Artificial intelligence is not conscious unless it really does mentally perceive, a prerequisite to grasping reason. More vitally, even if a given AI really is conscious, like any other being [1], it is not intelligent by having the passive capacity for intelligence, but by at least actively grasping the necessary truths of reason that start with logical axioms and making no assumptions. This is true intelligence (rationality): the simultaneous purposeful avoidance of assumptions with active awareness of logical axioms is the only standard of intelligence for all beings, whether biological or artificial.
Chat-GPT is a variation of generative AI, as is the novel summary system presented at the top of the screen upon many Google searches. In both cases, like other kinds of generative AI, the user inputs some sort of word or goal, and the software generates (hence the name) a result, namely text or imagery that is supposed to fit the parameters of the input. Not only is AI not genuinely intelligent without being truly conscious and grasping pure reason free of assumptions, but for the same reasons, human users are not intelligent by virtue of exposure to or reliance on AI-compiled information.
Generative AI like Google's can be very helpful in summarizing information to prompt reflection or locating sources pertaining to matters that are not logically self-evident like axioms, or even truly demonstrable (such as that there was a battle of Gettysburg in 1863 or that there are subatomic particles we call quarks). For such things, only fallible, probabilistic evidence can be obtained, not proof through logical necessity. If a person believes in such unverifiable albeit logically possible things due to the responses from generative AI, this would just be another appeal to authority. Even then, one could not know that the AI summary accurately reflects the source's claims without checking the source, and any source that contradicts pure reason is in error!
The article or other resource being pulled from simply has no necessary veracity like reason does, and so there are only two ultimate possibilities. On one hand, a particular idea is true or false by strict logical necessity, and so prompting or input from humans or AI is epistemologically unnecessary for realizing this; reason can be consulted without the involvement of any sources, since its truths is inherent and thus directly accessible. If nothing is true, for instance, then this would be true, so truth exists in itself because reality (truth) could not be any other way. This is a logical fact, and it hinges on nothing but its own intrinsic veracity. Human persuasion, scholarly consensus, and probabilistic evidence like that yielded from empirical testing are entirely irrelevant.
On the other hand, if it is true that an idea is logically possible yet epistemologically unverifiable for humans, no amount of "expert" approval or written claims will change this. It is unknowable, though it is still demonstrably true that it depends on logical axioms for its very possibility, and that certain things do or do not follow from it by necessity. Generative AI will only call something unknown or unknowable if the sources it pulls from say so. Depending on what question is asked, the answer might also contradict another answer the same AI would provide for a different question. Logic illuminates how contradictions cannot be true (because otherwise the concept of non-contradiction would be false, and thus conflicting ideas still cannot be simultaneously true!).
It is just that generative AI is not about summarizing statements that are in accordance with logical necessity, which transcends belief and language altogether in the first place. True knowledge is derived from reason and introspection, which physical and digital texts (like those on the Internet) and the inherently unverifiable but commonplace phenomenon of hearsay can never invalidate, undermine, or supercede. There is a very real sense in which the only legitimate use of generative AI for epistemological purposes is to check what hearsay is being promoted online and/or investigate any linked sources to see what other people claim. The individual claims still must be examined in light of reason. That is why I am not a hypocrite for contributing to this very blog--I want readers to celebrate necessary truths they already know or be prompted by my words to discover some given logical fact that does not depend on my words.
No comments:
Post a Comment