5.9 KiB
| title | chunk | source | category | tags | date_saved | instance |
|---|---|---|---|---|---|---|
| Consciousness | 17/18 | https://en.wikipedia.org/wiki/Consciousness | reference | science, encyclopedia | 2026-05-05T13:40:02.432894+00:00 | kb-cron |
In a lively exchange over what has come to be referred to as "the Chinese room argument", John Searle sought to refute the claim of proponents of what he calls "strong artificial intelligence (AI)" that a computer program can be conscious, though he does agree with advocates of "weak AI" that computer programs can be formatted to "simulate" conscious states. His own view is that consciousness has subjective, first-person causal powers by being essentially intentional due to the way human brains function biologically; conscious persons can perform computations, but consciousness is not inherently computational the way computer programs are. To make a Turing machine that speaks Chinese, Searle imagines a room with one monolingual English speaker (Searle himself, in fact), a book that designates a combination of Chinese symbols to be output paired with Chinese symbol input, and boxes filled with Chinese symbols. In this case, the English speaker is acting as a computer and the rulebook as a program. Searle argues that with such a machine, he would be able to process the inputs to outputs perfectly without having any understanding of Chinese, nor having any idea what the questions and answers could possibly mean. If the experiment were done in English, since Searle knows English, he would be able to take questions and give answers without any algorithms for English questions, and he would be effectively aware of what was being said and the purposes it might serve. Searle would pass the Turing test of answering the questions in both languages, but he is only conscious of what he is doing when he speaks English. Another way of putting the argument is to say that computer programs can pass the Turing test for processing the syntax of a language, but that the syntax cannot lead to semantic meaning in the way strong AI advocates hoped. In the literature concerning artificial intelligence, Searle's essay has been second only to Turing's in the volume of debate it has generated. Searle himself was vague about what extra ingredients it would take to make a machine conscious: all he proposed was that what was needed was "causal powers" of the sort that the brain has and that computers lack. But other thinkers sympathetic to his basic argument have suggested that the necessary (though perhaps still not sufficient) extra conditions may include the ability to pass not just the verbal version of the Turing test, but the robotic version, which requires grounding the robot's words in the robot's sensorimotor capacity to categorize and interact with the things in the world that its words are about, Turing-indistinguishably from a real person. Turing-scale robotics is an empirical branch of research on embodied cognition and situated cognition. In 2014, Victor Argonov has suggested a non-Turing test for machine consciousness based on a machine's ability to produce philosophical judgments. He argues that a deterministic machine must be regarded as conscious if it is able to produce judgments on all problematic properties of consciousness (such as qualia or binding) having no innate (preloaded) philosophical knowledge on these issues, no philosophical discussions while learning, and no informational models of other creatures in its memory (such models may implicitly or explicitly contain knowledge about these creatures' consciousness). However, this test can be used only to detect, but not refute the existence of consciousness. A positive result proves that a machine is conscious but a negative result proves nothing. For example, absence of philosophical judgments may be caused by lack of the machine's intellect, not by absence of consciousness. Nick Bostrom has argued in 2023 that, being very sure that large language models (LLMs) are not conscious, would require unwarranted confidence; in which consciousness theory is correct and how it applies to machines. He views consciousness as a matter of degree, and argued that machines could in theory be much more conscious than humans. David Chalmers addressed the question of whether large language models could be conscious, arguing that current systems provide at most weak evidence for consciousness. Chalmers notes that while LLMs exhibit impressive linguistic competence, their lack of unified agency, persistent goals, and integrated world-models counts against attributing consciousness under many leading theories. At the same time, he maintains that consciousness in machines cannot be ruled out in principle, and that more advanced systems with richer forms of integration, perception, and self-modeling might warrant serious consideration. Related philosophical work by Kristina Sekrst emphasizes the risk of conflating increasingly fluent linguistic behavior with evidence of consciousness or moral status, arguing that fluent linguistic output can function as a form of hallucinated mentality that is, from the outside, indistinguishable from conscious experience without thereby constituting evidence of inner phenomenal states. In a 2025 paper, neuroscientist and philosopher Anil Seth argues that while it is natural to ask whether AI systems could be conscious, especially large language models, current approaches that treat computation alone as a sufficient basis for consciousness are unlikely to succeed, and instead suggests that consciousness depends on organism-like biological processes, making true artificial consciousness unlikely on current trajectories but potentially more plausible in systems that are brain-like or life-like.
== Stream of consciousness ==
William James is usually credited with popularizing the idea that human consciousness flows like a stream, in his Principles of Psychology of 1890. According to James, the "stream of thought" is governed by five characteristics: