There is an interesting and relevent (to this discussion) article posted 4 days ago here:

http://arxiv.org/abs/...

The title and abstract are:

---

Cognition and Emotion: Perspectives of a Closing Gap
Authors: Claudius Gros
(Submitted on 16 Feb 2010)

Abstract: The primary tasks of a cognitive system is to survive and to maximize a life-long utility function, like the number of offsprings. A direct computational maximization of life-long utility is however not possible in complex environments, especially in the context, of real-world time constraints. The central role of emotions is to serve as an intermediate layer in the space of policies available to agents and animals, leading to a large dimensional reduction of complexity.
We review our current understanding of the functional role of emotions, stressing the role of the neuromodulators mediating emotions for the diffusive homeostatic control system of the brain. We discuss a recent proposal, that emotional diffusive control is characterized, in contrast to neutral diffusive control, by interaction effects, viz by interferences between emotional arousal and reward signaling. Several proposals for the realization of synthetic emotions are discussed in this context, together with key open issues regarding the interplay between emotional motivational drives and diffusive control.

Comments: A review. Cognitive Computation (in press)
Subjects: Neurons and Cognition (q-bio.NC)
Cite as: arXiv:1002.3035v1 [q-bio.NC]

---

As far as I understand the article he is arguing that emotions come from brain chemicals that diffuse throughout the brain to regulate cognition and behavior.

Frank

Frank, Sorry it’s taken me so long to get back at you regarding this article.( http://arxiv.org/abs/.... It makes some good points and the author seemed to be heading in a similar direction.

There are some differences though, extensions of his directions rather than disagreements. He didn’t seem to make a connection between the fireworks show of emotion and the system of needs that provide motivation. Maybe that’s because there isn’t one. But I don’t think so. From my perspective, initial experiences after birth consists of the conceptualizations of the physical environment. Wet, cold, movement, limb freedom, sound, etc. At this time that 6-10 lbs. of human confusion knows little of cognition. It is just gathering information about the environment. It does know feelings. From the womb it has developed an environmental expectation of consistency. This paper doesn’t address the gradual nature of the development of how feelings are intertwined within conceptualizations. I was left without understanding the underlying mechanisms and I believe that this is key to the emotional controls he refers to. He does recognize that there is a connection. He states: “Emotions correspond to homeostasis at the behavioral level.” And offers referances of others to support the statement. However I would have liked to have him make the connections.

Further, he states an assumption regarding a projection he makes about AI: “The artificial cognitive system would then, in this view, take into consideration all facts known about the environment, make a large computational effort, and maximize directly a preset utility function.” Previously he makes a case for the necessity of emotional control over cognition and than only refers to facts as being the only component of its ability to “maximize directly a preset utility function”. I don’t see the understanding.

Also he states: “Emotions do in general not direct the behavior towards a straight-forward utility maximization”. Than he states that “A vast body of clinical data shows that emotions and the organization of behavior through motivational drives are intrinsically related”. This seems to be a contradiction. The 2nd statement supports a connection to motivation and the first seemingly denies it.

In his conclusion he states: “The motivational problem cannot be solved by cognition alone, any highly developed cognitive system would remain goal-less stranded and disoriented without solid and well developed emotional groundings.” And I totally agree with this.

Overall I enjoyed the article. Thanks!

Randall, Rhetoric? My self-impression and intention was to do the opposite. I intended to discuss a different approach to AI. One that handles emotion and thought developmental progression from a different perspective. However, I have only begun to explore its basics and what I’ve received back demonstrates mostly misunderstanding. I consider this to be my problem! They are reacting to what they know of what I am saying and to date my perspectives seem to have little or no meaning to anyone else at all. But they do for me. This perspective describes what necessates thought clearly. It may only be clear to me but the critical thinking necessary to dispel this has yet to be presented. I can and will do this, if necessary, on my own! However, I am not here to talk to myself and I do encourage critical thinking. So, if you see anything amiss please inform me. I welcome questions, regardless of their nature, tone or intent, frank or otherwise. Bring it! I will answer or thank you for showing me the light. This project has and will continue to cost me a fortune! If it fails now it will be a finical boon to me. Failing later will certainly be a tail with a different theme.

I don’t see this as a one way flow of information. Hopefully, at least, we will have some lively discussion. Perhaps, even reveal some new approaches. I do have supporting research that assisted me with this. Mine is more a new arrangement than new thought.

Lex,

Good. The big problem with this group, the problem that could be a great benefit, is the vastly different backgrounds each of us bring to the topic of AI. The rough and tumble discussions that result are at least in part simply a mismatch in vernaculars. We have got to come to a shared vocabulary. That is why I keep pressing the rhetoric alert button. When the group is given enough information about a particular member's background and intent, the mapping of vernaculars is afforded in a natural way. Don't get discouraged. Just keep digging into your intent and exposing it for all to see. If we all do this, we will come naturally to a public understanding of a larger knowledge of the range of AI probability space.

Monica posts a book list for a reason. I think we should all take it upon ourselves to read through these basic tomes.

And yes, I am cantankerous. But my intent is always towards deeper understanding across a larger base of thinkers.

What drives your interest in AI? (answer in one paragraph or less)

Randall Reetz

Randall, Your skepticism regarding rhetoric is wise to say the least. It’s easy to talk a lot and say little.

My interest in AI comes from the first time I can remember seeing Robby the robot in that 50s sci-fi movie. I remember thinking “Man, I gota get one of those!” My background is Computer science/ Psychology. Back in the mid 70s it occurred to me that what we now refer to as reductiontionism was never going to result in true AI. At the time I hadn’t completely committed myself to comp sci and I was waist deep in my psych training. Then I realized that as far as I knew nobody had ever tried to apply a psych model to AI. It made absolutely sense! This would be a test of the fractured , puzzle like world, which is psych and an entire different direction for AI. I would be filthy rich and famous! As it turns out being rich and famous has more to with luck and circumstance than idea. Also a successful idea has as much to with communication as it does with inception. However, my idea may be right or wrong but AI still needs a new direction and psychology is still as fractured as ever. I am old enough to realize that I can no longer afford the “rhetoric”. That a psych model is, to me, still a valid approach. And that others might find my rational at least an interesting direction.

OK Lex,

That is a start. You have a life long fascination with robotics. You left robotics to your avocation pile and studied psychology (why?). And then for some reason late in that project you switched to AI because you thought your training in psych lent a plausible solution to the advancement of AI. You feel that AI needs a new direction (why?).

To my reasoning, there is much missing in this rhetorical expose. Most people are motivated by a need in their own life (riches, comfort, power, fame, recognition, or some sort of societal agenda). I assume there is more to your story than you have exposed.

Your claim that AI needs a new direction is popular. I liken this reaction to a group trying to find there way through the Northwest Passage. After months of iced-in ships, trekking by dog-sled, being eaten by polar bears, etc., it would be natural for some explorers to wonder if they were going in the right direction, if the continent was larger than theorized, if a passage existed at all, if humans would ever be able to weather the elements, etc. These thoughts might come to explorers just one ridge away from the Pacific. Failure to reach a destination doesn't necessarily mean the method or direction isn't correct. Sometimes it just means the task is more daunting than originally anticipated, sometimes it means one hasn't walked far enough, sometimes it means the direction is correct but more resources need to be tasked, sometimes it means the destination doesn't exist, sometimes it means, as you have suggested, that the method or direction needs to be re-evaluated.

I am curious as to your claims. I am an infrastructuralist. Meaning, I think behaviors are secondary to structure. Meaning, I believe things like emotion and psychology (at least as they are typically described by psychologists and neurologists and behaviorists and cognitive scientists) are more likely to be behaviors that emerge from deeper architectures once they cross some quantitative complexity and capacity threshold.

But I am also of the opinion that the typical descriptions given for the category "emotions" are likely to be far more complex than the artifice from which they emerge. My argument, is that emotions (very simple low-level situationally-attuned survival-prioritized attention-restrictors), play a large formative and structural roll at the base infrastructure level that is intelligence (in any form), but that emotions as you describe them (high level complex behaviors) are not formative or architectural at all – only the result of lower level essential essential systems.

I remain dubious that a human-centric high-level psychological model offers a much in the way of guiding the design of an intelligent system. I am concerned that high-level models like psychology only serve to rhetorically cloud our understanding of intelligence with a capital "I" in much the same way that early cars were thought of as "horse-less" carriages. Imagine had we decided 110 years ago that automobiles would only exist once we could build a mechanical steam or internal combustion powered four legged horse!

What I would like you to do Lex, is explain why you think that a psychological-engine needs to be pre-designed into synthetic intelligence and why that makes more sense than a model that expects psychology to emerge from a lower-level architectural base?

Randall Reetz

Randall, I, as most, would like to be able to roam the earth and go and do as I please. Whatever, wherever, whenever that may be! That was true 35 years ago as well as it is now. I have always felt that life could be easier regardless of how easy it currently is. I’m not trying to hide this, it just doesn’t seem that important. Maybe you’d like to discuss this with me personally, my guess is not. Other than the riches, fame and glory, I would like to participate. It would be terrific if I was the star (I’d be great at it) but chances are that’s not going to happen. Most of my past time in silence has been spent waiting for hardware to evolve to the point where I could afford the mass amounts of memory and mass storage a project of this nature is going to require. At the present I still don’t know if available hardware will be cost allowable to achieve my goals. I do think that it’s close. I should be able to produce a scaled down version.

Regarding Psychology, it is not my religion! I discuss these principles because they apply. You are right to point out that current psych methodology doesn’t or hasn’t explained behavior on a level low enough to enlighten us regarding intelligence. But that is the opportunity that inspires us to face now. The Psychologists needed to see the logic within the irrational. Sure an occasional “The world is flat!” type of idea has comes out of there, but, you can’t believe that AI has done better! If they have I haven’t seen it. There is great potential for AI! The world is literally starving for its possibilities. It is my intention to go where the ideas work.

“I am an infrastructuralist. Meaning, I think behaviors are secondary to structure. Meaning, I believe things like emotion and psychology (at least as they are typically described by psychologists and neurologists and behaviorists and cognitive scientists) are more likely to be behaviors that emerge from deeper architectures once they cross some quantitative complexity and capacity threshold.”

Me to! Almost. I do great right up to the “once they cross some quantitative complexity and capacity threshold.” Point. I see intelligence as a symptom, with more complex creatures displaying more of the symptom. It might be my misunderstanding, but I don’t see you offering much with this description. Maybe darkness comes early for me, but what seems clear to me is that animals use information from their input structures to indicate how they feel. These feelings are used to indicate what action if any needs to be taken. This is not the only function of sensory information. Another use is conceptualizations. This is where complexity comes into the picture. It seems to solely to be a product of memory. The more data you can fit in your accumulator at one time the more intelligent you are and the more complex your concepts can be.

I’ve left out the part of what gets conceptualized and what doesn’t. This is what feelings provide. Experiences are like clothes an entity tries on and can see what feels relevant. Concepts are taken from these experiences. Evolution has provided animal kind with the dimensions of these good feelings in the form of homeostasis. If an entity finds itself outside of these limits it means dire consequences. The part of what goes on within these limits is of the most interest to me. This is the beginning of concept aggregation. The more complex these aggregations become the more intelligent the entity.

The requirements of an AI entity so far are input devices complex enough to allow conceptualization of the environment. A list of homeostatic values (needs list) consistent with how it will survive in the environment and what we need of it to accomplish. Conceptualization programming. Control programming. And output devices capable of responding to the environment. I know that this doesn’t explain much. As is obvious this is barely a description and deffenatly not a plan. I can only want to point to the road at this time. And I haven’t begun with thought, creativity and a considerable amount of other relevant stuff.