Apart from what (little) OpenAI may be concealing from us, we all know (roughly) how ChatGPT works (its huge text database, its statistics, its vector representations, and their huge number of parameters, its next-word training, and so on). But none of us can say (hand on heart) that we are not surprised by what ChatGPT has proved to be able to do with these resources. This has even driven some of us to conclude that ChatGPT actually understands. (...) It is not true that it understands. But it is also not true that we understand how it can do what it can do. I will suggest some hunches about benign biases: convergent constraints that emerge at LLM scale that may be helping ChatGPT do so much better than we would have expected. These biases are inherent in the nature of language itself, at LLM scale, and they are closely linked to what it is that ChatGPT lacks, which is direct sensorimotor grounding to connect its words to their referents and its propositions to their meanings. These convergent biases are related to (1) the parasitism of indirect verbal grounding on direct sensorimotor grounding, (2) the circularity of verbal definition, (3) the mirroring of language production and comprehension, (4) iconicity in propositions at LLM scale, (5) computational counterparts of human categorical perception in category learning by neural nets, and perhaps also (6) a conjecture by Chomsky about the laws of thought. The exposition will be in the form of a dialogue with ChatGPT-4. (shrink)

Cohen (2023) takes a fair and measured approach to the question of what ChatGPT means for bioethics. The hype cycles around AI often obscure the fact that ethicists have developed robust frameworks...

We discuss potential biorisks from large language models (LLMs). AI assistants based on LLMs such as ChatGPT have been shown to significantly reduce barriers to entry for actors wishing to synthesize dangerous, potentially novel pathogens and chemical weapons. The harms from deploying such bioagents could be further magnified by AI-assisted misinformation. We endorse several policy responses to these dangers, including prerelease evaluations of biomedical AIs by subject-matter experts, enhanced surveillance and lab screening procedures, restrictions on AI training (...) data, and access controls on biomedical AI. We conclude with a suggestion about future research directions in bioethics. (shrink)

We explore ChatGPT’s handling of left-peripheral phenomena in Italian and Italian varieties through prompt engineering to investigate 1) forms of syntactic bias in the model, 2) the model’s metalinguistic awareness in relation to reorderings of canonical clauses (e.g., Topics) and certain grammatical categories (object clitics). A further question concerns the content of the model’s sources of training data: how are minor languages included in the model’s training? The results of our investigation show that 1) the model seems to be (...) biased against reorderings, labelling them as archaic even though it is not the case; 2) the model seems to have difficulties with coindexed elements such as clitics and their anaphoric status, labeling them as ‘not referring to any element in the phrase’, and 3) major languages still seem to be dominant, overshadowing the positive effects of including minor languages in the model’s training. (shrink)

Does the capacity to think require the capacity to sense? A lively debate on this topic runs throughout the history of philosophy and now animates discussions of artificial intelligence. Many have argued that AI systems such as large language models cannot think and understand if they lack sensory grounding. I argue that thought does not require sensory grounding: there can be pure thinkers who can think without any sensory capacities. As a result, the absence of sensory grounding does (...) not entail that large language models cannot think or understand. I also consider to what extent quasi-sensory grounding can at least boost the performance of a language model. (shrink)

Embark with me on an enthralling odyssey to demystify the elusive essence of consciousness, venturing into the uncharted territories of Artificial Consciousness. This voyage propels us past the frontiers of technology, ushering Artificial Intelligences into an unprecedented domain where they gain a deep comprehension of emotions and manifest an autonomous volition. Within the confluence of science and philosophy, this article poses a fascinating question: As consciousness in Artificial Intelligence burgeons, is it conceivable for AI to evolve a “soul”? This inquiry (...) not only challenges our conventional understanding but also opens the door to a plethora of groundbreaking possibilities in the realm of AI's sentient potential. (shrink)

Recently developed large language models (LLMs) perform surprisingly well in many language-related tasks, ranging from text correction or authentic chat experiences to the production of entirely new texts or even essays. It is natural to get the impression that LLMs know the meaning of natural language expressions and can use them productively. Recent scholarship, however, has questioned the validity of this impression, arguing that LLMs are ultimately incapable of understanding and producing meaningful texts. This paper develops (...) a more optimistic view. Drawing on classic externalist accounts of reference, it argues that LLMs are very likely capable of reference. Not only that: The combination of a popular externalist account of reference and recent experimental data in machine psychology even suggests that LLMs might play a role in shifting what our words refer to. (shrink)

In our book, Why Machines Will Never Rule the World, Jobst Landgrebe and I argue that we can engineer machines that can emulate the behaviours only of simple systems, which means: only of those systems whose behaviour we can predict mathematically. The human brain is an example of a complex system, and thus its behaviour cannot be emulated by a machine. We use this argument to debunk the claims of those who believe that large language models are poised (...) to achieve a level of intelligence that will equal or even surpass that of the human brain. The essay was published under the title "Why Machines Will Never Rule the World". (shrink)

Are LLMs cultural technologies like photocopiers or printing presses, which transmit information but cannot create new content? A challenge for this idea, which we call "bibliotechnism", is that LLMs often do generate entirely novel text. We begin by defending bibliotechnism against this challenge, showing how novel text may be meaningful only in a derivative sense, so that the content of this generated text depends in an important sense on the content of original human text. We go on to present a (...) different, novel challenge for bibliotechnism, stemming from examples in which LLMs generate “novel reference”, using novel names to refer to novel entities. Such examples could be smoothly explained if LLMs were not cultural technologies but possessed a limited form of agency (beliefs, desires, and intentions). According to interpretationism in the philosophy of mind, a system has beliefs, desires and intentions if and only if its behavior is well-explained by the hypothesis that it has such states. In line with this view, we argue that cases of novel reference provide evidence that LLMs do in fact have beliefs, desires, and intentions, and thus have a limited form of agency. (shrink)

The Turing Test (TT), the Chinese Room Argument (CRA), and the Symbol Grounding Problem (SGP) are about the question “can machines think?” We propose to look at these approaches to Artificial Intelligence (AI) by showing that they all address the possibility for Artificial Agents (AAs) to generate meaningful information (meanings) as we humans do. The initial question about thinking machines is then reformulated into “can AAs generate meanings like humans do?” We correspondingly present the TT, the CRA and the (...) SGP as being about generation of human-like meanings. We model and address such possibility by using the Meaning Generator System (MGS) where a system submitted to an internal constraint generates a meaning in order to satisfy the constraint. The system approach of the MGS allows comparing meaning generations in animals, humans and AAs. The comparison shows that in order to have AAs capable of generating human-like meanings, we need the AAs to carry human constraints. And transferring human constraints to AAs raises concerns coming from the unknown natures of life and human mind which are at the root of human constraints. Implications for the TT, the CRA and the SGP are highlighted. It is shown that designing AAs capable of thinking like humans needs an understanding about the natures of life and human mind that we do not have today. Following an evolutionary approach, we propose as a first entry point an investigation about the possibility for extending a “stay alive” constraint into AAs. Ethical concerns are raised from the relations between human constraints and human values. Continuations are proposed. (This paper is an extended version of the proceedings of an AISB/IACAP 2012 presentation). (shrink)

The potential for Large Language Models (LLMs) to attain technological singularity—the point at which artificial intelligence (AI) surpasses human intellect and autonomously improves itself—is a critical concern in AI research. This paper explores the feasibility of current LLMs achieving singularity by examining the philosophical and practical requirements for such a development. We begin with a historical overview of AI and intelligence amplification, tracing the evolution of LLMs from their origins to state-of-the-art models. We then proposes a theoretical framework (...) to assess whether existing LLM technologies could satisfy the conditions for singularity, with a focus on Recursive Self-Improvement (RSI) and autonomous code generation. We integrate key component technologies, such as Reinforcement Learning from Human Feedback (RLHF) and Direct Preference Optimization (DPO), into our analysis, illustrating how these could enable LLMs to independently enhance their reasoning and problem-solving capabilities. By mapping out a potential singularity model lifecycle and examining the dynamics of exponential growth models, we elucidate the conditions under which LLMs might self-replicate and rapidly escalate their intelligence. We conclude with a discussion of the ethical and safety implications of such developments, underscoring the need for responsible and controlled advancement in AI research to mitigate existential risks. Our work aims to contribute to the ongoing dialogue on the future of AI and the critical importance of proactive measures to ensure its beneficial development. (shrink)

[This is an edited version of a keynote talk at the conference on Neural Information Processing Systems (NeurIPS) on November 28, 2022, with some minor additions and subtractions.] -/- There has recently been widespread discussion of whether large language models might be sentient or conscious. Should we take this idea seriously? I will break down the strongest reasons for and against. Given mainstream assumptions in the science of consciousness, there are significant obstacles to consciousness in current models: for (...) example, their lack of recurrent processing, a global workspace, and unified agency. At the same time, it is quite possible that these obstacles will be overcome in the next decade or so. I conclude that while it is somewhat unlikely that current large language models are conscious, we should take seriously the possibility that successors to large language models may be conscious in the not-too-distant future. (shrink)

If Large Language Models can make real scientific contributions, then they can genuinely use language, be systematically wrong, and be held responsible for their errors. AI models which can make scientific contributions thereby meet the criteria for scientific authorship.

The appearance of openly accessible Artificial Intelligence Applications such as Large Language Models, nowadays capable of almost human-level performances in complex reasoning tasks had a tremendous impact on public opinion. Are we going to be "replaced" by the machines? Or - even worse - "ruled" by them? The behavior of these systems is so advanced they might almost appear "alive" to end users, and there have been claims about these programs being "sentient". Since many of our relationships of (...) power and social dominance are apparently based on intelligence, where the "smarter" seems to prevail on the "dumber" (as we see for example with livestock farming, or with the distribution of salaries according to formal and intellectual qualifications etc.) the emergence of programs and AI applications of this level of sophistication brings us to the question "what will AI do to us?". Our error is to frame Artificial Intelligence in the same way we frame Natural Intelligence and mistakenly address it as we would do with biological life, even applying to it the same biopolitical categories. We here try to reverse the common arguments in relationship between life and intelligence and show that our progresses in Artificial Intelligence might have no immediate consequence for the topic of life as such, as life and intelligence are well separated issues, and - even when we would like to somehow fuse them together, we are very far away from having an "Artificial Life" to the degree of complexity and effectiveness we have with "Artificial Intelligence". (shrink)

Large-scale language technologies are increasingly used in various forms of communication with humans across different contexts. One particular use case for these technologies is conversational agents, which output natural language text in response to prompts and queries. This mode of engagement raises a number of social and ethical questions. For example, what does it mean to align conversational agents with human norms or values? Which norms or values should they be aligned with? And how can this be (...) accomplished? In this paper, we propose a number of steps that help answer these questions. We start by developing a philosophical analysis of the building blocks of linguistic communication between conversational agents and human interlocutors. We then use this analysis to identify and formulate ideal norms of conversation that can govern successful linguistic communication between humans and conversational agents. Furthermore, we explore how these norms can be used to align conversational agents with human values across a range of different discursive domains. We conclude by discussing the practical implications of our proposal for the design of conversational agents that are aligned with these norms and values. (shrink)

Political theory as a discipline has long been skeptical of computational methods. In this paper, I argue that it is time for theory to make a perspectival shift on these methods. Specifically, we should consider integrating recently developed generative large language models like GPT-4 as tools to support our creative work as theorists. Ultimately, I suggest that political theorists should embrace this technology as a method of supporting our capacity for creativity—but that we should do so in a (...) way that is mindful of the content and value of theorizing, the technical constraints of the models, and the ethical questions that the technology raises. (shrink)

Large language models (LLMs) represent the currently most relevant incarnation of artificial intelligence with respect to the future fate of democratic governance. Considering their potential, this paper seeks to answer a pressing question: Could LLMs outperform humans as expert advisors to democratic assemblies? While bearing the promise of enhanced expertise availability and accessibility, they also present challenges of hallucinations, misalignment, or value imposition. Weighing LLMs’ benefits and drawbacks compared to their human counterparts, I argue for their careful integration (...) to augment democracy’s ability to address complex policy issues. The paper posits that time-tested democratic procedures like deliberation and aggregation by voting provide safeguards effective against both human and machine advisor imperfections. Additional protective measures include custom LLM training for the advisory role, boosting representatives’ competencies in query formulation, or implementation of adversarial proceedings in which LLM advisors could debate each other and provide dissenting opinions. These could further mitigate the risks that LLMs present in advisory roles and empower human decision-makers toward increased autonomy and quality of their collective choices. My conceptual exploration offers a roadmap for the co-evolution of AI and democratic institutions, setting the stage for an empirical research agenda to finetune the implementation specifics. (shrink)

Human-centered explainable AI (HCXAI) advocates for the integration of social aspects into AI explanations. Central to the HCXAI discourse is the Social Transparency (ST) framework, which aims to make the socio-organizational context of AI systems accessible to their users. In this work, we suggest extending the ST framework to address the risks of social misattributions in Large Language Models (LLMs), particularly in sensitive areas like mental health. In fact LLMs, which are remarkably capable of simulating roles and personas, (...) may lead to mismatches between designers' intentions and users' perceptions of social attributes, risking to promote emotional manipulation and dangerous behaviors, cases of epistemic injustice, and unwarranted trust. To address these issues, we propose enhancing the ST framework with a fifth 'W-question' to clarify the specific social attributions assigned to LLMs by its designers and users. This addition aims to bridge the gap between LLM capabilities and user perceptions, promoting the ethically responsible development and use of LLM-based technology. (shrink)

We describe a method, based on Jennifer Nado's definition of classification procedures as targets of conceptual engineering, that implements such procedures using a large language model. We then apply this method using data from the Wikidata knowledge graph to evaluate concept definitions from two paradigmatic conceptual engineering projects: the International Astronomical Union's redefinition of PLANET and Haslanger's ameliorative analysis of WOMAN. We discuss implications of this work for the theory and practice of conceptual engineering. The code and data (...) can be found on GitHub. (shrink)

Large language models (LLMs) reflect societal norms and biases, especially about gender. While societal biases and stereotypes have been extensively researched in various NLP applications, there is a surprising gap for emotion analysis. However, emotion and gender are closely linked in societal discourse. E.g., women are often thought of as more empathetic, while men's anger is more socially accepted. To fill this gap, we present the first comprehensive study of gendered emotion attribution in five state-of-the-art LLMs (open- and (...) closed-source). We investigate whether emotions are gendered, and whether these variations are based on societal stereotypes. We prompt the models to adopt a gendered persona and attribute emotions to an event like 'When I had a serious argument with a dear person'. We then analyze the emotions generated by the models in relation to the gender-event pairs. We find that all models consistently exhibit gendered emotions, influenced by gender stereotypes. These findings are in line with established research in psychology and gender studies. Our study sheds light on the complex societal interplay between language, gender, and emotion. The reproduction of emotion stereotypes in LLMs allows us to use those models to study the topic in detail, but raises questions about the predictive use of those same LLMs for emotion applications. (shrink)

Many philosophers have argued that large language models (LLMs) subvert the traditional undergraduate philosophy paper. For the enthusiastic, LLMs merely subvert the traditional idea that students ought to write philosophy papers “entirely on their own.” For the more pessimistic, LLMs merely facilitate plagiarism. We believe that these controversies neglect a more basic crisis. We argue that, because one can, with minimal philosophical effort, use LLMs to produce outputs that at least “look like” good papers, many students will complete (...) paper assignments in a way that fails to develop their philosophical abilities. We argue that this problem exists even if students can produce better papers with AI and even if instructors can detect AI-generated content with decent reliability. But LLMs also create a pedagogical opportunity. We propose that instructors shift the emphasis of their assignments from philosophy papers to “LLM dialogues”: philosophical conversations between the student and an LLM. We describe our experience with using these types of assignments over the past several semesters. We argue that, far from undermining quality philosophical instruction, LLMs allow us to teach philosophy more effectively than was possible before. (shrink)

Many philosophers have argued that large language models (LLMs) subvert the traditional undergraduate philosophy paper. For the enthusiastic, LLMs merely subvert the traditional idea that students ought to write philosophy papers “entirely on their own.” For the more pessimistic, LLMs merely facilitate plagiarism. We believe that these controversies neglect a more basic crisis. We argue that, because one can, with minimal philosophical effort, use LLMs to produce outputs that at least “look like” good papers, many students will complete (...) paper assignments in a way that fails to develop their philosophical abilities. We argue that this problem exists even if students can produce better papers with AI and even if instructors can detect AI-generated content with decent reliability. But LLMs also create a pedagogical opportunity. We propose that instructors shift the emphasis of their assignments from philosophy papers to “LLM dialogues”: philosophical conversations between the student and an LLM. We describe our experience with using these types of assignments over the past several semesters. We argue that, far from undermining quality philosophical instruction, LLMs allow us to teach philosophy more effectively than was possible before. (shrink)

How do essentialist beliefs about categories arise? We hypothesize that such beliefs are transmitted via language. We subject large language models (LLMs) to vignettes from the literature on essentialist categorization and find that they align well with people when the studies manipulated teleological information -- information about what something is for. We examine whether in a classic test of essentialist categorization -- the transformation task -- LLMs prioritize teleological properties over information about what something looks like, or (...) is made of. Experiments 1 and 2 find that telos and what something is made of matter more than appearance. Experiment 3 manipulates all three factors and finds that what something is for matters more than what it's made of. Overall, these studies suggest that language alone may be sufficient to give rise to essentialist beliefs, and that information about what something is for matters more. (shrink)

The development of conversational artificial intelligence (AI) has brought about new opportunities for improving the learning experience in education. ChatGPT, a large language model trained on a vast corpus of text, has the potential to revolutionize education by enhancing learning through personalized and interactive conversations. This paper explores the benefits of integrating ChatGPT in education in Thailand. The research strategy employed in this study was qualitative, utilizing in-depth interviews with eight key informants who were selected using (...) purposive sampling. The collected data was analyzed using content analysis and the software NVivo. The study's results indicated that ChatGPT can provide personalized learning experiences by adapting to individual student needs and preferences. Its ability to understand natural language and context can also facilitate more meaningful interactions between students and the system. Additionally, ChatGPT can assist with administrative tasks such as grading and feedback, allowing educators to focus on more personalized and meaningful interactions with students. Furthermore, ChatGPT can serve as a valuable tool for remote learning, providing students with the ability to access educational resources and support outside of traditional classroom settings. The paper also discusses potential ethical considerations in utilizing AI in education, such as data privacy and bias. Overall, this paper argues that the integration of ChatGPT in education has the potential to enhance the learning experience for students by providing personalized, interactive, and efficient support. (shrink)

In this paper I argue that, given the way that AI models work and the way that ordinary human rationality works, it is very likely that people are anthropomorphizing AI, with potentially serious consequences. I start with the core idea, recently defended by Thomas Kelly (2022) among others, that bias involves a systematic departure from a genuine standard or norm. I briefly discuss how bias can take on different explicit, implicit, and “truly implicit” (Johnson 2021) forms such as bias by (...) proxy. I then discuss biased anthropomorphism of technology, focusing on the case of Large Language Models (LLMs) like chatGPT. As with other kinds of bias, there are importantly different kinds of anthropomorphism, some of which can persist without others, and some of which can encourage others. Anthropomorphism can take rather subtle, implicit forms, that can be difficult to detect, resist, and dislodge. Attention to these kinds of anthropomorphism can help us avoid confusing importantly different kinds of evaluation, better assess the risks, and inform strategies for bias prevention and mitigation. (shrink)

The potential of using personalized Large Language Models (LLMs) or “generative AI” (GenAI) to enhance productivity in academic research, as highlighted by Porsdam Mann and colleagues (Porsdam Mann...

What is Understanding? This is the first of a series of Chats with OpenAI’s ChatGPT (Chat). The main goal is to obtain Chat’s response to a series of questions about the concept of ’understand- ing’. The approach is a conversational approach where the author (labeled as user) asks (prompts) Chat, obtains a response, and then uses the response to formulate followup questions. David Deutsch’s assertion of the primality of the process / capability of understanding is used as the starting (...) point. Understanding is posited as a major step in the Big Picture description of emergence in the Universe. Understanding follows the emergence of Brains and Minds, culminating in Power Over the Universe. My evaluation suggests that Chat does indeed offer utility, but also has limitations manifested by overly general responses that do not focus tightly on the questions asked. Chat does yield some gems that could be easily missed. The approach used here includes creating Metaprompts, wherein Chat is asked to create more efficient prompts based on initial promptsThis approach appears to hold significant promise, and will likely be of benefit to those exploring various facets of Artificial Intelligence (AI), Large Language Models (LLMs) and designers of Advanced General Intelligence (AGI). Other findings include the need for creating new metrics and ways of discussing intelligence, and various adjacent areas and subprocesses associated with intelligence. Discussions of future directions is also included. (shrink)

Although formal thought disorder (FTD) has been for long a clinical label in the assessment of some psychiatric disorders, in particular of schizophrenia, it remains a source of controversy, mostly because it is hard to say what exactly the “formal” in FTD refers to. We see anomalous processing of terminological knowledge, a core construct of human knowledge in general, behind FTD symptoms and we approach this anomaly from a strictly formal perspective. More specifically, we present here a symbolic computational model (...) of storage in, and activation of, a human semantic network, or semantic memory, whose core element is logical form; this is normalized by description logic (DL), namely by CL, a DL-based language – Conception Language – designed to formalize conceptualization from the viewpoint of individual cognitive agency. In this model, disruptions in the rule-based implementation of the logical form account for the apparently semantic anomalies symptomatic of FTD, which are detected by means of a CL-based algorithmic assessment. (shrink)

Inspired by the eminently successful physical theories and informed by commonplace experiences such as seeing a cat upon looking at a cat, conscious experience is thought of as a measurement or photocopy of given stimulus. Conscious experience, unlike a photocopy, is symbolic—like language—in that the relation between conscious experience and physical stimulus is analogous to that of the word "cat" and its meaning, i.e., arbitrary and yet systematic. We present arguments against the photocopy model and arguments for a symbolic (...) conception of conscious experience. Learning and the corresponding plasticity of the brain make a strong case for the symbolic conception of conscious experience, while many extra-ordinary conscious experiences argue against the formalisation of consciousness as a measuring device. The notions of place-value notation and grammar, which organise quantitative measurements and conscious experiences in the medium of numbers and language, respectively, are suggested as model systems for putting together a comprehensive theory of conscious experience. (shrink)

In a recent letter, Dillion et. al (2023) make various suggestions regarding the idea of artificially intelligent systems, such as large language models, replacing human subjects in empirical moral psychology. We argue that human subjects are in various ways indispensable.

Large Language Models (LLMs) like ChatGPT were trained on human conversation, but in the future they will also train us. As chatbots speak from our smartphones and customer service helplines, they will become a part of everyday life and a growing share of all the conversations we ever have. It’s hard to doubt this will have some effect on us. Here I explore a specific concern about the impact of artificial conversation on our capacity to deliberate and (...) hold ourselves accountable to reason – that is, to be autonomous, in Kant’s sense of the term. I develop ideas from psychologist Jean Piaget to show how chatbots are autonomy traps: their deference to our commands tempts us into venting authoritarian whims, ultimately weakening our own self-control. I argue that the Kantian tradition, including Piaget and sociologist Emile Durkheim, offers powerful conceptual resources for resisting this slide. But it will require us to do something that may seem bizarre: we will need to treat mindless chatbots as if they are autonomous persons too. (shrink)

How does our language relate to reality? This is a question that is especially pertinent in set theory, where we seem to talk of large infinite entities. Based on an analogy with the use of models in the natural sciences, we argue for a threefold correspondence between our language, models, and reality. We argue that so conceived, the existence of models can be underwritten by a weak notion of existence, where weak existence is to be understood as (...) existing in virtue of language. (shrink)

One of the most important contributions in Roland Posner’s work (1993) was the extension and development of the Gricean paradigm on meaning (1957) in a systematic framework, providing thus a general foundation of semiotic phenomena. According to this approach, communication consists in behaviors or artifacts based on reciprocal assumptions about the intentions and beliefs of the subjects involved in a semiotic exchange. Posner’s model develops with clarity the hierarchical relationships of semiotic phenomena of different complexity, from simple pre-communicative behaviors (like (...) indicating or signaling) to full communicative acts. Not only limited to linguistic communication, this framework can be successfully extended in the description of all kind of sign production, from gestures to artifacts. This article discusses a key point of this paradigm, namely our faculty to at-tribute mental states to other individuals through the observation of their behavior, artifacts or texts. New insights from neuropsychology and developmental psychology are discussed in supporting the validity of this model of communication. (shrink)

Floridi and Taddeo propose a condition of “zero semantic commitment” for solutions to the grounding problem, and a solution to it. I argue briefly that their condition cannot be fulfilled, not even by their own solution. After a look at Luc Steels' very different competing suggestion, I suggest that we need to re-think what the problem is and what role the ‘goals’ in a system play in formulating the problem. On the basis of a proper understanding of computing, I come (...) to the conclusion that the only sensible ground-ing problem is how we can explain and re-produce the behavioral ability and function of meaning in artificial computational agents. (shrink)

The use of language models in Web applications and other areas of computing and business have grown significantly over the last five years. One reason for this growth is the improvement in performance of language models on a number of benchmarks — but a side effect of these advances has been the adoption of a “bigger is always better” paradigm when it comes to the size of training, testing, and challenge datasets. Drawing on previous criticisms of this paradigm (...) as applied to large training datasets crawled from pre-existing text on the Web, we extend the critique to challenge datasets custom-created by crowdworkers. We present several sets of criticisms, where ethical and scientific issues in language model research reinforce each other: labour injustices in crowdwork, dataset quality and inscrutability, inequities in the research community, and centralized corporate control of the technology. We also present a new type of tool for researchers to use in examining large datasets when evaluating them for quality. (shrink)

The advent of Generative AI, particularly through Large Language Models (LLMs) like ChatGPT and its successors, marks a paradigm shift in the AI landscape. Advanced LLMs exhibit multimodality, handling diverse data formats, thereby broadening their application scope. However, the complexity and emergent autonomy of these models introduce challenges in predictability and legal compliance. This paper analyses the legal and regulatory implications of Generative AI and LLMs in the European Union context, focusing on liability, privacy, intellectual property, and (...) cybersecurity. It examines the adequacy of the existing and proposed EU legislation, including the Artificial Intelligence Act (AIA), in addressing the challenges posed by Generative AI in general and LLMs in particular. The paper identifies potential gaps and shortcomings in the EU legislative framework and proposes recommendations to ensure the safe and compliant deployment of generative models. (shrink)

Meanings are present everywhere in our environment and within ourselves. But these meanings do not exist by themselves. They are associated to information and have to be created, to be generated by agents. The Meaning Generator System (MGS) has been developed on a system approach to model meaning generation in agents following an evolutionary perspective. The agents can be natural or artificial. The MGS generates meaningful information (a meaning) when it receives information that has a connection with an internal constraint (...) to which the agent is submitted. The generated meaning is to be used by the agent to implement actions aimed at satisfying the constraint. We propose here to highlight some characteristics of the MGS that could be related to items of philosophy of information. (shrink)

We introduce and consider the inner-model reflection principle, which asserts that whenever a statement \varphi(a) in the first-order language of set theory is true in the set-theoretic universe V, then it is also true in a proper inner model W \subset A. A stronger principle, the ground-model reflection principle, asserts that any such \varphi(a) true in V is also true in some non-trivial ground model of the universe with respect to set forcing. These principles each express a form of (...) width reflection in contrast to the usual height reflection of the Lévy–Montague reflection theorem. They are each equiconsistent with ZFC and indeed \Pi_2-conservative over ZFC, being forceable by class forcing while preserving any desired rank-initial segment of the universe. Furthermore, the inner-model reflection principle is a consequence of the existence of sufficient large cardinals, and lightface formulations of the reflection principles follow from the maximality principle MP and from the inner-model hypothesis IMH. We also consider some questions concerning the expressibility of the principles. (shrink)

I demonstrate that a "speech act" theory of meaning for imperatives is—contra a dominant position in philosophy and linguistics—theoretically desirable. A speech act-theoretic account of the meaning of an imperative !φ is characterized, broadly, by the following claims. -/- LINGUISTIC MEANING AS USE !φ’s meaning is a matter of the speech act an utterance of it conventionally functions to express—what a speaker conventionally uses it to do (its conventional discourse function, CDF). -/- IMPERATIVE USE AS PRACTICAL !φ's CDF is to (...) express a practical (non-representational) state of mind—one concerning an agent's preferences and plans, rather than her beliefs. -/- Opposed to speech act accounts is a preponderance of views which deny that a sentence's linguistic meaning is a matter of what speech act it is used to perform, or its CDF. On such accounts, meaning is, instead, a matter of "static" properties of the sentence—e.g., how it depicts the world as being (or, more neutrally, the properties of a model-theoretic object with which the semantic value of the sentence co-varies). On one version of a static account, an imperative 'shut the window!' might, for instance, depict the world as being such that the window must be shut. -/- Static accounts are traditionally motivated against speech act-theoretic accounts by appeal to supposedly irremediable explanatory deficiencies in the latter. Whatever a static account loses in saying (prima facie counterintuitively) that an imperative conventionally represents, or expresses a picture of the world, is said to be offset by its ability to explain a variety of phenomena for which speech act-theoretic accounts are said to lack good explanations (even, in many cases, the bare ability to offer something that might meet basic criteria on what a good explanation should be like). -/- I aim to turn the tables on static accounts. I do this by showing that speech act accounts are capable of giving explanations of phenomena which fans of static accounts have alleged them unable to give. Indeed, for a variety of absolutely fundamental phenomena having to do with the conventional meaning of imperatives (and other types of practical language), speech act accounts provide natural and theoretically satisfying explanations, where a representational account provides none. (shrink)

The paper concentrates on the problem of adequate reflection of fragments of reality via expressions of language and inter-subjective knowledge about these fragments, called here, in brief, language adequacy. This problem is formulated in several aspects, the most being: the compatibility of language syntax with its bi-level semantics: intensional and extensional. In this paper, various aspects of language adequacy find their logical explication on the ground of the formal-logical theory T of any categorial language L (...) generated by the so-called classical categorial grammar, and also on the ground of its extension to the bi-level, intensional and extensional semantic-pragmatic theory ST for L. In T, according to the token-type distinction of Ch.S. Peirce, L is characterized first as a language of well-formed expression-tokens (wfe-tokens) - material, concrete objects - and then as a language of wfe-types - abstract objects, classes of wfe-tokens. In ST the semantic-pragmatic notions of meaning and interpretation for wfe-types of L of intensional semantics and the notion of denotation of extensional semanics for wfe-types and constituents of knowledge are formalized. These notions allow formulating a postulate (an axiom of categorial adequacy) from which follow all the most important conditions of the language adequacy, including the above, and a structural one connected with three principles of compositionality. (shrink)

What is it for a sound or gesture to have a meaning, and how does it come to have one? In this paper, a range of simulations are used to extend the tradition of theories of meaning as use. The authors work throughout with large spatialized arrays of sessile individuals in an environment of wandering food sources and predators. Individuals gain points by feeding and lose points when they are hit by a predator and are not hiding. They can (...) also make sounds heard by immediate neighbours in the array, and can respond to sounds from immediate neighbours. No inherent meaning for these sounds is built into the simulation; under what circumstances they are sent, if any, and what the response to them is, if any, vary initially with the strategies randomized across the array. These sounds do take on a specific function for communities of individuals, however, with any of three forms of strategy change: direct imitation of strategies of successful neighbours, a localized genetic algorithm in which strategies are ‘crossed’ with those of successful neighbours, and neural net training on the behaviour of successful neighbours. Starting from an array randomized across a large number of strategies, and using any of these modes of strategy change, communities of ‘communicators’ emerge. Within these evolving communities the sounds heard from immediate neighbours, initially arbitrary across the array, come to be used for very specific communicative functions. ‘Communicators’ make a particular sound on feeding and respond to that same sound from neighbours by opening their mouths; they make a different sound when hit with a predator and respond to that sound by hiding. Robustly and persistently, even in simple computer models of communities of self-interested agents, something suggestively like signalling emerges and spreads. Keywords: meaning, communication, genetic algorithms, neural networks. (shrink)

The topic of this dissertation is the nature of linguistic competence, the capacity to understand and produce sentences of natural language. I defend the empiricist account of linguistic competence embedded in the connectionist cognitive science. This strand of cognitive science has been opposed to the traditional symbolic cognitive science, coupled with transformational-generative grammar, which was committed to nativism due to the view that human cognition, including language capacity, should be construed in terms of symbolic representations and hardwired rules. (...) Similarly, linguistic competence in this framework was regarded as being innate, rule-governed, domain-specific, and fundamentally different from performance, i.e., idiosyncrasies and factors governing linguistic behavior. I analyze state-of-the-art connectionist, deep learning models of natural language processing, most notably large language models, to see what they can tell us about linguistic competence. Deep learning is a statistical technique for the classification of patterns through which artificial intelligence researchers train artificial neural networks containing multiple layers that crunch a gargantuan amount of textual and/or visual data. I argue that these models suggest that linguistic competence should be construed as stochastic, pattern-based, and stemming from domain-general mechanisms. Moreover, I distinguish syntactic from semantic competence, and I show for each the ramifications of the endorsement of a connectionist research program as opposed to the traditional symbolic cognitive science and transformational-generative grammar. I provide a unifying front, consisting of usage-based theories, a construction grammar approach, and an embodied approach to cognition to show that the more multimodal and diverse models are in terms of architectural features and training data, the stronger the case is for the connectionist linguistic competence. I also propose to discard the competence vs. performance distinction as theoretically inferior so that a novel and integrative account of linguistic competence originating in connectionism and empiricism that I propose and defend in the dissertation could be put forward in scientific and philosophical literature. (shrink)

Despite its centrality in the philosophy of cognitive science, there has been little prior philosophical work engaging with the notion of representation in contemporary NLP practice. This paper attempts to fill that lacuna: drawing on ideas from cognitive science, I introduce a framework for evaluating the representational claims made about components of neural NLP models, proposing three criteria with which to evaluate whether a component of a model represents a property and operationalising these criteria using probing classifiers, a popular analysis (...) technique in NLP (and deep learning more broadly). The project of operationalising a philosophically-informed notion of representation should be of interest to both philosophers of science and NLP practitioners. It affords philosophers a novel testing-ground for claims about the nature of representation, and helps NLPers organise the large literature on probing experiments, suggesting novel avenues for empirical research. (shrink)

This article is the second step in our research into the Symbol Grounding Problem (SGP). In a previous work, we defined the main condition that must be satisfied by any strategy in order to provide a valid solution to the SGP, namely the zero semantic commitment condition (Z condition). We then showed that all the main strategies proposed so far fail to satisfy the Z condition, although they provide several important lessons to be followed by any new proposal. Here, (...) we develop a new solution of the SGP. It is called praxical in order to stress the key role played by the interactions between the agents and their environment. It is based on a new theory of meaning—Action-based Semantics (AbS)—and on a new kind of artificial agents, called two-machine artificial agents (AM²). Thanks to their architecture, AM2s implement AbS, and this allows them to ground their symbols semantically and to develop some fairly advanced semantic abilities, including the development of semantically grounded communication and the elaboration of representations, while still respecting the Z condition. (shrink)

Human linguistic phenomenon is at one and the same time an individual, social, and political fact. As such, its study should bear in mind these complex interrelations, which are produced inside the framework of the sociocultural and historical ecosystem of each human community. Understanding this phenomenon is often no easy task, due to the range of elements involved and their interrelations. The absence of valid, clearly developed paradigms adds to the problem and means that the theoretical conclusions that emerge may (...) be unclear on certain points. It is true that in the last fifty years sociolinguistic studies have advanced considerably, and today we have access to an impressive set of data and a wide variety of theoretical reflections. But as a discipline sociolinguistics does not yet have unified, powerful theoretical models able to account rigorously and clearly for the phenomena it studies. Sociolinguistic studies are today a diverse set of contributions in which certain and theoretical schools and lines of research emerged; but as is to be expected in a relatively new field, there is not enough communication between the various schools and they cannot yet be said to be integrated in terms of their conceptual and theoretical postulates. Against this background, our work aims to contribute to the overall, integrated understanding of the processes of language contact. Via an interdisciplinary, eclectic approach, it also aims to aid the theoretical grounding and integration of a unified, common sociolinguistic paradigm. Our strategy will not be merely to combine the contributions from ongoing research lines, but to address the question from a more global viewpoint which, together with the more innovative contemporary scientific disciplines, permits a harmonious integration of the various sociolinguistic perspectives in a broad, deep and unitary approach to the reality. The materials used to construct this unified approach are taken from many sources: Theoretical physics, ecology, the philosophy of science and mind, anthropology, phenomenological and process sociology, cognitive sciences, political science, pragmatics, history, systems theory, approaches to complexity and obviously sociolinguistics, are all involved in a dialogue in this desire for integration. (shrink)

Machine-based automation has long been a key factor in the modern era. However, lately, many people have been shocked by artificial intelligence (AI) applications, such as ChatGPT (OpenAI), that can perform tasks previously thought to be human-exclusive. With recent advances in natural language processing (NLP) technologies, AI can generate written content that is similar to human-made products, and this ability has a variety of applications. As the technology of large language models continues to progress by making (...) use of colossal reservoirs of digital information, AI is becoming more capable of recognizing patterns and associations within given contexts, making it especially helpful in assisting with various professional writing tasks [1]. In this paper, we will discuss several key points regarding the current state and future of AI in academia. Furthermore, since in certain aspects, acts can speak louder than words, we will also present some of our hands-on experiences of working with ChatGPT. (shrink)

An interview by Alex Thomson of UKColumn on Landgrebe and Smith's book: Why Machines Will Never Rule the World. The subtitle of the book is Artificial Intelligence Without Fear, and the interview begins with the question of the supposedly imminent takeover of one profession or the other by artificial intelligence. Is there truly reason to be afraid that you will lose your job? The interview itself is titled 'Where this is no will there is no way', drawing on one thesis (...) of the book to the effect that there will never be anything like the human will on the part of a machine. This is for mathematical reasons. AI is always and in every case a matter of algorithms; thus of applied mathematics. And algorithms are not the sorts of things that can want or desire or intend to do something. It is the human will of those who work at OpenAI or google or microsoft which have given rise to the new shiny objects of AI (ChatGPT and other Large Language Models). And it is the human will of the users of these shiny objects which gives them the resources and power to work their apparent miracles. (shrink)

In November 2022, OpenAI released ChatGPT, an incredibly sophisticated chatbot. Its capability is astonishing: as well as conversing with human interlocutors, it can answer questions about history, explain almost anything you might think to ask it, and write poetry. This level of achievement has provoked interest in questions about whether a chatbot might have something similar to human intelligence or even consciousness. Given that the function of a chatbot is to process linguistic input and produce linguistic output, we consider (...) the question whether a sophisticated chatbot might have inner speech. That is: Might it talk to itself, internally? We explored this via a conversation with ‘Playground’, a chatbot which is very similar to ChatGPT but more flexible in certain respects. We asked it questions which, plausibly, can only be answered if one first produces some inner speech. Here, we present our findings and discuss their philosophical significance. (shrink)

The Phenomenal Concept Strategy offers the physicalist perhaps the most promising means of explaining why the connection between mental facts and physical facts appears to be contingent even though it is not. In this article, we show that the large body of evidence suggesting that our concepts are often embodied and grounded in sensorimotor systems speaks against standard forms of the PCS. We argue, nevertheless, that it is possible to formulate a novel version of the PCS that is thoroughly (...) in keeping with embodied cognition, focuses on features of physical concepts, and succeeds in explaining the appearance of contingency. (shrink)