CATEGORY:
Wandering Towards a Goal Essay Contest (2016-2017)
[back]
TOPIC:
Minimal goal-directed systems by Joscha Bach
[refresh]
Login or
create account to post reply or comment.
Author Joscha Bach wrote on Mar. 7, 2017 @ 16:36 GMT
Essay AbstractHow is it possible for a physical system to have goals? In the way we commonly use the notion, goals have qualities that are different from attractor states that characterize the dynamics of a system’s progression. Goals are part of the characterization of at least minimally intentional systems. Thus, our question cannot be extricated from: What is the minimal descriptive frame that allows us to describe a system as representing, choosing, committing to and pursuing goals? And when we can answer that: what processes and dynamics are necessary and sufficient for the genesis of such a system?
Author BioJoscha Bach is a cognitive scientist and AI researcher, specializing in cognitive architectures, Artificial General Intelligence, and models of motivation and decision making. He obtained his PhD from the Institute of Cognitive Science in Osnabrück, worked at Humboldt University of Berlin and the MIT Media Lab, and is currently affiliated with Harvard's Program of Evolutionary Dynamics in Cambridge, MA.
Download Essay PDF File
Ines Samengo wrote on Mar. 7, 2017 @ 20:04 GMT
Bravo! I think you made an excellent job in pinning down the computational components required for agency and goal-directed behavior. In particular, you seem to have answered the question that you posed at the beginning: "What are the minimal requirements to construct a system that is capable of causal modeling, assigning preferences, and making decisions?" I buy all your arguments in this respect. I would like to ask you, however, if you believe that you have also provided arguments of why all these processes are so ubiquitously instantiated in the world we live in (if you believe they are. Why should observers naturally tend to develop encoding functions, prediction functions, preference functions, and decision functions. Or should they? I understood that these functions are required for goals to exist at all, but why do they so often develop naturally? If you believe to have explained this, could you just summarize it in a few keywords, so I can go back and spot them? If not, is it because you believe we still do not know the answer to that question, or because you preferred not to focus in that point in this limited-word essay?
Thanks for the good material! ines.
report post as inappropriate
Author Joscha Bach replied on Mar. 9, 2017 @ 21:44 GMT
Thank you for the friendly commentary, Ines!
I think that the ability of modeling the world and the needs of the system that attempts to maintain stability in that world are driven by the utility of developing that model, and enabled by evolution.
In other words, if you have the resources for adaptive regulation, via information processing that is general enough for universal computation, you will on average outperform all similar systems if you develop a model that allows for goal-directed action.
Luis Patiño-Cuadrado wrote on Mar. 8, 2017 @ 09:09 GMT
Dear Joscha:
From the abstract of your essay, I gather that you credit computational processes for goal-oriented behavior -- including life and consciousness. Could you please explain to me briefly in terms of your essay why I feel fear, love, joy sadness? Have insights? Why I know I exist? Why I am not you?
Of course, I probably should read your whole essay, but you sound like a very bright person and would love to read just a brief explanation first.
Thank you. I would refer you to my essay, but it has not been posted yet.
All the best,
Luis
report post as inappropriate
Author Joscha Bach replied on Mar. 9, 2017 @ 21:57 GMT
Dear Luis, I think that affective states are
configurations of cognition, given by
modulators, such as arousal, valence, attentional focus, and our ability to deal with and assess the situation (coping potential, uncertainty). The modulators adapt us to the requirements of the situation at hand, so our cognition works differently when we are rewarded, aroused, focused etc. We perceive these differences as differences in action tendencies, and the "color" of our cognitive content as emotions.
Insights result from settling in a new model state for a given situation, i.e. from building a representation that lets you integrate and predict our observations better than before.
You don't know that you exist. What you perceive as "I" is a story that your brain tells itself, and that is created or re-created whenever you try to reflect on it. You are a fiction that your brain reinvents, usually many times in each minute.
If your self model gets dissociated from your short-term biography, body image, sense of perspective, and sense of agency, you may lose the feeling that you are different from any object in your mind, including me. This may happen in dreams, via meditation, or as a result of psychiatric disorder or the influence of drugs. You may lose the ability for self-differentiation before you lose the ability for self-identification. In this state, it may appear to you that you are "one with me" and every other object your self identifies with.
sherman loran jenkins wrote on Mar. 8, 2017 @ 19:13 GMT
Joscha,
Any one interested in experiments with living agents, make friends with a slime mold. They are agents that have goals and the will to reach for them.
report post as inappropriate
Stefan Keppeler wrote on Mar. 9, 2017 @ 21:35 GMT
Dear Joscha,
I find it interesting how you argue that observers and other "causal systems implemented in a reversible substrate will have only a temporary existence" (within in the section "Reversible computational structures"). In the same section you also mention "ergodic systems". I think I understand what you write immediately before and after this point, but I fail to see how ergodicity enters the game. Could you expand on that?
Cheers, Stefan
report post as inappropriate
Author Joscha Bach replied on Mar. 10, 2017 @ 04:26 GMT
If you observe an organism for a period of time, each of the observed states has a nonzero probability of being observed again in the future. (This is admittedly not strictly true for the whole organism, because it may suffer injuries or form memories etc., but for large subsets of its functionality, it is a useful characterization.)
If a system revisits previous states, it must be either in a perfect loop, or it must perform irreversible state transitions.
Stefan Keppeler replied on Mar. 10, 2017 @ 09:58 GMT
I see what you mean. Is it obvious that this is equivalent to some standard notion of ergodicity, see e.g.
ergodicity or
ergodic process?
report post as inappropriate
Anonymous replied on Mar. 13, 2017 @ 02:13 GMT
Dear Stefan, I think it is not, and you are right to point that out. I considered organismic regulation, for instance of body temperature, as a Markov process. If all states are recurrent and aperiodic, it is ergodic; once your states are transient, it usually means that the organism is in decay. I was unaware of the differences in understanding ergodicity in other contexts, such as dynamical systems and statistical mechanics, and I should have referenced and explained my thought more clearly. More importantly, even the Markov notion of ergodicity is not sufficient (or necessary) for the overall argument, and had no proper place here. Thank you for bringing that to my attention.
report post as inappropriate
Author Joscha Bach replied on Mar. 13, 2017 @ 02:15 GMT
(prev post was mine; I was not properly logged in)
Stefan Keppeler replied on Mar. 14, 2017 @ 13:02 GMT
Thanks for clarifying. Due to
my own essay I had so much Boltzmann in my mind that the word just led my astray. It doesn't affect your argument, of course. Good luck, Stefan
report post as inappropriate
hide replies
Conrad Dale Johnson wrote on Mar. 11, 2017 @ 13:50 GMT
Joscha –
Excellent essay… this helps me think more clearly about several interesting issues. For example, your distinction at the beginning between mathematical, computational and computable is very helpful. I was also struck by your notion of fundamental physics as a causally closed layer of information processing, which can contain no information about the system that implements...
view entire post
Joscha –
Excellent essay… this helps me think more clearly about several interesting issues. For example, your distinction at the beginning between mathematical, computational and computable is very helpful. I was also struck by your notion of fundamental physics as a causally closed layer of information processing, which can contain no information about the system that implements it. Altogether you give a very clear and insightful description of the levels of computational complexity arising in nature.
Though it’s not intended as a criticism of your arguments, I do want to suggest that the computational perspective is limited in ways that I think are important. Most basically, it presupposes the existence of definite information in the form of logical bits, along with functional mechanisms that read and write, store and copy, transmit and logically manipulate this information. At the quantum level we have the more complex qbits, which additionally require the so-far unexplained functionality of “measurement” to produce determinate information. Clearly our universe provides all this functionality – but as I suggest
in my essay on the natural technologies that support meaningful information, this is not implemented by some underlying invisible hardware layer. Rather, essentially all of atomic and molecular physics operates as a recursively self-sustaining process that defines and continually redefines all its own information.
Likewise when it comes to the technologies of life and of human communication, I think the computational approach needs to be supplemented by an understanding of the different levels of meaning that emerge in the context of self-replicating systems, or systems that can transfer complex information between brains. There is surely a sense in which we can say, “all this is computation.” But as you point out, there are extremely unlikely events like the origin of life – and probably many others, like the origin of eukaryotic cells – that “have to be helped along by a generous number of cosmic dice throws.” These are game-changers, because they suddenly open up new contexts of possibility in which the same computational machinery can begin to operate at a new scale of effectiveness.
So to get a more complete picture, we need to describe both the computational functions and the contexts they operate in, that support them. You could be right that the computational perspective is the great contribution to philosophy made in the last century… but I think we’re only in the first stages of understanding the other side of the picture.
Thanks again for a very fine piece of work.
Conrad
view post as summary
report post as inappropriate
Author Joscha Bach replied on Mar. 19, 2017 @ 05:43 GMT
Conrad, thank you for your kind comment!
Your remark ultimately seems to concern the "it from bit" debate. To me, it seems that information is much more elementary than the derived notions of "atomic and molecular physics" you mention. It is a bit unfortunate that we usually discuss information in the Shannon sense, i.e. as information
about something, rather than in the computational sense, i.e. as atoms of state.
I tend to take a strong computationalist perspective, which embarrassingly requires the existence of a machine providing the necessary functionality for computing the universe. I don't understand how a "recursively self-sustaining process", as you seem to envision it, could pull the universe out of the swamp of non-existence all by itself, but I'd love to see it!
Shaikh Raisuddin wrote on Mar. 13, 2017 @ 12:14 GMT
Joscha Bach,
Each word of a dictionary has its "physical correlate".
You need to rephrase definitions in terms of physicality.
For example, INTENTIONALITY is a state of matter caused by shortage or surplus of something physical. A balanced state of matter is unintentional.
A heterostatic system alone can acquire intentionality.
The fact "Physics is Behavioural Science of Matter", can help us correlate material properties with mental concepts.
report post as inappropriate
Joe Fisher wrote on Mar. 14, 2017 @ 16:23 GMT
Dear Dr. Joscha Bach,
Please excuse me for I have no intention of disparaging in any way any part of your essay.
I merely wish to point out that “Everything should be made as simple as possible, but not simpler.” Albert Einstein (1879 - 1955) Physicist & Nobel Laureate.
Only nature could produce a reality so simple, a single cell amoeba could deal with it.
The real Universe must consist only of one unified visible infinite physical surface occurring in one infinite dimension, that am always illuminated by infinite non-surface light.
A more detailed explanation of natural reality can be found in my essay, SCORE ONE FOR SIMPLICITY. I do hope that you will read my essay and perhaps comment on its merit.
Joe Fisher, Realist
report post as inappropriate
Satyavarapu Naga Parameswara Gupta wrote on Mar. 16, 2017 @ 09:58 GMT
Nice essay Joscha Bach,
Your ideas and thinking are excellent for eg…
How is it possible for a physical system to have goals? In the way we commonly use the notion, goals have qualities that are different from attractor states that characterize the dynamics of a system’s progression.
I propose another answer for that direct...
view entire post
Nice essay Joscha Bach,
Your ideas and thinking are excellent for eg…
How is it possible for a physical system to have goals? In the way we commonly use the notion, goals have qualities that are different from attractor states that characterize the dynamics of a system’s progression.
I propose another answer for that direct question….…………
………………… At this point I want you to ask you to please have a look at my essay, where ……………reproduction of Galaxies in the Universe is described. Dynamic Universe Model is another mathematical model for Universe. Its mathematics show that the movement of masses will be having a purpose or goal, Different Galaxies will be born and die (quench) etc…just have a look at my essay… “Distances, Locations, Ages and Reproduction of Galaxies in our Dynamic Universe” where UGF (Universal Gravitational force) acting on each and every mass, will create a direction and purpose of movement…..
I think intension is inherited from Universe itself to all Biological systems For your information Dynamic Universe model is totally based on experimental results. Here in Dynamic Universe Model Space is Space and time is time in cosmology level or in any level. In the classical general relativity, space and time are convertible in to each other.
Many papers and books on Dynamic Universe Model were published by the author on unsolved problems of present day Physics, for example ‘Absolute Rest frame of reference is not necessary’ (1994) , ‘Multiple bending of light ray can create many images for one Galaxy: in our dynamic universe’, About “SITA” simulations, ‘Missing mass in Galaxy is NOT required’, “New mathematics tensors without Differential and Integral equations”, “Information, Reality and Relics of Cosmic Microwave Background”, “Dynamic Universe Model explains the Discrepancies of Very-Long-Baseline Interferometry Observations.”, in 2015 ‘Explaining Formation of Astronomical Jets Using Dynamic Universe Model, ‘Explaining Pioneer anomaly’, ‘Explaining Near luminal velocities in Astronomical jets’, ‘Observation of super luminal neutrinos’, ‘Process of quenching in Galaxies due to formation of hole at the center of Galaxy, as its central densemass dries up’, “Dynamic Universe Model Predicts the Trajectory of New Horizons Satellite Going to Pluto” etc., are some more papers from the Dynamic Universe model. Four Books also were published. Book1 shows Dynamic Universe Model is singularity free and body to collision free, Book 2, and Book 3 are explanation of equations of Dynamic Universe model. Book 4 deals about prediction and finding of Blue shifted Galaxies in the universe.
With axioms like… No Isotropy; No Homogeneity; No Space-time continuum; Non-uniform density of matter(Universe is lumpy); No singularities; No collisions between bodies; No Blackholes; No warm holes; No Bigbang; No repulsion between distant Galaxies; Non-empty Universe; No imaginary or negative time axis; No imaginary X, Y, Z axes; No differential and Integral Equations mathematically; No General Relativity and Model does not reduce to General Relativity on any condition; No Creation of matter like Bigbang or steady-state models; No many mini Bigbangs; No Missing Mass; No Dark matter; No Dark energy; No Bigbang generated CMB detected; No Multi-verses etc.
Many predictions of Dynamic Universe Model came true, like Blue shifted Galaxies and no dark matter. Dynamic Universe Model gave many results otherwise difficult to explain
Have a look at my essay on Dynamic Universe Model and its blog also where all my books and papers are available for free downloading…
http://vaksdynamicuniversemodel.blogspot.in/
Be
st wishes to your essay.
For your blessings please…………….
=snp. gupta
view post as summary
report post as inappropriate
Willy K wrote on Mar. 19, 2017 @ 05:22 GMT
Dear Bach
Along with Alicea’s essay dealing with the Law of Requisite Variety, your essay on Conant’s Good Regulator Theorem seem tailor made for me to understand the functioning of those two crucial cybernetics principles. I believe the two laws are both essential for me to take my own essay further in its development. Thanks so much for putting it in the right perspective.
You go on to state, “The same is often even true for coordinated groups of agents, i.e. organization of agents can implement supervenient causal structures that realize all criteria of agency by themselves”. I suspect that the model that I have built for social systems (Constitutional nation state) may be an instance of what you are describing.
I give the highest rating to your essay since it is a scientifically rigorous write up, besides being so clearly relevant for my own effort. Please do let me know if my interpretation in the above paragraph is incorrect. Cheers!
Warmest Regards, Willy
report post as inappropriate
Author Joscha Bach replied on Mar. 19, 2017 @ 05:32 GMT
Dear Willy, nation states are certainly agents, and those that invented their constitutional codes clearly understood that laws are a part of a normative software that is animating a golem, made from individuals not unlike individuals are made from cells. I think you are clearly onto something!
Miles Mutka wrote on Mar. 19, 2017 @ 15:41 GMT
The essay makes its points very clearly, although I disagree with most of it. The Abstract promises discussion of attractors, but I saw none in the text itself.
report post as inappropriate
Member Larissa Albantakis wrote on Mar. 22, 2017 @ 15:01 GMT
Dear Joscha,
I enjoyed reading your essay a lot. Definitely gave me much food for thought. Can I ask you something about the concept of reversible computers? Is it required that every micro process in the system is reversible, or only that the state of the computer as a whole is reversible? In other words can there be an AND logic gate in the computer as long as some other gates keep track of the state of the inputs to the AND gate, or is no AND operation allowed?
Best regards,
Larissa
report post as inappropriate
Author Joscha Bach replied on Mar. 22, 2017 @ 18:14 GMT
Dear Larissa, thank you so much for investing your time and attention to read this!
You can implement an AND gate in a reversible system if you keep track of the history, i.e. for a single binary AND, you will have two garbage bits (one is not enough, because 0 has three possible predecessor states). For instance, in a Fredkin gate, for A&B, you get the garbage outputs !A&B and A (some "recycling" is usually possible later on).
If you want to implement your reversible computer in hardware to deal with the Landauer limit, the implementation will of course matter, but from a logical perspective, it is irrelevant whether you store an UNDO record separate from your AND gate or use some kind of Fredkin logic.
If we treat the universe as a reversible computation, we might be interested in identifying a minimal structure that can yield reversibility, such as a reversible cellular automaton rule. An interesting property of any deterministic finite cellular automaton is that must eventually enter a periodic state (even if the period is 1). Once we are in the loop, we can always find a reversible rule that completely describes the behavior of the automaton. Thus, if we leave a cellular automaton universe alone for long enough, it will become reversible, even if it did not start out this way! Of course, it is not clear that this has any similarity to our universe, especially since we don't know that our universe is periodic.
Don Limuti wrote on Mar. 29, 2017 @ 00:53 GMT
Hi Joscha,
A comprehensive essay on computation and agency.
I do favor the computational method in that it can produce a program that can be observed and evaluated.
You have a statement that I question: "Mathematics is the realm of all possible specifications, computation is the realm of implementations." This makes mathematics the foundation of computation. My problem with accepting your thesis is that mathematics is axiomatic. Some have no problem with a point as being a dimensionless object. Some (like myself) find this nuts but useful. My feeling is that when mathematics is postulated as the foundation of reality, we enter a new realm that is interesting but is not reality. Any thoughts you have on this appreciated.
And I believe you are solving real problems that will enlighten the future. And computational methods and demonstrations are very useful.
Please visit my essay (and be as critical as you like....really).
Thanks for a very good essay,
Don Limuti
report post as inappropriate
George Kirakosyan wrote on Apr. 4, 2017 @ 09:05 GMT
Hi dear Joscha.
I have opened your article because of its meaningful title. It should be comprehensible to everybody to start study of subject from its possible simplest forms and properties. You have emphasized this important methodological point, which shows the serious approach to the theme. Meantime, it is possible to see in the contest many our advanced thinker-brothers, who immediately rushed to explain, let me say, - how these or other decisions arise in the brains of important leaders of our society! Unfortunately, I am not so enough specialist to say some useful for you remarks. I can say only that your essay seems corresponding to scientific approach, as there are definitions, criterion, as well description on the methodology. You have touched to observer’s important role that is the close to me dealt. You say; “Thus, a goal has complex prerequisites, it does not just entail the tendency of a system to increase the probability of an reaching a state!” I see here that you well realized whole complicity of the question. This, mainly, has pushed me to go a little bit side in my essay (I hope you will open it and you can say some words in my page.) I see yours work as properly written and meaningful - interesting that deserves to study and to a good support that I will do now!
Best wishes!
report post as inappropriate
Dizhechko Boris Semyonovich wrote on Apr. 7, 2017 @ 04:44 GMT
Dear Sirs!
Physics of Descartes, which existed prior to the physics of Newton returned as the New Cartesian Physic and promises to be a theory of everything. To tell you this good news I use spam.
New Cartesian Physic based on the identity of space and matter. It showed that the formula of mass-energy equivalence comes from the pressure of the Universe, the flow of force which on the corpuscle is equal to the product of Planck's constant to the speed of light.
New Cartesian Physic has great potential for understanding the world. To show it, I ventured to give "materialistic explanations of the paranormal and supernatural" is the title of my essay.
Visit my essay, you will find there the New Cartesian Physic and make a short entry: "I believe that space is a matter" I will answer you in return. Can put me 1.
Sincerely,
Dizhechko Boris
report post as inappropriate
Login or
create account to post reply or comment.