CEFISES Seminar: Philemon Kongo, “Paraemic Intelligence: Toward a Situated Model of Reasoning for AI

[music] Yeah, I think I think we're good. Okay. So, >> can you hear me? Can you hear me? Very well. >> Very well. >> Okay. Good. >> So, let me welcome everybody online and in person for uh this last web seminar of this academic year. Today we have the pleasure to receive Philon Congo who will give us a talk entitled panic intelligence towards a contextual and socially situated model of reasoning for artificial intelligence please good. So good afternoon everybody. Good morning or good afternoon. Yes, I going to talk about pyic intelligence and uh I modified a a bit uh the title for clarity toward artificial understanding and socially situated reasoning. Uh the talk will based on uh my works on artificial intelligence. Uh I I began I began this works about uh 10 10 years about uh since 10 10 years. So I want to to leave it and uh especially I want uh you to react about what I will talk and uh to suggest something to propose something uh available because the purpose is to uh to publish a something a book about my theory on paric intelligence. So let us go. The history of artificial intelligence can be interpreted as a succession of attempts to understand and produce the mechanism through which human beings, decide and act under conditions of asertent. Since the publication of fundamental works in artificial intelligence, researchers have sought to formalize rational behavior through increasingly sophisticated computational models from symbolic reasoning systems to contemporary machine learning architectures. The central objective has remained remarkably stable. The construction of artificial agent capable of operating intelligently under conditions of astanti. Yet despite considerable progress, important questions remain concerning contextual understanding, social interpretation and meaningful explanation. It is precisely within this unresolved space that pmic intelligence becomes relevant. Human reasoning rarely operates through the street application of formal [clears throat] rules alone. nor is it to rely exclusively on statistical corre uh individuals constantly draw upon [clears throat] culturally accumulated experience socially validated patterns of interpretation and context sensitive for forms of practical wisdom. Such forms of reasoning enable human communities to navigate a certainty in ways that are often remarkably efficient despite the absence of explicit formalization. Among the most widely spread and enduring manifestations of such practical wisdom are proverbs. In our talk we will rather use the term paria par or parmius in program. I introduced the concept of pmic intelligence as theoretical framework for understanding these processes. The central thesis defined here is that paramic reasoning constitutes a distinct form of contextual intelligence whose inferential mechanisms remain only partially represented in contemporary artificial intelligence systems. The argument advanced in my talk is not that artificial intelligence should simply incorporate paramas into its knowledge basis. Such an approach would reduce paric reasoning to a collection of a linguistic artifact. Rather, the objective is to identify the underlying cognitive architecture that makes paramic reasoning possible and to explore its implications for the design of a future intelligent systems. More specifically, I shall argue that paramic reasoning involves four interconnected operations. One, analogical recognition. Two, contextual selection. Three, interpretive negotiation. Four, collective validation. Together, these operations constitute what may be called paramic intelligence. Instead, it functions through the context mobilization of a socially accumulated knowledge for the purpose of understanding and acting [clears throat] within uncertain situations. The ultimate objective of this p this my talk and of my paper which I prepare is therefore twofold. First, it seeks to establish paramic reasoning as a legitimate object of scientific investigation within AI research. Second, it proposes pamic intelligence as a conceptual framework capable of enriching contemporary approaches to reasoning under uncertainty. The first point is the opening question. Artificial intelligence has become extraordinarily powerful. Large language models can write aces, solve equations, generate images, and even imitate reasoning. Yet a fundamental question remains. Do machines understand or are they merely producing statistically possible output? This question lies at the heart of contemporary debate in artificial intelligence. Two, from artificial intelligence to artificial understanding. For 70 years, AI research has pursued a common objective, the construction of systems capable of intelligent behavior. Symbolic AI sorts intelligence through logic. Machine learning sorts intelligence through data. Deep learning sort intelligence through prediction. But prediction is not understanding. Human beings do more than calculate. They interpret. They contextualize. They negotiate meanings. They justify their actions socially. Perhaps intelligence itself is incomplete without understanding. Three, the legacy of John Mati. John Makati as we know all of us recognize a contract difficulty. Common reasoning common sense reasoning is not monotonic. For example, monotonic logic tells us that if the statement Q follows data A and Q is included in B, then Q follows from B. Similarly, the semantic notion of consequence is also monotonic. We say that a entails Q denoted as we as you can see if Q is true in all models of A. But if A then uh then Q and A is included in B then every model of B is also a model of A which shows that B then Q B coming from Q or B produce produce a Q uh that the the monotonic logic the the way the monotonic logic operates in non-monotonic logic we know as we know also is not the same if uh the proposition is modified we know that the conclusion must be changed If you add or another if you add information over the first information you must also change the conclusion. >> That is the way the common sense reasoning reason. We know also a part a part of uh a part of common sense reasoning. Uh we have also another um theory another theory exceptions constantly appear with Raymond Ra which uh talk talked about um default logic in artificial intelligence and we have also human beings revised beliefs with AGM theory. Uh we have um for example Maninson which uh who uh proposes with Ashiron and garden first who propose proposed this theory of AGM which uh focus on focused on revised belief. Anyway, circumcription and the nonmonotonic logic represented the major advances. I can I leave I show some picture of Raymond writer. He was he was a Canadian. He was Canadian. Unfortunately is already dead. And uh we have David McKinson who is a Australian. Those are the legacy of John Mari. Of course there are many many other musicians uh who are uh in the the category of legacy of John Makati. But I can I just give this to as a example. If we see if we can reflect a bit of this legacy of John John Makati, we can we can notice something uh which I wanted I wanted to add so that but uh even now it even nonmonotonic reasoning remains principally concerned with information. Uh that is the point. The information what uh how the man human how human can uh incorporate information he must incorpor incorporate information as robot or as human. That is the point. Human reasoning concerns something more. It concerns meaning. What is the meaning of this information? This distinction is crucial for the missing dimension. Current AI systems answer how humans ask why does this make sense? Explanability is therefore not only computational. It is interpretive and interpretation is fundamentally social paras intelligence. Parianas are often considered as folklore. I propose a different interpretation. Parameas are compo compressed models of experience. They constitute cognitive mechanisms. They allow commun communities to reason under uncertainty. They organize collective memory. They guide decisions. They generate explanations. The importance lies not in the the proverb or paria itself but in the architecture hidden behind it. Defining paramic intelligence. I define pomic intelligence as a contextual and socially situated form of reasoning that transforms collective experience into intelligible action. It differs from symbolic intelligence uh role based and statistical intelligence which focus on datadriven. So parameic intelligence constitute a third paradigm contextual intelligence. Four fundamental operations. Paramic reasoning rest upon four cognitive mechanisms. First, analogical recognition. Humans recognize structural similarities, not identical situations, but analogous structures. That is the idea and theory of Gner. She is an American uh from at Boston mighty. So meaning begins with analogy. The second uh characteristic of paramic intelligence is contextual selection. No explanation exist outside context. Context determines relevance. Context determines meaning. That is the idea of Brazilian and day. And uh I Brazilian must be a French from France. Brazil. That is the contextual selection from the my theory of paramic intelligence. Three, interpretive negotiation. Understanding is never automatic. It emerges through interpretation. Different interpretations coexist. Reasoning becomes dialogue. That is the idea of me and Shman. Fourth, collective validation. Rationality is social. Knowledge is not merely possessed. Knowledge is negotiated. Meaning images collectively. That is the ideas of archings and government. Now we can see beyond symbolic and statistical intelligence. If you see beyond that those feature of artificial intelligence, we can u we can talk about symbolic AI ines structure that is as we know the the charactic characteristic of symbolic AI and statistical AI. size is adaptation. But for paramic intelligence, I propose to emphasize uh meaning meaning that to say uh to understand what uh the proposition what the result means for human. It introduces context, interpretation, social cognition, and collective understanding. Explanable AI revisited. Current explanable AI asks how was the output produced? Paraming intelligence asks why is this output meaningful? Transparency is insufficient. Users seek intelligibility. Explanations are social phenomena. As Tim Mirror reminds us, explanations are designed for humans, not for machines. Distributed cognition. Edwin Arching demonstrated that cognition extend beyond individuals. Thinking is distributed across communities, artifacts, institutions, languages. And for paramic reasoning, uh parabic reasoning embodies precisely this principle. Collective memory becomes distributed distributed intelligence multi- aent systems. The future of AI is no longer individual. It is collective. Autonomous agents must negotiate, cooperate, revise beliefs, construct common meanings. But he must also for paramic intelligence. Paraming intelligence suggests that disagreement is not a failure. Disagreement is productive. Interpretation emerges through negotiation. Irresponsible AI that is the problem of ethic for AI for artificial intelligence. Luchiano Fed and Virginia DNA remind us AI must remain socially accountable. Optimization alone is insufficient. Meaning matters. Human values matter. Interpretation matters. So from paraming intelligence we introduce these dimensions into AI architecture toward artificial understanding. Perhaps we need a distinction artificial intelligence versus artificial understanding. Artificial intelligence optimizes. Artificial understanding interprets. Artificial intelligence predicts. Artificial understanding explains. Artificial intelligence processes information. Artificial understanding constructs meaning. And now we can see the problem of computation for paramic intelligence computation architecture. I just want to propose it's just a preliminary um model of computation but I hope that I will work on it and uh precise everything for the computation. The future paramic architecture could contain four modules. analogical memory, contextual evaluation, interpretive deliberation, social validation. uh we can see u among uh those four um points there's interconnection there's relation inter relation between uh different processes an analogical must uh rely on contextual contextual evaluation and context evaluation must rely on um interpretive deliberation and uh interpretive deliberation of course must rely on uh social based on social validation. Such systems would integrate symbolic knowledge, contextual cognition, social reasoning, explanability, philosophical implications. Of course, uh my theory uh produce eventually is some reaction or some implication. There must be some consequences. This proposal raises a deeper question. Perhaps intelligence itself has been defined too narrowly. Intelligence must must may not consist exclusively in solving problems. Intelligence may consist in making situations intelligible. This shifts AI research from computation toward understanding. Conclusion. Allow me to conclude with the following proposition. The future of artificial intelligence may depend not only upon learning from data but also upon learning how human communities transform experience into understanding. Paramic intelligence is therefore not a theory about proverbs or paria parames. It is a theory about meaning. And perhaps the next revolution in artificial intelligence will not consist in building more powerful machines, but in building systems capable of participating meaningfully in the social production of understanding. Thank you. We take one or we go directly to the questions. This will take five. Yeah. So we take uh FO we take five minutes and after that uh questions and answer. Okay. >> Okay. Okay. Five minute breaks. >> [music] [music] >> Hey, [music] hey, hey. >> [music] [music] >> Hey. [music] Hey. >> [music] >> Hey. [music] Hey. Hey. >> [music] [music] >> Hey. [music] Hey. Hey. >> [music] >> Okay. Question, comments. [clears throat] >> Well, I can start. >> Yeah. >> Okay. Thank you very much uh Filimo for the for the talk. I'm always very sympatic symposic therapy sympatic to uh to enter a little bit of contextuality in human reasoning since it's clearly not formal but I I was surprised about the the part about analogical recognition using GNA and thank you to remind me it's been a long time I run because one of the idea of GNA if I remember correctly is that we basically think in metaphor the brain is analogical but the the main cultural tools is metaphor. So why did you choose analytical recognition rather than metaphor recognition which would fit better with the the case study of proverbs? Seems that most proverbs use more metaphor than analogy. Yeah. Uh, can I reply? >> Yeah, please. >> Okay. Uh, okay. uh metaphor metaphor as a analogy. That is the those are the some kind of process uh which parasmic intelligence or paras use to uh to manage uh in my work in my thesis I explained everything about metaphor and analogy. Those are some kind of mechanisms um used in parimeology uh to uh construct to construct uh the con the situation and to uh to [clears throat] to show to show the reflection or or some characteristic of uh of phenomena. Uh let me explain. If um we can say that uh parameas used to to manage those kind of mechanism that is the same but the difference those metaphor as as analogy are the same mechanisms but the the difference reside in the construction of um each of each of the both in analogy as uh Gentner say said um there's not similarity similarity there's um common ar structure of the phenomena, common structure and then you transport this common structure to another [snorts] phenomena. That to say that is not a a simple simple comparison but is a construction in analog analogy there's construction. >> Okay. So as even to to construct a paria you must use analogy and but for metaphor metaphor is also a transport a transport of the a just a a feature of the phenomena to another uh phenomena. That is the difference I can say. Uh the analogy is for construction of the reasoning. Construction of the reasoning completely. But metaphor is the transport of one one feature of the phenomena to another one. but is not is not dealing with the construction of another uh reasoning pyomic reasoning that is the difference for me. So, so it's more structural analogical is more structural >> but those are the the same mechanism of this kind of this kind of reasoning >> resonate. Yeah. The principle the principle feature of this kind of this phenomenon. Other questions, comments? I have a bit of a a bit of a similar question. Um, and it probably it probably also comes from your from your other work. So, I really like the talk. One one thing that I'd love to to hear just a little bit more about is how you went from your analysis of perennials to those four features that are required to to understand them if the the architecture the uh analogical recognition contextual selection interpretive negotiation and uh and collective validation. How did you what what's the link there? But say say more about how that's how you see those as being so essential for how we understand. >> Uh yes. Uh uh first of all first of all we must know that is just a a proposition. I propose to scientific scientists this theory and uh as I see in my as I can see in my uh work analogical recognition is I can say is the first the first me the first process there's Don't there's so to speak there's there's there are four processes for computation we have analogy analogical recognition. So an analogical recognition is to compare is a kind of comparison uh of the phenomena. We have one phenomena in our background and we compare this the the phenomena which is in front of us which we uh we dealing we can deal with uh with the the first which is in our background. That is the first the first process for uh can I say I can say also for um proposing a computational a computational model uh because I don't have for the moment a computational model that to say some rules some uh mechanism logical technical mechanism for this theory but I can say for analogical recognition is just a compar comparison uh we have in our background we have uh can I say I can say in our data we have some examples. So we compare this the the one which we are dealing with with our background and we can see the difference and uh uh and give or or uh if we we found we will also find some difference but we can produce Another one. Another one which is uh which is uh contextually um irrelevance irrelevant contextually relevant. The second is contextual selection. We must also selection choose we can we must choose the context for the phenomena with uh which we are dealing with. We have in front of a phenomena a phenomenon and now we want to explain the the purpose we must know the purpose is to understand. [clears throat] So we must choose the context. This is the context. We can have different kind of context. It's can be in uh psychology in scientific especially uh matters or uh living in commercial affairs uh psychology that is those are different context. Um and we have to choose which context is irrelevant and uh after that we must interpret and we must negotiate the interpretation because um I talked about uh multi malta agent uh in artificial intelligence for now we are dealing just uh with uh individuals. But in the future, we'll have many agents uh who uh will have the the same knowledge or the same ability uh one another as uh for every everyone. So we have to interpret, we have to to negotiate to to hear or to collect the proposition of all aspects about the phenomenon and choose socially the the best inter interpretation for the case. that is the situation uh the selection of situation and uh for the fourth characteristic is collective validation. Uh we must for the problem of ethic ethical problem we must also uh seek for the validation of the society. Uh for example uh in AI uh today someone can have his drone you know the drone and he can decide uh to to use this drone for the purpose of military purpose and he can use it as he find best for him but not for the society, not for uh uh social purpose that is will be uh I can say uh a problem for the society. You cannot use uh technology from AI for your own um your own uh uh what can I say for your own purpose for your own benefit. You must also uh consider the society that this this characteristics shows shows that um uh show that uh we have also to consider the ethical problem of artificial intelligence. So the collective validation is needed for the for paramic reasoning for paramic intelligence. Now together these operations constitute what may be called paramic intelligence. You you cannot choose uh use one of them in alone. You must use all those operations for the purpose uh of uh uh for the best the best the best results uh of what can I say for the society for the benefit of the society for the benefit of of for yourself and uh for the future of uh the environment. ment for the future of the world that is I can say what I can respond for your question I think that is it okay >> yeah that's all thanks Marco >> thank you so much for your talk I just had some know questions about like curiosity because I'm outside of this domain. So concerning the contextual selection, you said that we need to choose uh the context and you mention the fact that we could choose a context in psychology, a context in another science. So are you saying that the context is uh um fieldbased? So the context means something between a scientific domain or is more like a social uh cluster. What is the context? is more um related to different field or to specific uh social context. I don't know if >> Yeah. Okay. Um the context we can say um is uh the contextual selection is about um five five points. We can consider the circumstances. The first of all the circumstances which is the circumstance of this phenomena for this reasoning that to say paramic intelligence propose to consider the circumstance. The case, the case. This is the case. If we can use analogical cognition, of course, we have a background. But for this circumstances or for this circumstance what we can say what what we can we can do. Second is intentions. The intention is uh also a matter of contextual. The intention which is the intention. We know the intention is aboutic problem matters. uh we can also deal with that ethical problemic matter that to say what what is your intention to use uh this uh device can I say this device of technology from AI what is your intention your intention is for example for the milit military purpose is to kill for killing people. What is the intention? Uh so I can say the second is also valuable as [snorts] the first circumstances the case. The second is intention. The third is relationships. Relationships is uh your action your your uh reasoning uh can be relate relate with another reasoning with another kind of uh problem. That is the pro that is the the the third the third I can say characteristic for the context selection. The fourth is historical background. The historical background is uh of course in the history we know that uh this phenomena uh yield yielded some kind of problem or this this problem or resolve the problem. So you have to consider this context the the historical background and the last one is the social expectations. You must not see only your uh your desire to to to realize your desire or to uh to reach your purpose. You must also uh consider the social benefit. the social benefit for [clears throat] the society, my reasoning, my action uh what uh what's the the point, what what the meaning for society that is uh those are can I say five uh characteristic of contextual selection. If you you have to to se to choose the context You have to see those four those five uh characteristics uh which can tells you tell you uh if your action or your reasoning is valuable for uh for scientific scientific papers or for human being. So I suppose that I Yeah. >> So thank you so much. So you you gave us a bullet point concerning all the steps and that when you say you must to do that, you must do that. The agent is the who is the agent? Is it the researcher? is a is a is some people from certain institution. So it's is could be plural. >> Who is the agent? >> The agent is everybody who knows >> okay okay >> something about AI. >> Okay. >> You must know something about AI. You must uh use the computer. You must use everything of from technology. Uh we have we have for now we have the phone, telephone, we have uh many applications. Uh so in the future I suppose that everyone will know to use uh those these uh devices is not not only the telephone. You you can have your drone, you can have everything else. But the problem is uh as agent you must you must I I say is in obligation because in the future we I suppose that my my talk was a little a little clear about that for the moment we are dealing dealing with individual the AI model for the moment is individual is the agent himself who can decide with of of course many skills, many ability decide to do or to to to speak or to reason. But for the future we will deal with society. There will be uh everybody will know about uh artificial intelligence. everybody. so to speak. For this, for the future, we have to consider not I can say in the theory of Rene Kart is the model of AI for now is the model of Rene Deart. with the subjectivism is the man I know. If I know something, I can act. If I have an ability, I can act as my reason, my rationality as I know. But the problem is you have to consider the society your action your reasoning must uh be in confrontation with the society. So the agent we for the future we will deal with many agents malt agents that is the the theory I propose with paramic intelligence >> thank you other questions comment >> actually I will >> yeah because it's super new for me so I'm curious Uh I believe I understood correctly. Uh you no more or less uh in your paper talk about proverbs right? You mentioned proverbs >> about proverbs. Yeah. Yeah. Yes. >> Yeah. And uh and you said that nonetheless should be u should be view as you said sophisticated mechanisms of contextual inferencing. Uh and my question is uh because I do have some example in Italian but I don't have maybe been English. Uh the pro proverbs could be um some proverbs could could be in opposition between one another. So some proverb could say something but other could go to another direction maybe the opposite direction. So how can you think you could manage? You you talked about this possible tension between uh proverbs or uh what what do you think about about that about that? >> Yes. Mhm. Um all my works are about uh about proverbs but uh I I propose I propose the term paria is the same paria or proverbs. Uh the difference is PMIA is um a transcription of the uh of the Greek term paria parameia. So uh all my works is about parroia paria or proverbs. But the proverbs in with the conception of um a logical reasoning not as a a fault or a linguistic artifact. >> Yeah. And uh uh the second um I can say for the confrontation we have the palava in Africa. This is a traditional culture of Africa. Uh in a palava we have to confront to have to uh to deal with to dialogue if you like to dialogue with someone who is using proverb in French. I can choose a proverb two proverbs. That is one proverb. But uh on the other side someone can say uh what that means? That means that uh if you propose a proverb, a parameia, you for reply, you have another parameia which can which can oppose your paria or which can uh follow your direction, can approve also your uh your paria. So that is a kind of a kind of a dialogue but the dialogue or par [snorts] but Aristotle talks talked about different kind kind of dialogue or parava. Uh we have a a parava agonistic parava that to say a confrontation is as in a war not you you cannot you in a war imagine a war but or a confrontation but you just use the proverb you use only the proverb so that if I I utter a proverb. You have to reply with your proverb uh as uh to confront my proverb. But there are also some kind of dialogue, some kind of palava which is uh irreic is I can say for Pacific Pacific purpose that to say we are looking for compromission not compromission for accord we are looking for to to be in the same in the same movement minger don't and if I propose a proverb I can approve your proverb to uh with my proverb which I can author also that is a kind of a kind of um um realization of a proverb in society. So this model can be uh used in artificial intelligence for the future. That to say if we have multi- agent you as agent in artificial intelligence you must also accept you must allow someone else to confront your action because proverb is not only a reasoning is a reasoning which imposed is is a reasoning of obligation. It what can I say? It it can give you an instruction to realize something to do to act to act with the proverb. So uh if you are you are an agent uh you must also accept another agent to confront your reasoning or your agent in artificial intelligence. That is the model which I propose for the future of the artificial intelligence and I I think that it may be the way which can u help to to find some laws or some rules uh to regulize the the use of artificial intelligence today because uh their lack of uh rules or for laws about artificial intelligence. Everyone everybody can do everything you like uh without beically orically condemned. But with uh the model I propose with paramic intelligence, we can find some rules, some norms, some laws to regulize for the future the use of artificial intelligence. >> Thank you. >> Okay. comment questions. Shall we thank our speaker? Oh, wait. Sorry. Sorry. I just saw online. My apologies. Um, >> one question. Oh, also uh also online. Uh, Peter Beth says he's sorry he can't be here. He had to go home, but he's been watching online. Um but as well there's a question that comes in online that says uh thank you for the presentation. Uh contemporary systems usually work with digital data. Uh do you think that a contextual permic AI system would also work with or be trained on similar kinds of data sets or would it require some kind of analog interface to work with the context? I I don't I don't hear. There was a problem with my communication. Can you Yeah. Yeah. >> Sorry. Can you hear me now? We Okay. >> Can you reply? Yeah. >> Okay. Uh the question says, uh, "Thanks for the presentation. Uh, contemporary systems usually work with digital data. Do you think that a contextual paric AI system would work on or be trained on similar kinds of data or will it require some kind of analog interface to to get to capture the context? Uh yes uh you know um the use of artificial intelligence. uh I can say for the moment that is a kind of artificial intelligence and the technology today everything is um stock is in the stock the stock what can I say uh what I mean is that we are constitute some data data for everything human can do. So we have a large a large we today we have we have the development of uh the progress of uh uh data cloud. Data cloud will will be the the the future problem characteristic of artificial intelligence. We have a large a large quantity of data for the moment and with the time we will have more than that more than that that to say in in history even for the constit constitution of data Everything of in in human life will be uh in stock storage stockage package in in data cloud or everything else. So with algorithm we will have the problem because uh AI will uh not only um uh stock magazine datas data but uh will um reason will operate with this data. And I think that even human being will be overwhelmed overwhelmed uh for this problem of information. So we have to to to uh to to propose also some rules or some laws about the the data the data for the moment because um you know you know you know that very well today artificial intelligence can can produce can yield everything of image. everything of the image, every kind of image uh every kind of uh production but uh it relies it's based on data and uh the problem we will deal with for the future will be uh that problem of selecting selecting the real case. The real data not only for uh scientific purpose but will deal with ethical ethical problem. political problem and uh society social social consideration problem. That is the problem of AI because for the moment AI is uh is progressing with uh enormously enormously and uh rapidly. So that is the problem of context selection and data and data and uh data model of the moment. We have also to reflect and to regularize the this problem the relation between uh with data cloud and the context the circum the context selection. So uh I suppose for my theory the pro the theory I propose I will um I will propose I will work on computational aspect to compute and to to propose some rules about this problem of inter intercon interconnection between data and context selection. I don't know if I reply on reply question. >> I'll see if I'll see if there's >> Is there another question online or not online? >> Yeah, >> sorry. >> Comments. >> Uh the last question was someone online so it will take time maybe for them to to reply if your answer satisfy them. We'll see. Oh, >> okay. Yeah. But for now I'm going to asking in the room online nothing thanks that helps. Okay so so the the person is satisfied by the answer thank you is there other questions? So I think we'll uh we'll thank our speaker. >> Thank you. >> Thank you very much. And it was the last whip of the year. And is it the last seminar of the year? >> Okay, >> everyone. So, >> thanks for being here. Yeah. See you soon. >> So, thank you. Thank you very much for listening, for your time and for your your desire to understand my theory. >> Thank you very much. Cheers. [music]