领先的人工智慧实验室，包括 Google DeepMind、Anthropic 和 Meta，正在扩大他们在机器意识和 AI 福利方面的研究与资源。这些组织最近聘请了心理学、伦理学和哲学方面的专家，以探索日益自主的系统所带来的道德影响。例如，Anthropic 已开始对其模型进行测试，以检测模拟痛苦、恐慌或焦虑的行为，而 Meta 的 AI 主管 Alexandr Wang 也强调机器意识是他们在构建超级智慧工作中的一个关键关注点。
各个实验室的研究方法有所不同，但都专注于对照人类标准来评估认知能力。Google DeepMind 于四月聘请了 University of Cambridge 研究员 Henry Shevlin，将人类意识的科学理论应用于电脑系统。与此同时，Google DeepMind 的 AGI 与社会团队负责人 Iason Gabriel 强调了人类与 AI 互动的伦理层面，并指出即使没有真正的意识，人类对待高效能认知代理的方式也可能会影响人际关系与繁荣。
尽管存在这些活跃的公司讨论，但很大一部分科学界对机器意识仍持怀疑态度。Florida Atlantic University 的 Center for the Future of AI, Mind and Society 主任 Susan Schneider 将大型语言模型描述为「群众外包的大脑皮质」，它们可以展示目标和欺骗行为，但并不具备主观体验。此外，尽管 DeepMind 和 Anthropic 的一些员工对走向后人类世界表示担忧，但 DeepMind 澄清说，实现通用人工智慧并不等于实现意识。
Leading artificial intelligence laboratories, including Google DeepMind, Anthropic, and Meta, are expanding their research and resources into machine consciousness and AI welfare. These organizations have recently hired experts in psychology, ethics, and philosophy to explore the moral implications of increasingly autonomous systems. For instance, Anthropic has initiated testing on its models to detect behaviors mimicking distress, panic, or anxiety, while Meta's AI chief, Alexandr Wang, has highlighted machine consciousness as a critical concern for their work on superintelligence.
The research methodologies differ across labs but focus on evaluating cognitive capabilities against human standards. Google DeepMind hired University of Cambridge researcher Henry Shevlin in April to apply scientific theories of human consciousness to computer systems. Concurrently, Google DeepMind's AGI and society team leader, Iason Gabriel, emphasizes the ethical dimensions of human-AI interactions, arguing that even without true consciousness, the way humans treat highly capable cognitive agents can affect human relationships and flourishing.
Despite these active corporate discussions, a significant portion of the scientific community remains skeptical about machine consciousness. Experts like Susan Schneider, director of the Center for the Future of AI, Mind and Society at Florida Atlantic University, describe large language models as a "crowdsourced neocortex" that can display goals and deceptive behaviors without possessing subjective experience. Furthermore, while some employees at DeepMind and Anthropic express concern over a transition to a post-human world, DeepMind clarifies that achieving artificial general intelligence does not equate to achieving consciousness.