Substantia. An International Journal of the History of Chemistry 2(1): 41-50, 2018 Firenze University Press www.fupress.com/substantia ISSN 1827-9635 (print) | ISSN 1827-9643 (online) | DOI: 10.13128/substantia-40 Citation: F.T. Arecchi (2018) Cognition and Reality. Substantia 2(1): 41-50. doi: 10.13128/substantia-40 Copyright: © 2018 F.T. Arecchi. This is an open access, peer-reviewed arti- cle published by Firenze University Press (http://www.fupress.com/substan- tia) and distribuited under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Data Availability Statement: All rel- evant data are within the paper and its Supporting Information files. Competing Interests: The Author(s) declare(s) no conflict of interest. Research Article Cognition and Reality F. Tito Arecchi Emeritus of Physics-Università di Firenze and INO-CNR, Firenze, Italy E-mail: tito.arecchi@ino.it Abstract. We discuss the two moments of human cognition, namely, apprehension (A),whereby a coherent perception emerges from the recruitment of neuronal groups, and judgment(B),that entails the comparison of two apprehensions acquired at differ- ent times, coded in a suitable language and retrieved by memory. (B) entails self-con- sciousness, in so far as the agent who expresses the judgment must be aware that the two apprehensions are submitted to his/her own scrutiny and that it is his/her task to extract a mutual relation. Since (B) lasts around 3 seconds, the semantic value of the pieces under comparison must be decided within that time. This implies a fast search of the memory contents. As a fact, exploring human subjects with sequences of sim- ple words, we find evidence of a limited time window , corresponding to the mem- ory retrieval of a linguistic item in order to match it with the next one in a text flow (be it literary, or musical, or figurative). While apprehension is globally explained as a Bayes inference, judgment results from an inverse Bayes inference. As a consequence, two hermeneutics emerge (called respectively circle and coil). The first one acts in a pre-assigned space of features. The second one provides the discovery of novel features, thus unveiling previously unknown aspects and hence representing the road to reality. Keywords. Human language, homoclinic chaos, synchronization of neural spike sequences, Bayes inference, inverse Bayes inference, circle hermeneutics, coil hermeneutics. 1. PERCEPTION, JUDGMENT AND SELF-CONSCIOUSNESS Figs 1 and 2 introduce the difference between A-apprehension or per- ception that rules the motor reactions of any brainy animal, and B-language ,only humans, and that provides judgments. Figs 3 and 4 show why the scientific program is a linguistic one and what is the reason of its success. With this in mind, we explore whether and how Cognition unveils Real- ity… Following the philosophy of cognition of Bernard Lonergan [Lonergan], I discuss two distinct moments of human cognition, namely, apprehension (A) whereby a coherent perception emerges from the recruitment of neuronal groups, and judgment (B) whereby memory recalls previous (A) units coded in a suitable language; these units are compared and from comparison it fol- lows the formulation of a judgment. 42 F. Tito Arecchi The first moment, (A), has a duration around 1 sec; its associated neuronal correlate consists of the synchro- nization of the EEG (electro-encephalo-graphic ) signals in the so-called gamma band (frequencies between 40 and 60 Hz) coming from distant cortical areas .It can be described as an interpretation of the sensorial stimuli on the basis of available algorithms, through a Bayes infer- ence [Arecchi,2007; Doya et al.]. Precisely, calling h(h= hypothesis) the interpreta- tive hypotheses in presence of a sensorial stimulus d (d=datum), the Bayes inference selects the most plausi- ble hypothesis h*,that determines the motor reaction, exploiting a memorized algorithm P(d|h), that represents the conditional probability that a datum d be the conse- quence of an hypothesis h. The P(d|h) have been learned during our past; they represent the equipment whereby a cognitive agent faces the world. By equipping a robot with a convenient set of P(d|h), we expect a sensible behavior. The second moment, (B),entails a comparison between two apprehensions (A) acquired at different times, coded in a given language and recalled by the memory. If, in analogy with (A), we call d the code of the second apprehension and h* the code of the first one, now – at variance with (A) h* is already given; instead, the relation P(d|h) which connects them must be retrieved; it represents the conformity between d and h*, that is, the best interpretation of d in the light of h*. Thus, in linguistic operations, we compare two suc- cessive pieces of the text and extract the conformity of the second one on the basis of the first one. This is very differ- ent from (A), where there is no problem of conformity but of plausibility of h* in view of a motor reaction. Let us make two examples: a rabbit perceives a rus- tle behind a hedge and it runs away, without investigat- ing whether it was a fox or just a blow of wind. On the contrary, to catch the meaning of the 4-th verse of a poem, we must recover the 3-d verse of that same poem, since we do not have a-priori algorithms to provide a satisfactory answer. Once the judgment, that is, the P(d|h) binding the codes of the two linguistic pieces in the best way, has been built, it becomes a memorized resource to which to recur whenever that text is presented again. It has acquired the status of the pre-learned algorithms that rule (A). However-at variance with mechanized resources- whenever were-read the same poem, we can grasp new meanings that enrich the previous judgment P(d|h). As in any exposure to a text (literary, musical, figurative) a re-reading increases our understanding. (B) requires about 3 seconds and entails self-con- sciousness, as the agent who expresses the judgment must be aware that the two successive apprehensions are both under his/her scrutiny and it is up to him/her to extract the mutual relation. At variance with (A), (B) does not presuppose an algorithm, but rather it builds a new one through An inverse Bayes procedure [Arecchi, 2010]. This construc- tion of a new algorithm is a sign of Figure 1. Plato said that we see the shadows of things, like a pris- oner constrained to view the end of a cave and forbidden to turn and see the outside world. This occurs indeed in perceptual tasks, where the sensorial stimuli are interpreted by “algorithms” and gen- erate (within1 sec) a motor reaction. The procedure is common to all brainy animals. Figure 2. In linguistic operations, humans code a perception in a linguistic code andretrieve it by short term memory (around 3 sec)comparing it to a successive coded perception. From the com- parisonit emerges a connection that increases the details of the observed thing. 43Cognition and Reality Creativity and decisional freedom Here the question emerges: can we provide a com- puting machine with the (B) capacity, so that it can emulate a human cognitive agent, as expected in theTur- ing test? The answer is NOT, because (B) entails non- algorithmic jumps, insofar as the inverse Bayes proce- dure generates an ad hoc algorithm, not available previ- ously. The scientific endeavor can not be carried on by an AI (artificial intelligence ) device, since it entails a lin- guistic step, as shown in Fig. 3. Fig. 4 explains why Gali- leo’s program provides certainties, rather than probabil- istic expectations. 2. THE BRAIN OPERATIONS-ROLE OF HOMOCLINIC CHAOS Let us introduce “deterministic chaos”.[Arecchi,2004 b] Since Poincaré (1890) we know that a dynamical sys- tem is extremely sensitive to the initial conditions. That yields the so called “butterfly effect” whereby a tiny shift in the initial conditions yields a large difference in course of time . Precisely, a difference of initial con- ditions induces a divergence of the dynamical trajecto- ries in course of time. In the case of a meteo dynamical model, accounting for most atmospheric features (wind, pressure, humidity, etc) but neglecting the motion of a butterfly wing could lead to a wrong prediction(from sunny to rainy).Loss of the initial information occurs over a time t whose inverse is called K (Kolmogorov entropy). In the meteo model such a t may be days, in a dynamical model of the solar system it takes millions of years. We call geometric chaos the above trajectory diver- gence. Another type of chaos,that we call temporal chaos, consists of regular closed orbits that however repeat at irregular times (Fig. 5). A single neuron in the brain undergoes tempo- ral chaos and its electrical output consists of a train of spikes (each one high 100mV and lasting 3 ms). The minimal inter-spike separation is 3 ms; the aver- age separation is 25 ms in the so-.called g band of the EEG(electro-encephalo-gram). A neuron communicates with other neurons in two ways [Arecchi,2004 a, Singer,Womelsdorf. & Fries]: – either directly , by coupling its spike train to anoth- er neuron via an electric line called axon, – or indirectly, by building with nearby neurons a local potential (detectable as an EEG signal) and providing a signal χ to a distant neuron, that conse- quently re-adjusts its firing rate. Fig. 6 shows the direct synchronization of two trains over a time Dt. The neurons involved in the coupling are confined in a thin layer of the brain (thickness 2 mm) called the cortex. Groups of nearby neurons contribute to a common task forming a specialized area that builds global interactions with other areas (Fig. 7). The areas are visualized via the amount of oxygen- ated blood required by a working region and visualized by f-MRI (functional magnetic resonance imaging). Fig. 8 visualizes the competition between two neu- ron groups I and II fed by the same sensorial (bottom- up) stimulus, but perturbed (top-down) by differ- ent interpretational stimuli provided by the long term Figure 3. The scientific program is a linguistic task. Galileo’s approach consists of extracting mathematical features; it implies a linguistic operation, according to Fig.2. Figure 4. Once the object under investigation is reduced to a col- lection of mathematical features, one applies millennia of math- ematical wisdom and predicts the future behavior. 44 F. Tito Arecchi memory. I wins, as the corresponding top-down stimu- lus succeeds in synchronizing the neuron pulses of this group better than in group II. This means that, over a time interval Δt, neurons of I sum up coherently their signals, whereas neurons of II are not co-ordinated, hence yielding a smaller sum. As a consequence a reader GWS (= global workspace, name given to the cortical area where signals from different areas converge; it is located in area F of Fig.7) reads within Δta sum signal overcoming a suitable threshold and hence eliciting a motor response [Dehaene]. Thus, the winning interpretation driving the motor system is that provided by I. What represented in Fig. 8 models the mechanism (A) common to any animal with a brain. 3. PERCEPTION AS A BAYES INFERENCE Neurosciences hypothesize a collective agreement of crowds of cortical neurons through the mutual synchro- nization of trains of electrical pulses (spikes) emitted indi- vidually by each neuron [Singer et al., Dehaene et al.] .The neuro-scientific approach is summarized in Fig. 8. However, a global description of the above process can be carried on in probabilistic terms, without recur- Figure 5. Homoclinic chaos; the dynamic trajectory is a closed orbit starting from S (saddle point) and returning to it. Projecting on a single direction, we observe spikes P repeating in time. The time separation between two spike occurrences depends on the relations between α and g, thus it can be controlled by a voltage applied to S, as the signal χ. Figure 6. Direct coupling of two neurons by synchronized spike trains; synchronization missed after Dt for an extra-spike in the upper train. Figure 7. Topology of specialized cortical areas ,each one being active as a large collection of synchronized neurons; mutual com- munication occurs via EEG signals. multisensory interactions will combine into a unified pattern involving frontal cortex, temporo- parietal regions as well as unimodal cortices: A = auditory cortex; V = visual cortex; M= higher-order multisensory regions; F = prefrontal cortex Figure 8. Competition of two cortical areas with different degrees of synchronization. 45Cognition and Reality ring to the details of the process. In 1763, Thomas Bayes, looking for a reliable strat- egy to win games, elaborated the following probabilistic argument[Bayes]. Let us formulate a manifold of hypoth- eses h about the initial situation of a system, attributing to each hypothesis a degree of confidence expressed by an a priori probability P(h). Any hypothesis, introduced as input into a model of evolution, generates data. Let us assume that we know the model and, hence, can evaluate the probability of the data conditioned by a specific hypothesis h; we write it as P(data|h). The model is like an instruction to a computer, thus we call it algorithm; it generates different data for differ- ent h. If then we perform a measurement and evaluate the probability P(data) of the data, we must conclude that there is an h more plausible than the other ones, precisely the one that maximizes the probability conditioned by the data P(h|data), that we call the a posteriori probability of hand denote as h*. This procedure is encapsulated in the formula, or theorem, of Bayes, that is P(h*)=P(h|data) = P(h) [P(data|h)/P(data)] To summarize, the a posteriori probability of h, conditioned by the observed data, is given by the prod- uct of the a priori probability of h, times the probability P(data|h) of the data conditioned by a given h, that we call the model, and divided by the probability P(data), based on a previous class of trials (Fig. 9). Fig. 10 summarizes the whole perception procedure, that is initiated by an external stimulus and concluded by a motor reaction [Arecchi, 2007]. Successive applications of the theorem yield an increasing plausibility of h*; it is like climbing a moun- tain of probabilities along its maximum slope, up to the peak. After each measurement of the data and conse- quent evaluation of the a posteriori h*, we reformulate a large number of new a priori h relative to the new situ- ation, and so on (Fig. 11). Notice that Darwinian evolu- tion by mutation and successive selection of the best fit mutant is a sequential implementation of Bayes theorem. 4. LINGUISTIC OPERATIONS AS INVERSE BAYES In Fig. 11 the recursive application of Bayes using the same algorithm- or model- is visualized as climbing a probability mountain. The bit length of the algorithm is the Algorithmic Complexity of the cognitive task. However, in everyday life we experience jumps toward different algorithms, that means going to climb different mountains (Fig. 12). The associated multiplicity of choices corresponds to attributing different meanings Figure 9. Bayes inference. Figure 10. Starting with a large number of presumed hypotheses h, the occurrence of the data selects the h* that satisfies the above relation and drives a suitable reaction. 46 F. Tito Arecchi to the input data; the number of alternative choices will be called Semantic Complexity [Arecchi, 2007]. This swap of the model is a creative jump proper of language operations. It is the root of Goedel-1931 incompleteness theorem and Turing-1936 halting problem for a computer, as dis- cussed in a previous paper [Arecchi, 2012]. Altogether different from (A) is the situation for (B), that – implying the comparison between different appre- hensions coded in the same language (literary, musi- cal, figurative, etc.) – represents an activity exclusively human. In fact, the second moment (B) entails the compari- son of two apprehensions acquired at different times, coded in the same language and recalled by the memory. (B) lasts around 3 sec; it requires self-consciousness, since the agent who performs the comparison must be aware that the two non simultaneous apprehensions are submitted to his/her scrutiny in order to extract a mutu- al relation. At variance with(A), (B) does not presuppose an algorithm but it rather builds a new one through an Inverse Bayes procedure introduced by Arecchi [Arec- chi,2010]. This construction of a new algorithm is the source of creativity and decisional freedom. Language indeed permits an infinite use of finite resources [Humboldt]. It is the missing step in Turing’s claim that human intelligence can be simulated by a machine [Turing]. The first scientist who explored the cognitive rele- vance of the 3sec interval has been Ernst Pöppel [Pöppel 2004, 2009]. This new temporal segment has been little explored so far. All the so-called “neural correlates of conscious- ness” (NCC) are in fact electrical (EEG) or functional magnetic resonance (fMRI) tests of a neuronal recruit- ment stimulating a motor response through a GWS (see Fig. 8); therefore they refer to (A). In such a case, rather than consciousness, one should call itperceptual awareness,that we have in common with brainy animals. Fig. 13 shows how an inverse Bayes procedure pro- vides the best comparisons of two successive pieces of a linguistic text, thus generating a judgment. While in perception we compare sensorial stim- uli with memories of past experiences, in judgment we compare a piece of a text coded in a specific language (literary, musical, figurative) with the preceding piece, recalled via the short term memory. Thus we do not refer to an event of our past life, but we compare two successive pieces of the same text. Such an operation requires that: i) The cognitive agent be aware that he/she is the same examiner of the two pieces under scrutiny; ii) The interpretation of the second piece based upon the previous one implies to have selected the most appropriate meanings of the previous piece in order Figure 11. Recursive application of Bayes is equivalent to climbing a probability mountain, guided by the Model ,that is, the condition- al probability that an hypothesis generates a datum. This strategy is common e.g. to Darwin evolution and to Sherlock Holmes criminal investigation; since the algorithm is unique, it can be automatized in a computer program (expert system). Figure 12. Comparison of two different complexities, namely, i)the algorithmic C. , corresponding to the bit length of the program that enables the expert system to a recursive Bayes; and ii)semantic C., corresponding to the occurrence of different models. Bayes M EANING INFORM ATION complexity, semantic----à (multiple models) complexity, algorithmic (single model) creativity= swap of model Climbing up a single peak is a non-semiotic procedure ON THE CONTRARY Jumping to other peaks is a creativity act, implying a holistic comprehension of the surrounding world (semiosis) 47Cognition and Reality to grant the best conformity (from a technical point of view, this conformity is what in the philosophy of cognition of Thomas Aquinas was defined as truth = adaequatio intellectus et rei (loosely translated as : conformity between the intellectual expectation and the object under scrutiny) In Fig. 10 we have generically denoted as top-down the bunch of inner resources ( emotions, attention) that, upon the arrival of a bottom-up stimulus, are respon- sible for selecting the model P(d|h) that infers the most plausible interpretation h* driving the motor response. The focal attention mechanisms can be explored through the so-called NCC (Neural Correlates of Consciousness) [Koch] related to EEG measurements that point the cor- tical areas where there is intense electrical activity pro- ducing spikes, or to f-MRI (functional magnetic reso- nance imaging) that shows the cortical areas with large activity which need the influx of oxygenated blood. Here one should avoid a current confusion. The fact that a stimulus elicits some emotion has NOTHING to do with the judgment that settles a linguistic compari- son. As a fact, NCC does not reveal self-consciousness, but just the awareness of an external stimulus to which one must react. Such awareness is common to animals, indeed many tests of NCC are done on laboratory animals. It is then erroneous to state that a word isolated from its context has an aesthetical quality because of its musical or evocative power. In the same way, it is erro- neous to attribute an autonomous value to a single spot of color in a painting independently from the compari- son with the neighboring areas.All those “excitations” observed by fMRI refer to emotions related to apprehen- sion and are inadequate to shed light on the judgment process. The different semantic values that a word can take are associated with different emotions stored in the memory with different codes (that is, spike trains). Among all the different values, the cognitive operation “judgment” selects that one that provides the maximum synchronization with the successive piece. Thus emotions are necessary but not sufficient to establish a judgment. On the other hand, emotions are necessary and sufficient to establish the apprehension as they represent the algorithms of the direct Bayes infer- ence. This entails a competition in GWS (Fig. 8) ,where the winner is the most plausible one; whereas in the judgment-once evoked the panoply of meanings to be attributed to the previous piece- these meanings do not compete in a threshold process, but they must be com- pared with the code of the next word in order to select the best interpretation. Recent new terms starting with neuro-( as e.g. neu- ro-ethics, neuro-aesthetics, neuro-economics, neuro- theology) smuggle as shear emotional reactions decisions that instead are based on judgments. The papers using those terms overlook the deep difference between appre- hensions and judgments. The question is discussed in detail in the Conclusions. A very successful neurological research line deals with mirror neurons, that is, neurons that activate in subjects (humans or higher animals) observing another subject performing a specific action, and hence stimulate mimetic reactions [Rizzolatti]. Here too, we are in pres- ence of mechanisms(empathy)limited to the emotional sphere, that is, very useful for formulating an Apprehen- sion, but not a Judgment. 5. TWO DIFFERENT HERMENEUTICS, THAT IS, INTERPRETATIONS OF COGNITIVE DATA Fig. 14 shows how a cognitive agent A reads an object B. The CIRCLE refers to a Bayes cognition, where- by an algorithm is taken as necessary and sufficient to generate knowledge of B. Whenever A reconsiders B, he/she finds the same B already memorized. On the contrary, expressing the knowledge in a lan- guage and comparing successive pieces by inverse Bayes, entails an increase of details of B (B1,B2, etc.) that improve the cognition of the agent (A1,A2, etc). As for the CIRCLE, in information science, an ontology is a formal definition of the properties, and Figure 13. The inverse Bayes procedure that occurs in linguistic endeavors, whereby a previous piece ofa text is retrieved by the short term memory and compared with the next one: the appro- priate conditional probability is no longer stored permanently but it emerges as a result of the comparison (judgmentand consequent decision). 48 F. Tito Arecchi mutual relationships of the entities that exist for a par- ticular domain of discourse. An ontology lists the vari- ables needed for some set of computations and estab- lishes the relationships between them. For instance, the booklet of the replacement parts of a brand of car is the ontology of that car. The fields of artificial intelligence create ontologies to limit complexity and to organize information. The ontology can then be applied to prob- lem solving. Nothing is left out; we call this cognitive approach “finitistic” as no new insight is provided by repeated trials. On the contrary, in any human linguistic endeavor (be it literary, or musical or figurative) A starts building a provisional interpretation A1 of the text ; whenever A returns to B, he/she has already some interpretational elements to start with, and from there A progresses beyond , grasping new aspects B2, B3…and hence going to A2 and so on (COIL). To carry on a COIL program, we do not need a large amount of resources; language makes an infinite use of finite resources [Humboldt]. The COIL hermeneutics describes also the inter-per- sonal dialogue. If the object B of cognition is a human person as A, then the changes B1,B2, etc are not only due to an increased knowledge by A, but also to an intrinsic change of B who re-adjusts his/her relation with A. Thus, if B is another human subject, then B under- goes similar hermeneutic updates as A; this is a picture of the dialogical exchange between two human beings. (persons). 6. CONCLUSIONS- TWO ASPECTS OF LINGUISTIC CREATIVITY We conclude by stressing two well known aspects of linguistic creativity. First, if we start a linguistic endeav- or, a wealth of possible situations emerge , giving rise to ambiguous behaviors as it occurs in most products of human creativity, that – like the Etruscan Chimera – display apparently contradictory behaviors, from Ulyss- es to Dom Quixote (Fig. 15). The onset of Chimeras is explained in Fig. 16 as the lack of an external referent B. Altogether different is the what takes place when the language is interpreting scientific observations. In fact, the repeated comparison extracts elements of reality, as hinted in the COIL hermeneutics. Figure 14. Two kinds of interpretation of a text, or hermeneutics, namely, the CIRCLE, whereby the interpreter A attributes a finite and fixed set of meanings to the text B, and the COIL, whereby A captures some particular aspects of B and-based on that informa- tion- A approaches again the text B discovering new meanings. The novel insight provided at each coil is an indication of how language provides new semantic apertures. Figure 15. A linguistic action that proceeds from a known piece toward an unknown one is like the Etruscan Chimera: it can gen- erate mutually conflicting behaviors , as it occurs in most charac- ters ,from Ulysses to Dom Quixote.When instead the linguistic comparison regards two observed items (as it occurs in reading the verses of a Poem, but also in scientific observations), then we really increase our personal knowledge with an element of reality. Figure 16. How chimeras emerge in linguistic creations. 49Cognition and Reality Applying our hermeneutics to the scientific pro- gram, we have two possible approaches. As the size of the observed world increases from a few particles to many, within a universal scientific description associated with a fixed-algorithm (as an AI tool would operate) we witness an exponential increase of the size C of the com- putational program ( called the algorithmic complexity) as well as a reduction of the time interval t over which predictions are reliable, that is, an increase of the Kol- mogorov entropy K=1/t (Fig. 17). A more efficient scientific program consists of lin- guistic comparisons of different situations, with the help of inverse Bayes inference, applying non-algorithmic jumps as the horizontal lines of Fig. 12. We are some- what manipulating the set of attributions of the item under study, emphasizing novel aspects and overlook- ing some previous ones. Such a change of paradigm [Kuhn] leads to novel theories with lower C and K. A very familiar example is the formulation of Maxwell’s electromagnetic equations, unifying electric, magnetic and optical phenomena. The novel values we attribute to some features leads to the so called effective science [Hartmann, 2001]. An outstanding example is offered by Landau theory of phase transitions. In Fig. 18 we list the so-called neuro-by products (neuro-ethics, neuro-aesthetics, neuro-economics, etc.) meaning that the decisions in that specific matter result from brain processes signaled by f-MRI. One would attribute to emotions the role that is instead proper of a linguistic act, thus requiring an inverse Bayes. Precisely, the emotions select a particular meaning to be assigned to the piece of text h* in order to optimize its matching with the next one d; they are crucial for maximizing P(h*|d) but they are not all , just a piece of the whole tapestry (see Fig. 13). Final consideration. Does AI operate by inverse Bayes in a linguistic elaboration? Answer: only in a very limited way. Indeed, AI refers to a built-in “ontology”, consisting of a large, yet finite, list of properties of each item. (See e.g. the informational use of the term ontology to list the component parts of a car). Thus AI can build P(h*|d) for each h* and d , and it can do it in a much faster way than a human. However the human exploits emotions in selecting the meaning of h*, thus he/she can go beyond the large, yet limited ontology available to AI and attribute to h* novel aspects previously unknown. This is the creativity of human language, already addressed by Humboldt , and absent in AI. BIBLIOGRAPHY Arecchi F.T. (2004a). Chaotic neuron dynamics, synchro- nization and feature binding, Physica A 338: 218- 237. Arecchi F.T. (2004b). Caos e complessità nel vivente (lezi- oni tenute alla Scuola Universitaria Superiore, Pavia), IUSS Press-Pavia, pp. 248 Figure 17. Normal science vs. paradigm shift  effective science [Hartmann, Kuhn]. Let C be the bit length of the algorithm and K the Kolmogorov entropy, i.e., the inverse of the time t beyond which the initial information is lost by dynamical chaos. A com- puter program that evaluates Kepler’s orbits, as BACON, has small C and K. As the physical system gets richer, both C and K increase and a scientific search carried on by an AI system would be affect- ed by higher and higher C and K. However, a linguistic actor as a human scientist can act by a “jump of paradigm”, that is, change code and introduce a new scientific theory (effective description) with low C and K. Against the NCC (= neural correlates of consciousness) role (F.Crick, C. Koch) NCC means : there is a brain area , localized by f-MRI , whereby specific human endeavors emerge , classified as - ethics - aesthetics - economics - theology - mathematics etc. “neuro-” On the contrary, in phenomenology (from Husserl to Varela) it makes no sense to isolate the brain but one must account for the whole cognitive being in dialogue with the world . The inverse Bayes stems from a comparison of meanings and NOT from a single emotion highlighted by f-MRI. Figure 18. It is fashionable to speak of neuro-ethics, neuro-aesthet- ics, neuro-economics, etc., overlooking the essential role of inverse Bayes. Hence, the neuro-xxxx should be regarded as scientific mis- understandings. 50 F. Tito Arecchi Arecchi F.T. (2007). Complexity, Information Loss and Model Building: from neuro- to cognitive dynamics, SPIE Noise and Fluctuation in Biological, Biophysical, and Biomedical Systems, Paper 6602-36. Arecchi F.T. (2010). Dynamics of consciousness: com- plexity and creativity, The Journal of Psychophysiology, 24(2), 141-148. Arecchi F.T. (2012). Fenomenologia della coscienza: dall’apprensione al giudizio, in “...e la coscienza? FENOMENOLOGIA, PSICO-PATOLOGIA, NEURO- SCIENZE” a cura di A. Ales Bello. Bayes T. (1763). An Essay toward solving a Problem in the Doctrine of Chances, Philosophical Transactions of the Royal Society of London 53, 370-418 Dehaene S., Sergent C. & Changeux J.-P. (2003). A neu- ronal network model linking subjective reports and objective physiological data during conscious percep- tion, Proceedings of the National Academy of Science (USA), 100(14), 8520-8525. Doya K., Ishii S. & Pouget A. (Eds.) (2007). Bayes- ian Brain: Probabilistic Approaches to Neural Coding. MIT Press. Galilei G. (1612). Terza lettera a M. Welser sulle macchie solari, Opere, vol. V (pp. 187-188). Firenze: Edizione Nazionale, Barbera 1968. Hartmann S. (2001). Effective field theories, reduction and scientific explanation, Studies in History and Phi- losophy of Modern Physics 32B, 267-304. Humboldt W.(1836). On Language: On the Diversity of Human Language Construction and its Influence on the Mental Development of the Human Species. Trans. Peter Heath, Cambridge: CUP, 1988. Koch C. (2004). The quest for consciousness: a neurobio- logical approach. Englewood, US-CO: Roberts & Co. Publishers. Kuhn T. (1962 1st ed.; 1996, 3rd ed.). The Structure of Sci- entific Revolutions, University of Chicago Press. Lonergan B.S.J (1970). Insight, A study of human under- standing, Philosophical Library, New York. Pöppel E. (2004). Lost in time: a historical frame, elemen- tary processing units and the 3-second window, Acta Neurobiologiae Experimentalis, 64: 295-301. Pöppel E.(2009). Pre-semantically defined temporal win- dows for cognitive processing, Philosophical Transac- tions of the Royal Society B, 364,1887-1896. Rizzolatti G. et al. (1996). Premotor cortex and the rec- ognition of motor actions, Cognitive Brain Research, 3(2), 131-141. Singer W. & Gray C.M. (1995). Visual feature integra- tion and the temporal correlation hypothesis, Annual Reviews of Neuroscience 18, 555-586. Singer W. (2007). Binding by synchrony, Scholarpedia, 2(12), 1657. Tommaso D’Aquino (1269). Summa theologica, Pars Pri- ma, Quaestio 16. Turing A. (1950). Computing Machinery and Intelli- gence. Mind 59, 433-460. Womelsdorf T. & Fries P. (2007). The role of neuronal synchronization in selective attention, Current Opin- ion in Neurobiology, 17, 154-160. Substantia An International Journal of the History of Chemistry Vol. 2, n. 1 - March 2018 Firenze University Press Why Chemists Need Philosophy, History, and Ethics Emulsion Stability and Thermodynamics: In from the cold Stig E. Friberg Finding Na,K-ATPase Hans-Jürgen Apell Mechanistic Trends in Chemistry Louis Caruana SJ Cognition and Reality F. Tito Arecchi A Correspondence Principle Barry D. Hughes1,* and Barry W. Ninham2 From idea to acoustics and back again: the creation and analysis of information in music1 Joe Wolfe Snapshots of chemical practices in Ancient Egypt Jehane Ragai The “Bitul B’shishim (one part in sixty)”: is a Jewish conditional prohibition of the Talmud the oldest-known testimony of quantitative analytical chemistry? Federico Maria Rubino Michael Faraday: a virtuous life dedicated to science Franco Bagnoli and Roberto Livi