remembering,it might be possible to overwrite all of a persons old memories.Turning on resources on at will could force a mind to spend all of itstime pursuing one particular goal.
[p. 341, Emotion machine].

And when we build machines to help us, they also have problems with limitedresources:The coming years will see more powerful and resourceful machines. But every system that we build will surprise us with new kind of flaws- until those machines becomes powerful enough to conceal their faultsfrom us.
[p. 345, Emotion Machine].

Learning.

Despite some resource problems and other ''design issues'' we do of course have excellent minds. One of the things that our minds do really well - is our ability to learn almost anything:
It took hundreds of millions of years for us to evolve from the earlist vertebrate fish,and it took those eons to develop the structures that became the the higher reflective layers of our brains. Human children makes extensive use of these high level structures to develop our uniquely human way of represent new kinds of knowledge and processes. This is why it is not easy to make a machine that learn just about anything. You can not learn what you can not represent!
[p. 180, Emotion Machine].

A smart way of learning is of course also necessary with so much to learn in human societies: No infant could by itsef event enough to develop an adult intelligence. So, perhaps ourmost important skill is how we learn, not only from own experiences, but alsofrom being told things by other people.
[p. 271, Emotion Machine].

Still, even for us, learning is a slow and difficult thing:Thomas Landauer concluded that, during any extended interval, none of his subjects could learn at at a rate of more than about two bits per second, whether the realm bevisual, verbal, musical or whatever.
[p. 178, Emotion Machine].

Consciousness.

Some might have hoped Minskys book would end with some grand statements aboutconsciousness. Not so!For starters, he doesn't like the word all that much.
And some of the unconscious thinking that goes on in our minds may actually be a lot more interesting?
Minsky thinks that consciousness is a suitcase-like word that we use to refer to many different mental activities. Which obviously is why it is so difficulet to talk about what consciousness really is! Is it imagery, thinking, decision making, awareness, attention, the concept of self?

Aaron Sloman (1994) makes it even more explicit: People are too impatient. They want a three line definition of consciousness and a five line-line proof that a computational system can or cannot have consciousness. They dont want the hard work of unravelling complex and muddled concepts.

Questions unanswered:

Minskys decision not to go into the nitty gritty of how our brains might actually work allows him to give a good overview, but his approach obviously leaves many questions unanswered.

Brainwaves.

E.g. it would have been nice with a few words about the binding problem , but sadly there is nothing about this in the book. Luckily, New Scientist deals with it
July 10th 2010:
Exactly where brainwaves are generated in the brain, and how they communicate information is somewhat of a mystery. But they must obviously do something these alpha-, theta-, beta- and gamma- waves. When we look at an apple, the apples redness and its roundness are picked up by different cells in the brain, but we don't see red and round - we see one thing.
The rhythmic activity of the brain waves ensures that all the relevant signals relating to a sensation arrives to a binding region at exactly the same time, where they can be recombined into a single sensation. Faulty synchronization between parts of the brain responsible for planning, executing and sensing speech could mean that a person fails to recognize that words he has just uttered is in fact his own words (as in schizophrenia).

Sleep.

And what about sleep? Why do we sleep? According to Wikipedia, there are multiple arguments supporting the restorative function of sleep. And memory seems to be affected differently by certain stages of sleep such as REM andslow-wave sleep (SWS). Indicating that procedural memory might benefit from late, REM-rich sleep,whereas declarative memory might benefit from early, SWS-rich sleep.
William Dement, a co- discover of REM sleep, isn't so sure though: ''As far as I know'', he answered, ''The only reason we need to sleep that is really, really solid is becasuse we get sleepy.''
[National Geographic, May 2010].

Still off we go for that nap. According to one theory, the Thalamus blocks input from the senses,allowing the brain to fokus on processing information from the day. The Hippocampus thenreplays memories to be stored during during REM sleep. At least studies suggest that memory consolidation may be one function of sleep (a transfer of memories from shortterm memoryto longterm memory can only take place, while unconscious, because the same networks are used for other information processing purposes while we are awake)
[National Geographic, May 2010].

==============
Perhaps the Amazon reviewer was right:

(By reading the book) You'll learn a lot about how your mind works, even if you won't be all that much wiser about what is actually going on within your brain.

Still, it is a wonderful book!

July 18th 2010.

-Simon

Simon Laub
www.simonlaub.net