Do large language models like Chat-GPT or Claude meaningfully use the words they produce? Or are they merely clever prediction machines, simulating language use by producing statistically plausible text? There have already been some initial attempts to answer this question by showing that these models meet the criteria for entering meaningful states according to metasemantic theories of mental content. In this paper, I will argue for a different approach—that we should instead consider whether language models meet the (...) criteria given by our best metasemantic theories of linguistic content. In that vein, I will illustrate how this can be done by applying two such theories to the case of language models: Gareth Evans’ ( 1982 ) account of naming practices and Ruth Millikan’s ( 1984, 2004, 2005 ) teleosemantics. In doing so, I will argue that it is a mistake to think that the failure of LLMs to meet plausible conditions for mental intentionality thereby renders their outputs meaningless, and that a distinguishing feature of linguistic intentionality—dependency on a pre-existing linguistic system—allows for the plausible result that LLM outputs are meaningful. (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)

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 generate novel text. We begin with a defense of bibliotechnism, showing how even novel text may inherit its meaning from original human-generated text. We then argue that bibliotechnism faces an independent challenge from examples in which LLMs generate novel reference, using new names to refer to new entities. Such examples could be (...) explained if LLMs were not cultural technologies but had beliefs, desires, and intentions. According to interpretationism in the philosophy of mind, a system has such attitudes if and only if its behavior is well explained by the hypothesis that it does. Interpretationists may hold that LLMs have attitudes, and thus have a simple solution to the novel reference problem. We emphasize, however, that interpretationism is compatible with very simple creatures having attitudes and differs sharply from views that presuppose these attitudes require onsciousness, sentience, or intelligence (topics about which we make no claims). (shrink)

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)

Large Language Models (LLMs) have been transformative. They are pre-trained foundational models that are self-supervised and can be adapted with fine tuning to a wide range of natural language tasks, each of which previously would have required a separate network model. This is one step closer to the extraordinary versatility of human language. GPT-3 and more recently LaMDA can carry on dialogs with humans on many topics after minimal priming with a few examples. However, there has (...) been a wide range of reactions and debate on whether these LLMs understand what they are saying or exhibit signs of intelligence. This high variance is exhibited in three interviews with LLMs reaching wildly different conclusions. A new possibility was uncovered that could explain this divergence. What appears to be intelligence in LLMs may in fact be a mirror that reflects the intelligence of the interviewer, a remarkable twist that could be considered a Reverse Turing Test. If so, then by studying interviews we may be learning more about the intelligence and beliefs of the interviewer than the intelligence of the LLMs. As LLMs become more capable they may transform the way we interact with machines and how they interact with each other. Increasingly, LLMs are being coupled with sensorimotor devices. LLMs can talk the talk, but can they walk the walk? A road map for achieving artificial general autonomy is outlined with seven major improvements inspired by brain systems. LLMs could be used to uncover new insights into brain function by downloading brain data during natural behaviors. (shrink)

To what degree can language be acquired from linguistic input alone? This question has vexed scholars for millennia and is still a major focus of debate in the cognitive science of language. The complexity of human language has hampered progress because studies of language–especially those involving computational modeling–have only been able to deal with small fragments of our linguistic skills. We suggest that the most recent generation of Large Language Models (LLMs) might finally provide the (...) computational tools to determine empirically how much of the human language ability can be acquired from linguistic experience. LLMs are sophisticated deep learning architectures trained on vast amounts of natural language data, enabling them to perform an impressive range of linguistic tasks. We argue that, despite their clear semantic and pragmatic limitations, LLMs have already demonstrated that human-like grammatical language can be acquired without the need for a built-in grammar. Thus, while there is still much to learn about how humans acquire and use language, LLMs provide full-fledged computational models for cognitive scientists to empirically evaluate just how far statistical learning might take us in explaining the full complexity of human language. (shrink)

Current work in language models (LMs) helps us speed up or even skip thinking by accelerating and automating cognitive work. But can LMs help us with critical thinking -- thinking in deeper, more reflective ways which challenge assumptions, clarify ideas, and engineer new concepts? We treat philosophy as a case study in critical thinking, and interview 21 professional philosophers about how they engage in critical thinking and on their experiences with LMs. We find that philosophers do not find LMs (...) to be useful because they lack a sense of selfhood (memory, beliefs, consistency) and initiative (curiosity, proactivity). We propose the selfhood-initiative model for critical thinking tools to characterize this gap. Using the model, we formulate three roles LMs could play as critical thinking tools: the Interlocutor, the Monitor, and the Respondent. We hope that our work inspires LM researchers to further develop LMs as critical thinking tools and philosophers and other 'critical thinkers' to imagine intellectually substantive uses of LMs. (shrink)

This essay explores the possible contribution of Large Language Models (LLMs) to human cognition. It investigates whether human cognition can be enhanced by advanced AI systems such as LLMs. Can LLMs make people as learners smarter, or, on the contrary, make them reason/think less? The author discusses the concepts of human and artificial intelligence and examines LLMs as advanced AI systems, which use deep learning techniques and can be considered as excelling in neural network architectures, data volume, generalisation and (...) scalability. The author suggests that while LLMs can assist in facilitating numerous cognitive tasks, more research and philosophical inquiry is needed to understand whether such kind of AI assistance would make people cultivate human intelligence more, and not less. Presumably, Large Language Models (LLMs) can contribute to human intelligence and cognition just under strict (addressed existing limitations, questioning prompting, time-sensitivity, etc.) conditions. However, it is important that these theoretical considerations could be verified by experimental research. (shrink)

Newly powerful large language models have burst onto the scene, with applications across a wide range of functions. We can now expect to encounter their outputs at rapidly increasing volumes and frequencies. Some commentators claim that large language models are bullshitting, generating convincing output without regard for the truth. If correct, that would make large language models distinctively dangerous discourse participants. Bullshitters not only undermine the norm of truthfulness (by saying false things) but the normative status of (...) truth itself (by treating it as entirely irrelevant). So, do large language models really bullshit? I argue that they can, in the sense of issuing propositional content in response to fact-seeking prompts, without having first assessed that content for truth or falsity. However, I further argue that they need not bullshit, given appropriate guardrails. So, just as with human speakers, the propensity for a large language model to bullshit depends on its own particular make-up. (shrink)

Large language models (LLMs) have now entered the realm of medical ethics. In a recent study, Balaset alexamined the performance of GPT-4, a commercially available LLM, assessing its performance in generating responses to diverse medical ethics cases. Their findings reveal that GPT-4 demonstrates an ability to identify and articulate complex medical ethical issues, although its proficiency in encoding the depth of real-world ethical dilemmas remains an avenue for improvement. Investigating the integration of LLMs into medical ethics decision-making appears to (...) be an interesting avenue of research. However, despite the promising trajectory of LLM technology in medicine, it is crucial to exercise caution and refrain from attributing their expertise to medical ethics. Our thesis follows an examination of the nature of expertise and the epistemic limitations that affect LLM technology. As a result, we propose two more fitting applications of LLMs in medical ethics: first, as tools for mining electronic health records or scientific literature, thereby supplementing evidence for resolving medical ethics cases, and second, as educational platforms to foster ethical reflection and critical thinking skills among students and residents. The integration of LLMs in medical ethics, while promising, requires careful consideration of their epistemic limitations. Consequently, a well-considered definition of their role in ethically sensitive decision-making is crucial. (shrink)

Artificial intelligence chatbots driven by large language models (LLMs) have the potential to increase public science literacy and support scientific research, as they can quickly summarize complex scientific information in accessible terms. However, when summarizing scientific texts, LLMs may omit details that limit the scope of research conclusions, leading to generalizations of results broader than warranted by the original study. We tested 10 prominent LLMs, including ChatGPT-4o, ChatGPT-4.5, DeepSeek, LLaMA 3.3 70B, and Claude 3.7 Sonnet, comparing 4900 LLM-generated summaries (...) to their original scientific texts. Even when explicitly prompted for accuracy, most LLMs produced broader generalizations of scientific results than those in the original texts, with DeepSeek, ChatGPT-4o, and LLaMA 3.3 70B overgeneralizing in 26–73% of cases. In a direct comparison of LLM-generated and human-authored science summaries, LLM summaries were nearly five times more likely to contain broad generalizations (OR = 4.85, 95% CI [3.06, 7.70], p < 0.001). Notably, newer models tended to perform worse in generalization accuracy than earlier ones. Our results indicate a strong bias in many widely used LLMs towards overgeneralizing scientific conclusions, posing a significant risk of large-scale misinterpretations of research findings. We highlight potential mitigation strategies, including lowering LLM temperature settings and benchmarking LLMs for generalization accuracy. (shrink)

Large language models (LLMs) such as ChatGPT or Google’s Bard have shown significant performance on a variety of text-based tasks, such as summarization, translation, and even the generation of new...

Which is better at doing medical ethics: conversational artificial intelligence bots like ChatGPT or tools based on fuzzy cognitive maps? The article compares the performance of chatbots that rely on large language models to that of our own METHAD algorithm. While both tools approach dilemmas in medical ethics through the lens of Beauchamp and Childress’ mid-level principles, ChatGPT and METHAD differ considerably in the format of their inputs and outputs, in their interpretability, and in the kinds of mistakes that (...) they make. An ideal advisory algorithm would combine their characteristics. (shrink)

This article sets in with the question whether current or foreseeable transformer-based large language models (LLMs), such as the ones powering OpenAI’s ChatGPT, could be language users in a way comparable to humans. It answers the question negatively, presenting the following argument. Apart from niche uses, to use language means to act. But LLMs are unable to act because they lack intentions. This, in turn, is because they are the wrong kind of being: agents with intentions need (...) to be autonomous organisms while LLMs are heteronomous mechanisms. To conclude, the article argues, based on structural aspects of transformer-based LLMs, that these LLMs have taken a first step away from mechanistic artificiality to autonomous self-constitution, which means that these models are (slowly) moving into a direction that someday might result in non-human, but equally non-artificial agents, thus subverting the time-honored Kantian distinction between organism and mechanism. (shrink)

In some recent work, some psychologists, linguists, AI researchers, and philosophers (e.g., Shanahan, 2022; Bender & Koller, 2020; Montemayor, 2021; Bender et al., 2021) have argued that, despite producing convincing human-like linguistic output, large language models do not possess linguistic competence on the ground that they lack communicative intention. Among the proponents of this position, the notion of communicative intention is entertained as the liveliest candidate for a distinguishing characteristic of human cognition vis-a-vis LLMs and as such as at (...) least one of the features which accounts for the characteristic mentality of humans. I demonstrate that such a position is largely driven by strong Gricean assumptions about the nature of linguistic competence that I argue is now known to be empirically untenable. Moreover, I show that once we give up these assumptions and fairly assess the architecture of both LLM and human cognitive architectures, we will find no good reasons to suppose large language models lack communicative intentions. I thus reject communicative intentions as a viable ground on which to understand the difference between LLMs and humans and thus as a characteristic mark of human mentality. (shrink)

Large Language Models (LLMs) could be a useful tool for lawyers. However, empirical research on their effectiveness in conducting legal tasks is scant. We study securities cases involving cryptocurrencies as one of numerous contexts where AI could support the legal process, studying GPT-3.5’s legal reasoning and ChatGPT’s legal drafting capabilities. We examine whether a) GPT-3.5 can accurately determine which laws are potentially being violated from a fact pattern, and b) whether there is a difference in juror decision-making based on (...) complaints written by a lawyer compared to ChatGPT. We feed fact patterns from real-life cases to GPT-3.5 and evaluate its ability to determine correct potential violations from the scenario and exclude spurious violations. Second, we had mock jurors assess complaints written by ChatGPT and lawyers. GPT-3.5’s legal reasoning skills proved weak, though we expect improvement in future models, particularly given the violations it suggested tended to be correct (it merely missed additional, correct violations). ChatGPT performed better at legal drafting, and jurors’ decisions were not statistically significantly associated with the author of the document upon which they based their decisions. Because GPT-3.5 cannot satisfactorily conduct legal reasoning tasks, it would be unlikely to be able to help lawyers in a meaningful way at this stage. However, ChatGPT’s drafting skills (though, perhaps, still inferior to lawyers) could assist lawyers in providing legal services. Our research is the first to systematically study an LLM’s legal drafting and reasoning capabilities in litigation, as well as in securities law and cryptocurrency-related misconduct. (shrink)

Can large language models produce expert‐quality philosophical texts? To investigate this, we fine‐tuned GPT‐3 with the works of philosopher Daniel Dennett. To evaluate the model, we asked the real Dennett 10 philosophical questions and then posed the same questions to the language model, collecting four responses for each question without cherry‐picking. Experts on Dennett's work succeeded at distinguishing the Dennett‐generated and machine‐generated answers above chance but substantially short of our expectations. Philosophy blog readers performed similarly to (...) the experts, while ordinary research participants were near chance distinguishing GPT‐3's responses from those of an “actual human philosopher”. (shrink)

As large language models (LLMs) such as ChatGPT, Claude, Gemini, and Perplexity become increasingly ubiquitous as both tools and objects of scientific study, in addition to their established roles as chatbots, text generators and translators, questions about their identity conditions become scientifically as well as philosophically and socially important. This paper is about how to count language models. We argue that much of the emerging literature on these systems presupposes an answer to the question of identity for these (...) AIs but lacks a framework for counting or individuating them. We propose four levels of analysis, inspired by an analogy with software as well as insights from social epistemology. This account produces practical consequences for both research on language models, the ethics of AI, and our everyday talk about them. (shrink)

A growing chorus of AI researchers and philosophers posit internal representations in large language models (LLMs). But how do these representations relate to the kinds of mental states we routinely ascribe to our fellow humans? While some research has focused on belief- or knowledge- like states in LLMs, there has been comparatively little focus on the question of whether LLMs have intentions. I survey five properties that have been associated with intentions in the philosophical literature, and assess two candidate (...) classes of LLM representations against this set of features. The result is mixed: LLMs have representations that are intention-like in many— perhaps surprising—respects, but they differ from human intentions in important ways. (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.

[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)

Recent work by Chatzi et al. and Ravfogel et al. has developed, for the first time, a method for generating counterfactuals of probabilistic Large Language Models. Such counterfactuals tell us what would - or might - have been the output of an LLM if some factual prompt x had been x* instead. The ability to generate such counterfactuals is an important necessary step towards explaining, evaluating, and eventually improving, the behavior of LLMs. I argue, however, that the existing method (...) rests on an ambiguous interpretation of LLMs: it does not interpret LLMs literally, for the method involves the assumption that one can change the implementation of an LLM's sampling process without changing the LLM itself, nor does it interpret LLMs as intended, for the method involves explicitly representing a nondeterministic LLM as a deterministic causal model. I here present a much simpler method for generating counterfactuals that is based on an LLM's intended interpretation by representing it as a nondeterministic causal model instead. The advantage of my simpler method is that it is directly applicable to any black-box LLM without modification, as it is agnostic to any implementation details. The advantage of the existing method, on the other hand, is that it directly implements the generation of a specific type of counterfactuals that is useful for certain purposes, but not for others. I clarify how both methods relate by offering a theoretical foundation for reasoning about counterfactuals in LLMs based on their intended semantics, thereby laying the groundwork for novel application-specific methods for generating counterfactuals. (shrink)

We explore the capabilities of Large Language Models (LLMs) by comparing the way they gather data with the way humans build knowledge. Here we examine how scientific knowledge is made and compare it with LLMs. The argument is structured by reference to two figures, one representing scientific knowledge and the other LLMs. In a 2014 study, scientists explain how they choose to ignore a ‘fringe science’ paper in the domain in the domain of gravitational wave physics: the decisions are (...) made largely as a result of tacit knowledge built up in social discourse, mostly spoken discourse, within closed groups of experts. It is argued that LLMs cannot or do not currently access such discourse, but it is typical of the early formation of scientific knowledge. LLMs ‘understanding’ builds on written literatures and is therefore insecure in the case of the initial stages of knowledge building. We refer to Colin Fraser’s ‘Dumb Monty Hall problem’ where in 2023 ChatGPT failed though a year or so later LLMs were succeeding. We argue that this is not a matter of improvement in LLMs ability to reason but in the change in the body of human written discourse on which they can draw (or changes being put in by humans ‘by hand’). We then invent a new Monty Hall prompt and compare the responses of a panel of LLMs and a panel of humans: they are starkly different but we explain that the previous mechanisms will soon allow the LLMs to align themselves to humans once more. Finally, we look at ‘overshadowing’ where a settled body of discourse becomes so dominant that LLMs fail to respond to small variations in prompts which render the old answers nonsensical. The ‘intelligence’, we argue, is in the humans not the LLMs. (shrink)

The extraordinary success of recent Large Language Models (LLMs) on a diverse array of tasks has led to an explosion of scientific and philosophical theorizing aimed at explaining how they do what they do. Unfortunately, disagreement over fundamental theoretical issues has led to stalemate, with entrenched camps of LLM optimists and pessimists often committed to very different views of how these systems work. Overcoming stalemate requires agreement on fundamental questions, and the goal of this paper is to address one (...) such question, namely: is LLM behavior driven partly by representation-based information processing of the sort implicated in biological cognition, or is it driven entirely by processes of memorization and stochastic table look-up? This is a question about what kind of algorithm LLMs implement, and the answer carries serious implications for higher level questions about whether these systems have beliefs, intentions, concepts, knowledge, and understanding. I argue that LLM behavior is partially driven by representation-based information processing, and then I describe and defend a series of practical techniques for investigating these representations and developing explanations on their basis. The resulting account provides a groundwork for future theorizing about language models and their successors. (shrink)

If some language models become welfare subjects, how could we find out what welfare states they are in? We develop an analogical-abductive approach for measuring language model welfare. This approach adapts paradigms used to measure human or non-human animal welfare, for instance verbal reports or non-verbal choice behavior (analogy). Then, one systematically searches for clusters of such indicators in language models. This search for clusters contributes to the cross-validation of welfare measures and motivates explanations in terms (...) of a welfare state that underlies multiple measures (abduction). Additionally, we argue that measures of welfare based on verbal report may already be applicable to current language models because they have substantial introspective abilities, semantic competence, and requisite inclinations for accurate reporting. We further discuss a detailed empirical case study that exemplifies the feasibility and fruitfulness of the analogical-abductive approach and reply to the objection that language model behavior is always better explained in terms of statistical pattern completion or memorization, rather than welfare. While language model welfare measurement undoubtedly faces some remaining theoretical as well as many detailed practical challenges, we conclude that there is a strong case that the analogical abductive approach offers a viable path forward. (shrink)

Large language models (LLMs) challenge Chomsky’s long-standing mysterian view of the creative aspect of language use (CALU). By exhibiting fluent, situation-appropriate linguistic behavior and offering concrete mechanistic hypotheses, they provide the first viable scientific models of CALU. We endorse Futrell and Mahowald’s call to integrate LLMs into linguistic inquiry and suggest a bolder aim: elucidating the mechanisms underlying linguistic creativity.

This study investigates the potential use of current neural language models in psychological practice, particularly in diagnosing patients through the analysis of textual content. The growing interest in the capabilities of neural language models, especially in areas where natural language serves as a primary information source, and specifically by psychology, is the motivation behind this study. As a case study, we tackled the assessment of coherence in autobiographical narratives, a diagnostic methodology widely recognized in psychology for its (...) pivotal role in understanding personal identity. This is a particularly intriguing case study for two reasons. On the one hand, it demonstrates how these tools can be a valid aid in psychological practice, in a task notoriously burdensome and time-consuming for human operators. On the other hand, it tests the ability of neural language models to grasp deep, global conceptual aspects from a text, a task not easily accomplished even by professional psychologists. Coherence is evaluated here quantitatively using the standard procedure known as the Narrative Coherence Coding Scheme. The results obtained are particularly promising, especially when using models subjected to fine-tuning with some examples of stories and their evaluations. However, even without fine-tuning, models prompted with appropriate prompts also perform acceptably. (shrink)

Large language model evaluation has become dominated by single-number leaderboards that rank models using aggregate scores across diverse tasks. While these leaderboards provide useful high-level comparisons, they obscure critical details about model behavior, capabilities, and limitations that matter for responsible deployment. This paper critiques current LLM benchmarking practices and proposes a framework for comparative analysis built on three principles: parity of information (standardized evaluation conditions), uncertainty and risk reporting (confidence intervals and safety metrics), and benchmark cards (comprehensive (...) metadata about datasets, tasks, and protocols). Through case studies in multidomain evaluation, I demonstrate how transparent, multidimensional comparisons can support responsible model development and deployment while enabling fairer assessment of competing approaches. The framework provides both theoretical foundations and practical tools for moving beyond reductive metrics toward evaluation practices that capture the dimensions of model performance most relevant for beneficial real-world deployment. (shrink)

Informed consent in surgical settings requires not only the accurate communication of medical information but also the establishment of trust through empathic engagement. The use of large language models (LLMs) offers a novel opportunity to enhance the informed consent process by combining advanced information retrieval capabilities with simulated emotional responsiveness. However, the ethical implications of simulated empathy raise concerns about patient autonomy, trust and transparency. This paper examines the challenges of surgical informed consent, the potential benefits and limitations of (...) digital tools such as LLMs and the ethical implications of simulated empathy. We distinguish between active empathy, which carries the risk of creating a misleading illusion of emotional connection and passive empathy, which focuses on recognising and signalling patient distress cues, such as fear or uncertainty, rather than attempting to simulate genuine empathy. We argue that LLMs should be limited to the latter, recognising and signalling patient distress cues and alerting healthcare providers to patient anxiety. This approach preserves the authenticity of human empathy while leveraging the analytical strengths of LLMs to assist surgeons in addressing patient concerns. This paper highlights how LLMs can ethically enhance the informed consent process without undermining the relational integrity essential to patient-centred care. By maintaining transparency and respecting the irreplaceable role of human empathy, LLMs can serve as valuable tools to support, rather than replace, the relational trust essential to informed consent. (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.

Large language models like GPT-4 have achieved remarkable proficiency in a broad spectrum of language-based tasks, some of which are traditionally associated with hallmarks of human intelligence. This has prompted ongoing disagreements about the extent to which we can meaningfully ascribe any kind of linguistic or cognitive competence to language models. Such questions have deep philosophical roots, echoing longstanding debates about the status of artificial neural networks as cognitive models. This article -- the first part of two (...) companion papers -- serves both as a primer on language models for philosophers, and as an opinionated survey of their significance in relation to classic debates in the philosophy cognitive science, artificial intelligence, and linguistics. We cover topics such as compositionality, language acquisition, semantic competence, grounding, world models, and the transmission of cultural knowledge. We argue that the success of language models challenges several long-held assumptions about artificial neural networks. However, we also highlight the need for further empirical investigation to better understand their internal mechanisms. This sets the stage for the companion paper (Part II), which turns to novel empirical methods for probing the inner workings of language models, and new philosophical questions prompted by their latest developments. (shrink)

Today, large language models are very powerful, informational and analytical tools that significantly accelerate most of the existing methods and methodologies for processing informational processes. Scientific information is of particular importance in this capacity, which gradually involves the power of large language models. This interaction of science and qualitative new opportunities for working with information lead us to new, unique scientific discoveries, their great quantitative diversity. There is an acceleration of scientific research, a reduction in the time spent (...) on its implementation – the freed up time can be spent both on solving new scientific problems and on scientific creativity, which, although it may not necessarily lead to a specific solution to a particular scientific problem, but is able to demonstrate the beauty of science in various disciplinary areas. As a result, the interaction of large language models and scientific information is at the same time a research for solutions to scientific problems, scientific problems, and scientific creativity. Solving scientific problems requires the ability to efficiently process big data, which cannot be done without an effective method – one of the significant methods was the Transformer architecture, introduced in 2017 and comprehensively integrated into the GPT‑3 model, which, as of September 2020, was the largest and most advanced language model in the world. Therefore, GPT‑3 can be called the basis of most scientific developments carried out in the context of using large language models. The interaction of science and large language models has become a factor in the emergence of a large number of questions, among which are: «Is the result of data analysis new knowledge?», «What are the prospects for scientific creativity in the era of big computing?». Currently, these issues are extremely important, because they allow us to develop the foundations for effective human‑computer interaction. Therefore, this study analyzes the issues presented. (shrink)

People reliably associate the meanings of both abstract and concrete words with colors distributed over color space, a phenomenon that influences aspects of visual cognition ranging from object recognition to interpreting information visualizations. Prior research has hypothesized that color–concept associations arise from the cross‐modal statistical structure of experience, but it remains unclear whether natural environments contain such structure or whether learning systems can discover it without strong prior constraints. To address these questions, we investigated whether GPT‐4, a multimodal large (...) class='Hi'>language model, can estimate color–concept association ratings that approximate those made by people. We tested 71 colors spanning perceptual color space and a variety of concepts varying in abstractness. GPT‐4 ratings correlated strongly with human ratings across a range of prompting strategies, outperforming prior state‐of‐the‐art methods for automatically estimating color–concept associations from images. In an empirical study assessing people's ability to interpret the meanings of colors in information visualizations, palettes generated from GPT‐4's rating data were not only interpretable but, in some cases, more effective than those based on human ratings. Taken together, our results suggest that high‐order covariance between language and perception, present in web‐scale data, provide sufficient information to learn color–concept associations without initial constraints, and that machine‐derived associations can support the optimization of information visualizations for visual communication. (shrink)

The philosophy of mathematical practice (PMP) looks to evidence from working mathematics to help settle philosophical questions. One prominent program under the PMP banner is the study of explanation in mathematics, which aims to understand what sorts of proofs mathematicians consider explanatory and what role the pursuit of explanation plays in mathematical practice. PMP researchers have recently turned to corpus analysis methods as a promising alternative to small-scale case studies. Such methods stand to benefit, it would seem, from the sophisticated (...) text-comprehension capabilities of today's frontier large language models (LLMs). This paper reports the results from such a corpus study making use of Google's Gemini 2.5 Pro. Based on a sample of 5000 mathematics preprints, the experiments yielded a dataset of hundreds of useful annotated examples. Its aim was to gain insight on questions like the following: How often do mathematicians make claims about explanation in the relevant sense? Do mathematicians’ explanatory practices vary in any noticeable way by subject matter? Which philosophical theories of explanation are most consistent with a large body of non-cherry-picked examples? How might philosophers make further use of AI tools to gain insights from large datasets of this kind? As the first PMP study making extensive use of LLM methods, it also seeks to begin a conversation about these methods as research tools in practice-oriented philosophy and to evaluate the strengths and weaknesses of current models for such work. (shrink)

This paper explores the intersection of AI hallucinations and the question of AI consciousness, examining whether the erroneous outputs generated by large language models (LLMs) could be mistaken for signs of emergent intelligence. AI hallucinations, which are false or unverifiable statements produced by LLMs, raise significant philosophical and ethical concerns. While these hallucinations may appear as data anomalies, they challenge our ability to discern whether LLMs are merely sophisticated simulators of intelligence or could develop genuine cognitive processes. By analyzing (...) the causes of AI hallucinations, their impact on the perception of AI cognition, and the potential implications for AI consciousness, this paper contributes to the ongoing discourse on the nature of artificial intelligence and its future evolution. (shrink)

What large language models are able to do can teach us valuable lessons about our own mental lives.

Our information and communication environment has fallen short of the ideals that networked global communication might have served. Existing recommender systems very likely contribute to this shortfall. In this paper, which draws on the normative tools of philosophy of computing, informed by empirical and technical insights from computer science, we make the moral case for an alternative approach. We argue that existing recommenders incentivise mass surveillance, concentrate power, fall prey to narrow behaviourism, and compromise user agency. Rather than just trying (...) to avoid algorithms entirely, or to incrementally alter the current paradigm, researchers and engineers should explore an alternative approach: the use of large language models (LLMs) and LLM-based agents to source and curate content that matches users’ preferences and values, expressed in natural language. The use of Language Model Agents for recommendation poses its own challenges, including those related to candidate generation, computational efficiency, preference modelling, and prompt injection. Nonetheless, if implemented successfully, these systems could: guide us through the digital public sphere without relying on mass surveillance; shift power away from platforms towards users; optimise for what matters instead of just for behavioural proxies; and scaffold our agency instead of undermining it. (shrink)

This letter explores the intricate historical and contemporary links between large language models (LLMs) and cognitive science through the lens of information theory, statistical language models, and socioanthropological linguistic theories. The emergence of LLMs highlights the enduring significance of information‐based and statistical learning theories in understanding human communication. These theories, initially proposed in the mid‐20th century, offered a visionary framework for integrating computational science, social sciences, and humanities, which nonetheless was not fully fulfilled at that time. The subsequent (...) development of sociolinguistics and linguistic anthropology, especially since the 1970s, provided critical perspectives and empirical methods that both challenged and enriched this framework. This letter proposes that two pivotal concepts derived from this development, metapragmatic function and indexicality, offer a fruitful theoretical perspective for integrating the semantic, textual, and pragmatic, contextual dimensions of communication, an amalgamation that contemporary LLMs have yet to fully achieve. The author believes that contemporary cognitive science is at a crucial crossroads, where fostering interdisciplinary dialogues among computational linguistics, social linguistics and linguistic anthropology, and cognitive and social psychology is in particular imperative. Such collaboration is vital to bridge the computational, cognitive, and sociocultural aspects of human communication and human−AI interaction, especially in the era of large language and multimodal models and human‐centric Artificial Intelligence (AI). (shrink)

The explosion of interest in large language models (LLMs) has been accompanied by concerns over the extent to which generated outputs can be trusted, owing to the prevalence of bias, hallucinations, and so forth. Accordingly, there is a growing interest in the use of ontologies and knowledge graphs to make LLMs more trustworthy. This rests on the long history of ontologies and knowledge graphs in constructing human-comprehensible justification for model outputs as well as traceability concerning the impact of (...) evidence on other evidence. Understanding the nature of arguments and argumentation is critical to each, especially when LLM output conflicts with what is expected by users. The central contribution of this article is to extend the Arguments Ontology (ARGO) - an ontology specific to the domain of argumentation and evidence broadly construed - into the space of LLM fact-checking in the interest of promoting justification and traceability research through the use of ARGO-based ‘blueprints’. (shrink)

The recent advances in Large Language Models (LLMs) and their deployment in social settings prompt an important philosophical question: are LLMs social agents? This question finds its roots in the broader exploration of what engenders sociality. Since AI systems like chatbots, carebots, and sexbots are expanding the pre-theoretical boundaries of our social ontology, philosophers have two options. One is to deny LLMs membership in our social ontology on theoretical grounds by claiming something along the lines that only organic or (...) X-type creatures belong in our social world. Second, expand our ontological boundaries. I take the second route and claim LLMs implemented as social chatbots are social agents. Utilizing Brian Epstein's concepts of grounding and anchoring, alongside Dee Peyton's criteria for what makes a property social, I posit that the capacity for conversations is a social property. Further, within a framework where social agency is considered multi-dimensional, akin to agency, possessing a social property is sufficient for attaining social agency. Thus establishing chatbots as social agents. A helpful concept here is Levels of Abstraction (LoA). The LoA framework allows for extracting relevant and important information from the target domain (the social world). For example, the property of being a 'boy' and the capacity for conversation are social properties; hence, they are part of the social domain. LLMs have the capacity for conversations. So, if we focus on the conversational-LoA and not the gender-LoA, then LLMs occupy a dimension of sociality (i.e., conversational dimensions). (shrink)

Large Language Models are useless for linguistics, as they are probabilistic models that require a vast amount of data to analyze externalized strings of words. In contrast, human language is underpinned by a mind-internal computational system that recursively generates hierarchical thought structures. The language system grows with minimal external input and can readily distinguish between real language and impossible languages.

Large language models (LLMs) have been proposed as candidate models of human semantics, and as such, they must be able to account for conceptual combination. This work explores the ability of two LLMs, namely, BERT-base and Llama-2-13b, to reveal the implicit meaning of existing and novel compound words. According to psycholinguistic theories, understanding the meaning of a compound (e.g., “snowman”) involves its automatic decomposition into constituent meanings (“snow,” “man”), which are then connected by an implicit semantic relation selected from (...) a set of possible competitors (FOR, MADE OF, BY, …) to obtain a plausible interpretation (“man MADE OF snow”). Here, we leverage the flexibility of LLMs to obtain contextualized representations for both target compounds (e.g., “snowman”) and their implicit interpretations (e.g., “man MADE OF snow”). We demonstrate that replacing a compound with a paraphrased version leads to changes to the embeddings that are inversely proportional to the paraphrase's plausibility, estimated by human raters. While this relation holds for both existing and novel compounds, results obtained for novel compounds are substantially weaker, and older distributional models outperform LLMs. Nonetheless, the present results show that LLMs can offer a valid approximation of the internal structure of compound words posited by cognitive theories, thus representing a promising tool to model word senses that are at once implicit and possible. (shrink)

We develop new experimental paradigms for measuring welfare in language models. We compare verbal reports of models about their preferences with preferences expressed through behavior when navigating a virtual environment and selecting conversation topics. We also test how costs and rewards affect behavior and whether responses to an eudaimonic welfare scale - measuring states such as autonomy and purpose in life - are consistent across semantically equivalent prompts. Overall, we observed a notable degree of mutual support between our measures. (...) The reliable correlations observed between stated preferences and behavior across conditions suggest that preference satisfaction can, in principle, serve as an empirically measurable welfare proxy in some of today's AI systems. Furthermore, our design offered an illuminating setting for qualitative observation of model behavior. Yet, the consistency between measures was more pronounced in some models and conditions than others and responses were not consistent across perturbations. Due to this, and the background uncertainty about the nature of welfare and the cognitive states (and welfare subjecthood) of language models, we are currently uncertain whether our methods successfully measure the welfare state of language models. Nevertheless, these findings highlight the feasibility of welfare measurement in language models, inviting further exploration. (shrink)

Large language models (LLMs) such as ChatGPT are opening up new areas of research and teaching potential across a variety of domains. The purpose of the present conceptual paper is to map this new terrain from the point of view of Critical Discourse Studies (CDS). We demonstrate that the usage of LLMs raises concerns that definitely fall within the remit of CDS; among them, power and inequality. After an initial explanation of LLMs, we focus on three key areas of (...) reflection. The first is a general stock-taking, where we look at CDS’ theoretical underpinnings and what they imply for working with AI-generated language data. The second issue is authorship, where we assess the traceability of linguistic metadata and the ethically sensitive situation with regard to ownership of texts. The third area is linguistic homogenisation, where we examine how LLM usage privileges the mainstream. Afterwards, we explore ways in which LLMs could be used in research, and we discuss the implications of exploring their use in the classroom through a CDS lens. We close the paper with some observations on likely future developments in AI and how CDS can contribute with its distinctive theoretical, methodological and critical apparatus. (shrink)

Humans can attribute beliefs to others. However, it is unknown to what extent this ability results from an innate biological endowment or from experience accrued through child development, particularly exposure to language describing others' mental states. We test the viability of the language exposure hypothesis by assessing whether models exposed to large quantities of human language display sensitivity to the implied knowledge states of characters in written passages. In pre‐registered analyses, we present a linguistic version of the (...) False Belief Task to both human participants and a large language model, GPT‐3. Both are sensitive to others' beliefs, but while the language model significantly exceeds chance behavior, it does not perform as well as the humans nor does it explain the full extent of their behavior—despite being exposed to more language than a human would in a lifetime. This suggests that while statistical learning from language exposure may in part explain how humans develop the ability to reason about the mental states of others, other mechanisms are also responsible. (shrink)

This article gives mathematical pseudocodes for large language model training based on a dataset from a corpus, inference, and chat with possibly lengthy human prompts and generated replies. It introduces the concepts of “microlect" and “resonantcommunity." Most generally, a microlect is a specialized behavioral "language" consisting of expressions called "cores" built upon units called "keys." It is a minimal and adequate representation of a mental model. An expression may resonate more or less with a mental (...) class='Hi'>model. A resonant- community is a set of human beings whose expressions resonate with one another. These notions are helpful in understanding of how the “frozen matrices" of a large language model are a conduit for resonance between human-beings and a zeitgeist represented in a corpus. The primary objectives are to encapsulate the functionality of large language models in a succinct technical representation, and to explain how this functionality connects human beings to their ambient cultural zeitgeist as represented in a textual corpus. The research draws on relevant literature in psychology, linguistics, mathematics, physics, and software engineering. The significance of the research is to embed the explanation in a perspective on the entirety of human expressiveness. The study can provide valuable insights based on resonances between human beings among themselves and with computing machines. (shrink)

Recent studies allege that large language models (LLMs) exhibit a range of cognitive biases familiar from human cognition. I argue that the case for many biases is weaker than it may appear. Using case studies of knowledge effects in the Wason selection task, availability bias in relation extraction, and anchoring bias in code generation, I show how a range of vindicatory strategies traditionally used to vindicate apparent biases in humans can be used to push back against allegations of bias (...) in LLMs. I discuss implications for the role of cognitive bias in evaluating LLM performance, the rationality of human cognition, and future work on cognitive bias in LLMs. (shrink)

Formalizations serve as cognitive tools. By enabling algorithmic reasoning over sets of statements in a formal language, they provide a cognitive boost for human reasoners. We argue that the emergence of large language models (LLMs) as a technology for the analysis and generation of natural language provides a new perspective on the relative roles of formal and natural languages in formalization.

Despite the exceptional performance of large language models (LLMs) on a wide range of tasks involving natural language processing and reasoning, there has been sharp disagreement as to whether their abilities extend to more creative human abilities. A core example is the interpretation of novel metaphors. Here we assessed the ability of GPT-4, a state-of-the-art large language model, to provide natural-language interpretations of a recent AI benchmark (Fig-QA dataset), novel literary metaphors drawn from Serbian poetry (...) and translated into English, and entire novel English poems. GPT-4 outperformed previous AI models on the Fig-QA dataset. For metaphors drawn from Serbian poetry, human judges – blind to the fact that an AI model was involved – rated metaphor interpretations generated by GPT-4 as superior to those provided by a group of college students. In interpreting reversed metaphors, GPT-4, as well as humans, exhibited signs of sensitivity to the Gricean cooperative principle. In addition, for several novel English poems GPT-4 produced interpretations that were rated as excellent or good by a human literary critic. These results indicate that LLMs such as GPT-4 have acquired an emergent ability to interpret literary metaphors, including those embedded in novel poems. (shrink)

This article is a commentary on Jonathan Birch's The edge of sentience. It considers the role of deference to consciousness experts in the citizens' assemblies that he calls for; evaluates his claim that the human fetus is a sentience candidate from 12 weeks' gestation; and explores some of the issues raised by his approach to sentience in large language models and other AI systems.