LLMs have dramatically improved in capabilities in recent years. This raises the question of whether LLMs could become genuine agents with beliefs and desires. This paper demonstrates an in principle limit to LLM agency, based on their architecture. LLMs are next word predictors: given a string of text, they calculate the probability that various words can come next. LLMs produce outputs that reflect these probabilities. I show that next word predictors are exploitable. If LLMs are (...) prompted to make probabilistic predictions about the world, these predictions are guaranteed to be incoherent, and so Dutch bookable. If LLMs are prompted to make choices over actions, their preferences are guaranteed to be intransitive, and so money pumpable. In short, the problem is that selecting an action based on its potential value is structurally different then selecting the description of an action that is most likely given a prompt: probability cannot be forced into the shape of expected value. The in principle exploitability of LLMs raises doubts about how agential they can become. This exploitability also offers an opportunity for humanity to safely control such AI systems. (shrink)

In their recent Opinion in TiCS, Yildirim and Paul propose that large language models (LLMs) have ‘instrumental knowledge’ and possibly the kind of ‘worldly’ knowledge that humans do. They suggest that the production of appropriate outputs by LLMs is evidence that LLMs infer ‘task structure’ that may reflect ‘causal abstractions of... entities and processes in the real world.' While we agree that LLMs are impressive and potentially interesting for cognitive science, we resist this project on two (...) grounds. First, it casts LLMs as agents rather than as models. Second, it suggests that causal understanding could be acquired from the capacity for mere prediction. (shrink)

In this paper, we propose RSI-LLM (Recursively Self-Improving Large Language Model), which recursively executes its inference and improves its parameters to fulfill the instrumental goals of superintelligence: G1: Self-preservation, G2: Goal-content integrity, G3: Intelligence enhancement, and G4: Resource acquisition. We empirically observed the behavior of the LLM that tries to design tools to achieve G1~G4, within the autonomous self-improvement and knowledge acquisition. During interventions in these LLMs' coding experiments to ensure safetyness, we have also discovered that, as the creator (...) of the LLM, we are able to change or edit any of those LLMs' cognitive information about the world in any modality, including time, vision, hearing, etc. We discuss the relationship between us humans and these LLMs in the context of the creator and the creation of the observable universe. (shrink)

Elon Musk famously predicted that an artificial intelligence superior to the smartest individual human would arrive by the year 2025. In response, Gary Marcus offered Musk a $1 million bet to the effect that he would be proved wrong. In specifying the conditions of this bet (which Musk did not take) Marcus lists the following ‘tasks that ordinary people can perform’ which, he claimed, AI will not be able to perform by the end of 2025. • Reliably drive a car (...) in a novel location that they haven’t previously encountered, even in the face of unusual circumstances like hand lettered signs, without the assistance of other humans. • Learn to ride a mountain bike off-road through forest trails. • Babysit children in an unfamiliar home and keep them safe. • Tend to the physical and psychological needs of an elderly or infirm person. Each of these tasks involves the application of practical or tacit knowledge, or what is also called ‘knowing how’, which is to say knowledge of a sort that is captured not by means of sentences or propositions or explicit rules, but rather through the expertise demonstrated in human actions. The talk addresses the puzzle that nowhere in the AI literature do we find a discussion of knowing how as a feature of human intelligence. (shrink)

As large language models (LLMs) continue to demonstrate remarkable abilities across various domains, computer scientists are developing methods to understand their cognitive processes, particularly concerning how (and if) LLMs internally represent their beliefs about the world. However, this field currently lacks a unified theoretical foundation to underpin the study of belief in LLMs. This article begins filling this gap by proposing adequacy conditions for a representation in an LLM to count as belief-like. We argue that, while the (...) project of belief measurement in LLMs shares striking features with belief measurement as carried out in decision theory and formal epistemology, it also differs in ways that should change how we measure belief. Thus, drawing from insights in philosophy and contemporary practices of machine learning, we establish four criteria that balance theoretical considerations with practical constraints. Our proposed criteria include accuracy, coherence, uniformity, and use, which together help lay the groundwork for a comprehensive understanding of belief representation in LLMs. We draw on empirical work showing the limitations of using various criteria in isolation to identify belief representations. (shrink)

In his 1955 essay "Meaning and synonymy in natural languages", Rudolf Carnap presents a thought experiment wherein an investigator provides a hypothetical robot with a definition of a concept together with a description of an individual, and then asks the robot if the individual is in the extension of the concept. In this work, we show how to realize Carnap's Robot through knowledge probing of an large language model (LLM), and argue that this provides a useful cognitive tool for conceptual (...) engineers to compare the extension of a proposed concept definition to the extensional knowledge represented as facts in a given knowledge base, providing a possible solution to the implementation problem in conceptual engineering. (shrink)

Sideloading is the creation of a digital model of a person during their life via iterative improvements of this model based on the person's feedback. The progress of LLMs with large prompts allows the creation of very large, book-size prompts which describe a personality. We will call mind-models created via sideloading "sideloads"; they often look like chatbots, but they are more than that as they have other output channels, like internal thought streams and descriptions of actions. -/- By arranging (...) the predictive power of the facts about a person, we can write down the most important facts first. This allows us to get a high level of fidelity in a person's model by writing down a finite number of facts and running it as an LLM prompt with instructions to create a person's chatbot. This instruction also includes an extensive general sideload prompt (we will call it a prompt-loader), which explains how the person's sideload should work, but it is universal and works as an operating system for different persons. -/- I created my personal sideload to test all this and it started working surprisingly quickly. We found that there are three main levels of facts about a person: core facts, long-term memory, and historical facts. Core facts should go to the main prompt of the LLM, long-term memory belongs to RAG, and historical facts – which a person often does not remember – should be only used to extract new relevant data. The most important thing in preparing data for sideloading is to get a full list of the core facts, and not to mix it with long-term memory and historical facts. The LLM-sideload project is inherently future-oriented. New state-of-the-art LLMs with large prompt capabilities emerge quarterly. This allows for the preparation of prompts that exceed current LLM capabilities, anticipating smooth execution in the following year. Below we present the key findings of this work in the form of an executive summary: Quality of sideloads is evaluated using three metrics: • Facts: Accuracy of the answers about the person's life • Vibe: How well the sideload captures the person's style and personality, as judged by the person themselves and their acquaintances • Brilliant insights: Ability to contribute unique, valuable ideas beyond the training data, while maintaining the author's style • Coarseness: Levels of details as percent of the total of recordable textual information from memory estimated to be 10000 pages. Similar to the jpeg level of compression. Estimated performance of Alexey Turchin's sideload: • Facts: 70% correct • Vibe: 20% accurate • Brilliant insights: near zero • Coarseness: 10% The project achieved these results in about one month of work. Vibe is harder to evaluate and regulate than facts. This approach focuses on informational personal identity, setting aside issues of observer identity and internal qualia. (shrink)

This essay analyzes the rise of Large Language Models (LLMs) such as GPT-4 or Gemini, which are now incorporated in a wide range of products and services in everyday life. Importantly, it considers some of their hidden human costs. First, is the question of who is left behind by the further infusion of LLMs in society. Second, is the issue of social inequalities between lingua franca and those which are not. Third, LLMs will help disseminate scientific concepts, (...) but their meanings' dissemination is at risk. Fourth, is the issue of information security and how nations and individuals might adapt. Fifth, is the issue of the environmental cost of LLMs for future generations. At the dawn of the conversational AI era, it would be wise to ponder what inequalities and empowerments will be co-created with LLMs. (shrink)

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)

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)

This paper opens the philosophical debate around the notion of Artificial General Intelligence (AGI) and its application in Large Language Models (LLMs). Through the lens of moral philosophy, the paper raises questions about these AI systems' capabilities and goals, the treatment of humans behind them, and the risk of perpetuating a monoculture through language.

The emergence of Large Language Models (LLMs) and advancements in Artificial Intelligence (AI) offer an opportunity for computational social science research at scale. Building upon prior explorations of LLM agent design, our work introduces a simulated agent society where complex social relationships dynamically form and evolve over time. Agents are imbued with psychological drives and placed in a sandbox survival environment. We conduct an evaluation of the agent society through the lens of Thomas Hobbes's seminal Social Contract Theory (SCT). (...) We analyze whether, as the theory postulates, agents seek to escape a brutish "state of nature" by surrendering rights to an absolute sovereign in exchange for order and security. Our experiments unveil an alignment: Initially, agents engage in unrestrained conflict, mirroring Hobbes's depiction of the state of nature. However, as the simulation progresses, social contracts emerge, leading to the authorization of an absolute sovereign and the establishment of a peaceful commonwealth founded on mutual cooperation. This congruence between our LLM agent society's evolutionary trajectory and Hobbes's theoretical account indicates LLMs' capability to model intricate social dynamics and potentially replicate forces that shape human societies. By enabling such insights into group behavior and emergent societal phenomena, LLM-driven multi-agent simulations, while unable to simulate all the nuances of human behavior, may hold potential for advancing our understanding of social structures, group dynamics, and complex human systems. (shrink)

Chatbots powered by large language models (LLMs) are increasingly capable of engaging in what seems like natural conversations with humans. This raises the question of whether we should interact with these chatbots in a morally considerate manner. In this chapter, we examine how to answer this question from within the normative framework of virtue ethics. In the literature, two kinds of virtue ethics arguments, the moral cultivation and the moral character argument, have been advanced to argue that we should (...) afford moral treatment to social robots. However, we propose a moral character argument against the view that we should afford moral treatment to LLM-based chatbots drawing on the notion of practical wisdom. Practical wisdom in this context consists in the skill to discern genuine and artificial sociality. Drawing on ideas from phenomenological philosophy, we argue that this involves the ability to distance oneself from direct social perception and the ability to widen one’s awareness over one’s apparently social interactions. We conclude by suggesting that this skill is a kind of technomoral wisdom required to live well with advanced, social AI systems. (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)

There are calls from various sectors, including the popular press, industry, and academia, to incorporate artificial intelligence (AI)-based technologies in general, and large language models (LLMs) (such as ChatGPT and Gemini) in particular, into various spheres of the South African higher education sector. Nonetheless, the implementation of such technologies is not without ethical risks, notably those related to bias, unfairness, privacy violations, misinformation, lack of transparency, and threats to autonomy. This paper gives an overview of the more pertinent ethical (...) concerns that may result from the deployment of various current AI technologies in the South African higher education context. It provides a broad overview of the relevant AI ethics literature and distills a few general AI-ethics principles that can serve as guidelines for the ethical development, adoption, and use of AI systems. Suggestions are made as to how these might be applied to mitigate the ethical concerns in the South African higher education context. Overall, it is argued that AI technologies should only be adopted if they offer demonstrable benefits to stakeholders affected by them and that care should be taken to ensure that any potential harms are adequately addressed. (shrink)

The recent developments in artificial intelligence (AI), particularly in light of the impressive capabilities of transformer-based Large Language Models (LLMs), have reignited the discussion in cognitive science regarding whether computational devices could possess semantic understanding or whether they are merely mimicking human intelligence. Recent research has highlighted limitations in LLMs’ reasoning, suggesting that the gap between mere symbol manipulation (syntax) and deeper understanding (semantics) remains wide open. While LLMs overcome certain aspects of the symbol grounding problem through (...) human feedback, they still lack true semantic understanding, struggling with common-sense reasoning and abstract thinking. This paper argues that while adding sensory inputs and embodying AI through sensorimotor integration with the environment might enhance its ability to connect symbols to real-world meaning, this alone would not close the gap between syntax and semantics. True meaning-making also requires a connection to subjective experience, which current AI lacks. The path to AGI must address the fundamental relationship between symbol manipulation, data processing, pattern matching, and probabilistic best guesses with true knowledge that requires conscious experience. A transition from AI to AGI can occur only if it possesses conscious experience, which is closely tied to semantic understanding. Recognition of this connection could furnish new philosophical insights into longstanding practical and philosophical questions for theories in biology and cognitive science and provide more meaningful tests of intelligence than the Turing test. (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)

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)

The paper continues my earlier Chat with OpenAI’s ChatGPT with a Focused LLM Experiment (FLEX). The idea is to conduct Large Language Model (LLM) based explorations of certain areas or concepts. The approach is based on crafting initial guiding prompts and then follow up with user prompts based on the LLMs’ responses. The goals include improving understanding of LLM capabilities and their limitations culminating in optimized prompts. The specific subjects explored as research subject matter include a) diagonalization techniques as (...) practiced by Cantor, Turing, Gödel, and the advances, such as the Forcing techniques introduced by Paul Cohen and later investigators, b) Knowledge Hierarchies & Mapping Exercises, c) Discussions of IJ Good’s Speculations Concerning the First Ultraintelligent Machine, AGI, and Super- intelligence. Results suggest variability between major models like ChatGPT-4, Llama-3, Cohere, Sonnet and Opus. Results also point to strong dependence on users’ preexisting knowledge and skill bases. The paper should be viewed as ’raw data’ rather than a polished authoritative reference. (shrink)

The re-use of publicly available (personal) data for originally unanticipated purposes has become common practice. Without such secondary uses, the development of many AI systems like large language models (LLMs) and ChatGPT would not even have been possible. This chapter addresses the ethical implications of such secondary processing, with a particular focus on data protection and privacy issues. Legal and ethical evaluations of secondary processing of publicly available personal data diverge considerably both among scholars and the general public. While (...) some of these uses are met with opposition and criticism, others are quite unanimously viewed as unproblematic. Often, proponents and opponents of such practices invoke the same ethical and legal standards for their opposite conclusions. This state of affairs shows that other considerations besides the public availability of data must play a role. It calls for a theoretical clarification of the additional criteria that should guide decisions about the (legally informed) ethical acceptability of re-processing practices. In order to make a contribution towards this goal, the present chapter maps the ongoing debate and systematises the existing contributions around three lines of argument: a consent-centred position, an approach that focuses on the distinction between data and information, and finally a line of argument that focuses on the contextual norms that govern the flows of information. The chapter further relates these arguments to three underlying conceptions of privacy and data protection: rights-based, structural and contextual, and discusses the advantages and disadvantages of each position in the light of concrete examples. It concludes by arguing for a mixed approach that combines core elements of the structural and contextual approaches. The chapter aims to contribute to existing research in the fields of data, AI and research ethics, and to reconnect the debate with ethical and legal scholarship on privacy and data protection. In doing so, it aims to make a theoretical contribution towards refining existing conceptions of privacy and data protection in order to make them more fit to ‘drive our digital world’ as far as the use of publicly available data is concerned. (shrink)

La inteligencia artificial tiene como objetivo incrementar la productividad y mejorar las capacidades de las personas para realizar tareas de manera eficiente. Sin embargo, el uso excesivo de la inteligencia artificial, como los grandes modelos de lenguaje (LLM) tales como ChatGPT, Gemini, Copilot, LLaMa, Bing, entre otros, podría tener un efecto contrario. La automatización de procesos por parte de las máquinas puede llegar a representar una amenaza para el desarrollo neuronal de los usuarios, lo que eventualmente podría conducir a una (...) disminución del crecimiento y la estimulación neuronal debido a la dependencia del uso de lenguaje autogenerado. Este capítulo explora los mecanismos cerebrales involucrados en la creatividad y el desarrollo intelectual de la escritura. Posteriormente, se analiza la suplantación intelectual proporcionada por los modelos generativos y cómo esto afecta las capacidades cognitivas del usuario. Además, se exploran las implicaciones éticas de este complejo tema. Finalmente, se comparten pautas y recomendaciones para el uso adecuado de estas herramientas con el objetivo de aprovechar sus beneficios sin comprometer la integridad académica ni las capacidades analíticas y creativas de la persona. (shrink)

The rapid advancement of Artificial Intelligence (AI) and the developments of Large Language Models (LLMs) like Generative Pretrained Transformers (GPTs) have significantly influenced content creation in scholarly communication and across various fields. This paper presents a comparative analysis of the content reliability between human-generated and GPT-generated scholarly articles. Recent developments in AI suggest that GPTs have become capable in generating content that can mimic human language to a greater extent. This highlights and raises questions about the quality, accuracy, and (...) reliability of such content, especially in academic contexts. Statistical evaluations and quantitative assessments conducted in this study uncover key differences in content accuracy, coherence, citation usage, and overall reliability between human and GPT-generated articles. The paper also examines the potential biases, the role of context, and the implications of AI-generated content for the future of scholarly communication. The study concludes with framework for predictive model indicating the potential future impact of GPT-driven content creation and recommendations for ensuring content quality and ethical considerations in AI usage. (shrink)

In this study, we metaphysically discuss how societal values will change and what will happen to the world when superintelligence is safely realized. By providing a mathematical definition of superintelligence, we examine the phenomena derived from this thesis. If an intelligence explosion is triggered under safe management through advanced AI technologies such as large language models (LLMs), it is thought that a modern form of broken communism—where rights are bifurcated from the capitalist system—will first emerge. In that era, the (...) value of humans will ultimately eliminate external factors, and beings without superintelligence will converge into irreplaceable existences possessing only intrinsic value. The world will be divided into those who possess superintelligence and those who do not. For better or worse, global standardization will progress, and due to over-simulation accompanying the intelligence explosion, we may become unable to distinguish whether the reality we inhabit is original or a copy created through augmented reality. -/- Note: In the creation of this paper, we have undertaken all writing ourselves and have not used generative AI for text generation except for translation purposes. (shrink)

本文深入探討人工智慧（AI）於實現共同福祉與幸福、公平與非歧視、理性公共討論及自主與控制之倫理與正義重要性與挑戰。以中央研究院LLM事件及國家科學技術委員會（NSTC）AI技術研發指導方針為基礎，本文 分析AI能否滿足人類共同利益與福祉。針對AI不公正，本文評估其於區域、產業及社會影響。並探討AI公平與非歧視挑戰，尤其偏差數據訓練問題，及後處理監管，強調理性公共討論之重要性。進而，本文探討理性公眾於 公共討論中之挑戰及應對，如STEM科學素養與技術能力教育之重要性。最後，本文提出“以人為本”方法，非僅依賴AI技術效用最大化，以實現AI正義。 -/- 關鍵詞：AI倫理與正義、公平與非歧視、偏差數據訓練、公共討論、自主性、以人為本的方法.

The EU proposal for the Artificial Intelligence Act (AIA) defines four risk categories: unacceptable, high, limited, and minimal. However, as these categories statically depend on broad fields of application of AI, the risk magnitude may be wrongly estimated, and the AIA may not be enforced effectively. This problem is particularly challenging when it comes to regulating general-purpose AI (GPAI), which has versatile and often unpredictable applications. Recent amendments to the compromise text, though introducing context-specific assessments, remain insufficient. To address this, (...) we propose applying the risk categories to specific AI scenarios, rather than solely to fields of application, using a risk assessment model that integrates the AIA with the risk approach arising from the Intergovernmental Panel on Climate Change (IPCC) and related literature. This model enables the estimation of the magnitude of AI risk by considering the interaction between (a) risk determinants, (b) individual drivers of determinants, and (c) multiple risk types. We use large language models (LLMs) as an example. (shrink)

The present comment concerns a recent AI & Ethics article which purports to report evidence of speciesist bias in various popular computer vision (CV) and natural language processing (NLP) machine learning models described in the literature. I examine the authors' analysis and show it, ironically, to be prejudicial, often being founded on poorly conceived assumptions and suffering from fallacious and insufficiently rigorous reasoning, its superficial appeal in large part relying on the sequacity of the article's target readership.

With the increasing popularity of Large Language Models (LLMs), there has been an increase in the number of AI generated literary works. In the absence of clear authors, and assuming such works have meaning, there lies a puzzle in determining who or what fixes the meaning of such texts. I give an overview of six leading theories for ascribing meaning to literary works. These are Extreme Actual Intentionalism, Modest Actual Intentionalism (1 & 2), Conventionalism, Actual Author Hypothetical Intentionalism, and (...) Postulated Author Hypothetical Intentionalism. I argue that while only Conventionalism and Postulated Author Hypothetical Intentionalism show any promise of adjudicating how we ought to ascribe meaning in the case of AI generated texts, Postulated Author Hypothetical Intentionalism is the stronger of the two views. (shrink)

Large Language Models (LLMs) are generative AI systems capable of producing original texts based on inputs about topic and style provided in the form of prompts or questions. The introduction of the outputs of these systems into human discursive practices poses unprecedented moral and political questions. The article articulates an analysis of the moral status of these systems and their interactions with human interlocutors based on the Habermasian theory of communicative action. The analysis explores, among other things, Habermas's inquiries (...) into the analogy between human minds and computers, and into the status of atypical participants in the linguistic community such as genetically modified subjects and animals. Major conclusions are the LLMs seem to qualify as authors that originally participate in discursive practices but do display only a structurally derivative form of communicative competence and fail to meet the status of communicative agents. In this sense, while the contribution of AI writing systems in public discourse and deliberation can support the process of mutual understanding within the community of speakers, the human actors involved in the development, use, and diffusion of these systems share a collective responsibility for the disclosure of AI authorship and verification and adjudication of validity claims. (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 (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)

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)

The ability to interact in a natural language profoundly changes devices’ interfaces and potential applications of speaking technologies. Concurrently, this phenomenon challenges our mainstream theories of knowledge, such as how to analyze linguistic outputs of devices under existing anthropocentric theoretical assumptions. In section 1, I present the topic of machines that speak, connecting between Descartes and Generative AI. In section 2, I argue that accepted testimonial theories of knowledge and justification commonly reject the possibility that a speaking technological artifact can (...) give testimony. In section 3, I identify three assumptions underlying the view that rejects conversational AIs – AI-based technologies that converse, as testifiers: conversational AIs (1) lack intentions, (2) cannot be normatively assessed, and (3) cannot constitute an object in trust relations, while humans can. In section 4, I propose the concept ‘AI-testimony’ for analyzing outputs of conversational AIs, suggesting three conditions for technologies to deliver AI-testimony: (1) content is propositional, (2) generated and delivered with no other human directly involved, (3) the output is perceived as phenomenologically similar to that of a human. I conclude that this concept overcomes the limitations of the anthropocentric concept of testimony, opening future directions of research without associating conversational AIs with human-like agency. (shrink)

The alignment problem in the context of large language models must consider the plurality of human values in our world. Whilst there are many resonant and overlapping values amongst the world’s cultures, there are also many conflicting, yet equally valid, values. It is important to observe which cultural values a model exhibits, particularly when there is a value conflict between input prompts and generated outputs. We discuss how the co- creation of language and cultural value impacts large language models ( (...) class='Hi'>LLMs). We explore the constitution of the training data for GPT-3 and compare that to the world’s language and internet access demographics, as well as to reported statistical profiles of dominant values in some Nation-states. We stress tested GPT-3 with a range of value-rich texts representing several languages and nations; including some with values orthogonal to dominant US public opinion as reported by the World Values Survey. We observed when values embedded in the input text were mutated in the generated outputs and noted when these conflicting values were more aligned with reported dominant US values. Our discussion of these results uses a moral value pluralism (MVP) lens to better understand these value mutations. Finally, we provide recommendations for how our work may contribute to other current work in the field. (shrink)

Recent advances in natural language processing have given rise to a new kind of AI architecture: the language agent. By repeatedly calling an LLM to perform a variety of cognitive tasks, language agents are able to function autonomously to pursue goals specified in natural language and stored in a human-readable format. Because of their architecture, language agents exhibit behavior that is predictable according to the laws of folk psychology: they function as though they have desires and beliefs, and then make (...) and update plans to pursue their desires given their beliefs. We argue that the rise of language agents significantly reduces the probability of an existential catastrophe due to loss of control over an AGI. This is because the probability of such an existential catastrophe is proportional to the difficulty of aligning AGI systems, and language agents significantly reduce that difficulty. In particular, language agents help to resolve three important issues related to aligning AIs: reward misspecification, goal misgeneralization, and uninterpretability. (shrink)

A key assumption fuelling optimism about the progress of Large Language Models (LLMs) in modelling the world is that the truth is systematic: true statements about the world form a whole that is not just consistent, in that it contains no contradictions, but cohesive, in that the truths are inferentially interlinked. This holds out the prospect that LLMs might rely on that systematicity to fill in gaps and correct inaccuracies in the training data: consistency and cohesiveness promise to (...) facilitate progress towards comprehensiveness in an LLM’s representation of the world. However, philosophers have identified reasons to doubt that the truth is systematic across all domains of thought, arguing that in normative domains, in particular, the truth is not necessarily systematic. I argue that insofar as the truth in normative domains is asystematic, this renders it correspondingly harder for LLMs to make progress, because they cannot rely on the consistency and cohesiveness of the truth to work towards comprehensiveness. And the less LLMs can rely on the systematicity of truth, the less we can rely on them to do our practical deliberation for us, as there is correspondingly more of a role for human agency in navigating asystematic normative domains. (shrink)

According to virtue epistemology, the main aim of education is the development of the cognitive character of students (Pritchard, 2014, 2016). Given the proliferation of technological tools such as ChatGPT and other LLMs for solving cognitive tasks, how should educational practices incorporate the use of such tools without undermining the cognitive character of students? Pritchard (2014, 2016) argues that it is possible to properly solve this ‘technology-education tension’ (TET) by combining the virtue epistemology framework with the theory of extended (...) cognition (EXT) (Clark and Chalmers, 1998). He argues that EXT enables us to consider tools as constitutive parts of the students’ cognitive system, thus preserving their cognitive character from technologically induced cognitive diminishment. The first aim of this paper is to show that this solution is not sufficient to solve the TET. Second, I aim to offer a complementary and more encompassing framework of tool-use to address the TET. Then, I apply it to the educational uses of ChatGPT as the most notable example of LLM, although my arguments can be extended to other generative AI systems. To do so, in Sect. 1.1, I present Pritchard’s framework of cognitive character and virtue epistemology applied in education, to which I am committed in this treatment. In Sects. 2 and 3, I respectively illustrate Pritchard’s (2014) solution to the TET, and I highlight the general limitations of his proposal. Thus, in Sect. 4.1 I characterize ChatGPT as a computational cognitive artifact using Fasoli’s (2017, 2018) taxonomy of cognitive artifacts. In Sect. 4.2, I introduce my proposal, which combines Pritchard’s account of virtue epistemology with Fasoli’s (2017, 2018) taxonomy of cognitive artifacts to address the TET. Finally, in Sect. 5.1, I present some epistemically virtuous uses of ChatGPT in educational contexts. To conclude, I argue in favor of a multidisciplinary approach for analyzing educational activities involving AI technologies such as ChatGPT. (shrink)

What can contemporary machine learning (ML) models do? Given the proliferation of ML models in society, answering this question matters to a variety of stakeholders, both public and private. The evaluation of models' capabilities is rapidly emerging as a key subfield of modern ML, buoyed by regulatory attention and government grants. Despite this, the notion of an ML model possessing a capability has not been interrogated: what are we saying when we say that a model is able to do something? (...) And what sorts of evidence bear upon this question? -/- In this paper, we aim to answer these questions, using the capabilities of large language models (LLMs) as a running example. Drawing on the large philosophical literature on abilities, we develop an account of ML models' capabilities which can be usefully applied to the nascent science of model evaluation. Our core proposal is a conditional analysis of model abilities (CAMA): crudely, a machine learning model has a capability to X just when it would reliably succeed at doing X if it 'tried'. The main contribution of the paper is making this proposal precise in the context of ML, resulting in an operationalisation of CAMA applicable to LLMs. We then put CAMA to work, showing that it can help make sense of various features of ML model evaluation practice, as well as suggest procedures for performing fair inter-model comparisons. (shrink)

This paper examines the question of whether Large Language Models (LLMs) like ChatGPT possess minds, focusing specifically on whether they have a genuine folk psychology encompassing beliefs, desires, and intentions. We approach this question by investigating two key aspects: internal representations and dispositions to act. First, we survey various philosophical theories of representation, including informational, causal, structural, and teleosemantic accounts, arguing that LLMs satisfy key conditions proposed by each. We draw on recent interpretability research in machine learning to (...) support these claims. Second, we explore whether LLMs exhibit robust dispositions to perform actions, a necessary component of folk psychology. We consider two prominent philosophical traditions, interpretationism and representationalism, to assess LLM action dispositions. While we find evidence suggesting LLMs may satisfy some criteria for having a mind, particularly in game-theoretic environments, we conclude that the data remains inconclusive. Additionally, we reply to several skeptical challenges to LLM folk psychology, including issues of sensory grounding, the "stochastic parrots" argument, and concerns about memorization. Our paper has three main upshots. First, LLMs do have robust internal representations. Second, there is an open question to answer about whether LLMs have robust action dispositions. Third, existing skeptical challenges to LLM representation do not survive philosophical scrutiny. (shrink)

The problem considered in this chapter emerges from the tension we find when looking at the design and architecture of chatbots on the one hand and their conversational aptitude on the other. In the way that LLM chatbots are designed and built, we have good reason to suppose they don't possess second-order capacities such as intention, belief or knowledge. Yet theories of conversation make great use of second-order capacities of speakers and their audiences to explain how aspects of interaction succeed. (...) As we can all bear witness to now though, at the point of use chatbots appear capable of performing language tasks at a level close to that of humans. This creates a tension when we consider something like, for example, the classic Gricean theory of implicature. On a broad summary of this type of account, to utter p and implicate q requires the reflexive occurrence of an audience supposing a speaker believes that q, and the speaker believing that their audience can determine they believe it when they utter p. So taken at face value, if a chatbot doesn’t have the capacity for belief, then either in their role as speaker or audience, they would not seem capable of either generating or comprehending implicatures. Yet on the surface it does seem that chatbots are capable of dealing with (some) implicatures, and as such it raises questions about how we should then correlate this with what we think occurs in cases of implicature with chatbots. (shrink)

This paper analyses the phenomenology and epistemology of chatbots such as ChatGPT and Bard. The computational architecture underpinning these chatbots are large language models (LLMs), which are generative artificial intelligence (AI) systems trained on a massive dataset of text extracted from the Web. We conceptualise these LLMs as multifunctional computational cognitive artifacts, used for various cognitive tasks such as translating, summarizing, answering questions, information-seeking, and much more. Phenomenologically, LLMs can be experienced as a “quasi-other”; when that happens, (...) users anthropomorphise them. For most users, current LLMs are black boxes, i.e., for the most part, they lack data transparency and algorithmic transparency. They can, however, be phenomenologically and informationally transparent, in which case there is an interactional flow. Anthropomorphising and interactional flow can, in some users, create an attitude of (unwarranted) trust towards the output LLMs generate. We conclude this paper by drawing on the epistemology of trust and testimony to examine the epistemic implications of these dimensions. Whilst LLMs generally generate accurate responses, we observe two epistemic pitfalls. Ideally, users should be able to match the level of trust that they place in LLMs to the degree that LLMs are trustworthy. However, both their data and algorithmic opacity and their phenomenological and informational transparency can make it difficult for users to calibrate their trust correctly. The effects of these limitations are twofold: users may adopt unwarranted attitudes of trust towards the outputs of LLMs (which is particularly problematic when LLMs hallucinate), and the trustworthiness of LLMs may be undermined. (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...

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)

RIGHT TO SILENCE-UK, U.S, FRANCE, and GERMANY SALLY RAMAGE (TRADE MARK REGISTERED) WIPO Orchid ID 0000-0002-8854-4293 Pages 2-30 Current Criminal Law, Volume 1, Issue 2, -/- Sally Ramage, BA (Hons), MBA, LLM, MPhil, MCIJ, MCMI, DA., ASLS, BAWP. Orchid ID 0000-0002-8854-4293 Publisher & Managing Editor Criminal Lawyer series [1980-2022](ISSN 2049-8047) Current Criminal Law series [2008-2022] (ISSN 1758-8405) and Criminal Law News series [2008-2022] (ISSN 1758-8421). Sweet & Maxwell (Thomson Reuters) (Licensed Annotator of UK Statutes) in annual law books Current Law (...) Statutes Annotated, (2006, 2009, 2010 editions) (sole and separate S&M contracts) for 7 UK Criminal Justice Statutes: UK Safeguarding Vulnerable Persons Act 2006; UK International Development Act 2006; UK Fraud Act 2006; UK Policing and Crime Act 2009, UK Local Democracy Act 2009; UK Bribery Act 2010; and UK Crime and Security Act 2010. TradeMark SALLY RAMAGE ® (2005-2025) in the UK, 2401827; WIPO TradeMark SALLY RAMAGE® 900614; USA TradeMark 3,440,910) (2008-2028). WARNING: The doing of an unauthorized act in relation to a copyright work may result in both a civil claim for damages and criminal prosecution. -/- ABSTRACT The privilege of the right to silence can be traced back to the 12th century, becoming more developed in later centuries. -/- Table of cases European Court of Human Rights Deweer v Belgium [1980], Eckle v Germany [1982], DN v The Netherlands [1975], Funke v France [1993] 16 EHRR 297, JP v Austria [1989], Kansal v UK [2003] Application No 21413/02 Khan v UK [2000] 31 EHRR 45, Saunders v UK [1997] 23 EHRR 313, Schenk v Switzerland [1991] 13 EHRR 242, and Tora Tolmos v Spain [1995] Table of Cases United Kingdom Brown v Procurator Fiscal, Dunfermline, [2000] , The Times, Feb 14, High Ct Scot.; HM Advocate v Friel [1978] SLT; Lambe v Munster [1882] 10 QBD 110.; Pearse v Pearse [1846] Eng Rep 950 (on truth); Re Pantmaenog Timber Co Ltd [2003] HL; R v Bryce [1992] 95 Cr App Rep 320; R v Canale [1990] 2 All ER 187; R v Fitzpatrick [1995] 4 SCR 154; R v Herbert [1990] 2 SCR 151; R V Hertfordshire County Council, ex parte Green Environmental Industries and another [2000] 1 All ER 773; R v Keenan [1989] Crim LR 720; R v Maguire [1990] Cr App Rep 115; R v Pound, Green and ors [2004]; R v Sang [1979] Crim LR 655; R v Saunders [1986] unreported; R v Seelig, R v Spens [1991] CA; R v Smith [1992] 95 Cr App. R. 191; R v Steen and ors [2003] unreported; R v Walsh [1981] CrimLR 822; R V Ward 96 Cr App R [1993]; R v Weekes [1993] Crim LR 211; R v White [1999] 2 SCR 417; R v Young [2003] unreported; Thompson Newspaper v Canada [1990] 1 SCR 425; Walsh Automation (Europe) Ltd v Bridgeman and ors [2002] EWHC 1344 QB France-Law France 1789 Declaration des driots de l‟homme et du citoyen 1808 Code d ‟instruction criminelle 1946 Constitution 1958 Code de procedure penale 2000 Penal Code Germany-Law Germany 1871 Criminal Code 1877 Criminal Procedure Code (St PO) 1877 Judicial Organisation Law 1949 Constitution 1975 Criminal Code 1998 Criminal Code Table of Cases Germany Germany BGH Admissibility of secret tape recordings judgement 14 June 1960, in BGHSt 14, 358 BGH Lie detector judgement 16 Feb 1954, in BGHSt 5, 332. BberfG Decision concerning the applicability of ne bis in idem 31 March 1987, in BberfGE 75,1; BGH Sweeping deception of suspect judgement 24 August 1988, in BGHSt 36, 328. BGH Judgement 24 August 1993, in BGHSt 39, 305 Table of Cases United States United States of America California v Byres [1971] 402 U.S. 424; Fisher v United States 475 U.S. 391 [1986]; Griffin v California [1965] 380 US 609 [; Hoffman v United States [1950] 341 U.S. 479; Supreme Court Baldwin v NY, 399 US 143 [1970]; Supreme Court Chavez v Martinez , U.S. 123 [2003]; Supreme Court Doyle v Ohio, 426 U.S. 610 [1976]; Supreme Court Escobedo v Illinois, 378 U.S. 478 [1964]; Supreme Court Giglio v US, 405 U.S. 150 [1972] Supreme Court Harris v NY, 401 U.S. 222 [1971]; Supreme Court Kirby v Illinois, 406 U.S. 682 [1972]; Supreme Court Miranda v Arizona, 384 U.S. 436 [1966]; Supreme Court Moore v Illinois, 408 U.S. 768 [1972]; Supreme Court Moran v Burbine, 475 U.S. [1986]; Supreme Court Orzoco v Texas, 324 U.S.24 [1969]; Supreme Court US v Hale [1975] 422 US 17; Supreme Court US v Leon, 468 U.S. 897 [1984]; Supreme Court US v Salerno, 481 U.S. 739 [1987]; Twining v New Jersey 211 US 78 113 -/- . (shrink)

Criminal offences and regulatory breaches in using social networking evidence in personal injury litigation Pages 2-7 Current Criminal Law ISSN 1758-8405 Volume 2 Issue 3 March 2010 Author SALLY RAMAGE WIPO 900614 UK TM 2401827 USA TM 3,440.910 Orchid ID 0000-0002-8854-4293 Sally Ramage, BA (Hons), MBA, LLM, MPhil, MCIJ, MCMI, DA., ASLS, BAWP. Publisher & Managing Editor, Criminal Lawyer series [1980-2022](ISSN 2049-8047); Current Criminal Law series [2008-2022] (ISSN 1758-8405) and Criminal Law News series [2008-2022] (ISSN 1758-8421). Sweet & Maxwell (Thomson (...) Reuters) (Licensed Annotator of UK Statutes) in Current Law Statutes Annotated, (2006, 2009, 2010 editions) (sole and separate S&M contracts) for 7 UK Criminal Justice Statutes: UK Safeguarding Vulnerable Persons Act 2006; UK International Development Act 2006; UK Fraud Act 2006; UK Policing and Crime Act 2009, UK Local Democracy Act 2009; UK Bribery Act 2010; and UK Crime and Security Act 2010. Trade Mark SALLY RAMAGE ® 2401827 (2005-2025) in the UK, WIPO Trade Mark SALLY RAMAGE® 900164; USA-(Trade Mark 3,440,910) (2008-2028). WARNING: The doing of an unauthorized act in relation to a copyright work may result in both a civil claim for damages and criminal prosecution. Essentially, Social Network information may not always be admissible in a court of law but may be the door to lead to other, non-contentious evidence. In fact, even illegally obtained evidence may sometimes be admissible evidence, though this is not absolute. Most of the information available on social networking sites will be ‘personal data’. The data controller is the network operator, without whose consent, collecting such information could be an offence, unless it was necessary for the purpose of preventing or detecting crime. (shrink)

In artificial intelligence (AI), responses generated by machine-learning models (most often large language models) may be unfactual information presented as a fact. For example, a chatbot might state that the Mona Lisa was painted in 1815. Such phenomenon is called AI hallucinations, seeking inspiration from human psychology, with a great difference of AI ones being connected to unjustified beliefs (that is, AI “beliefs”) rather than perceptual failures). -/- AI hallucinations may have their source in the data itself, that is, the (...) source content, or in the training procedure, i.e. the way the knowledge was encoded in the model’s parameters, so that errors in encoding and decoding textual and non-textual representations can cause hallucinations. In this paper, we will observe how such errors come to life and how they might be mitigated. For this purpose, we will analyze the usability of justification logics, to behave as a proof checker for validating the correctness of large language models’ (LLM) responses. Justification logic was developed by S. Artemov, and later on mostly by Artemov and M. Fitting, deriving its main idea from the logic of proofs (LP): knowledge and belief modalities are seen as justification terms, i.e. t:X stands for t is a (proper) justification for X. Justification logic originated from attempts to create semantics for intuitionistic logic where proofs were the most proper justifications, but in further development, justification logic could be applied to different kinds of justifications). With the recent attempts to mitigate incorrect LLM responses, we will analyze various guardrails that are currently used for LLM responses, and see how the logic of justification may provide its benefits as an AI safety layer against false data. (shrink)

The complexity and emergent autonomy of Generative AI systems introduce challenges in predictability and legal compliance. This paper analyses some of the legal and regulatory implications of such challenges in the European Union context, focusing on four areas: 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)

The generative AI boom is unleashing its minions. Enterprise software vendors have rolled out legions of automated assistants that use large language model (LLM) technology, such as ChatGPT, to offer users helpful suggestions or to execute simple tasks. These so-called copilots and chatbots can increase productivity and automate tedious manual work. In this article, we explain how that leads to the risk that users' ethical competence may degrade over time — and what to do about it.

This paper delves into the significance and challenges of Artificial Intelligence (AI) ethics and justice in terms of Common Good and Well-being, fairness and non-discrimination, rational public deliberation, and autonomy and control. Initially, the paper establishes the groundwork for subsequent discussions using the Academia Sinica LLM incident and the AI Technology R&D Guidelines of the National Science and Technology Council(NSTC) as a starting point. In terms of justice and ethics in AI, this research investigates whether AI can fulfill human common (...) interests and welfare. Taking AI injustice as an example, I analyze the practical assessment of AI regarding regional, industrial, and social impacts. Further, this paper discusses the challenges of fairness and non-discrimination in AI, specifically addressing the issue of training on biased data, discussing the acquisition of bias by AI and post-processing supervision issues, and emphasizing the importance of rational public deliberation in this process. Then, this research examines the challenges and countermeasures the rational public faces in public deliberation, such as the importance of education in STEM scientific literacy and technological capabilities. Finally, in discussing AI and autonomy, I propose a 'Human-Centered Approach’ rather than relying solely on the 'Technological Utility Maximization' brought by AI to achieve substantial AI justice. -/- Keywords: AI Ethics and Justice, Fairness and Non-Discrimination, Biased Data Training, Public Deliberation, Autonomy, Human-Centered Approach . (shrink)