Computo Ergo Non Sum

If René is to be considered the father of any tangible measure of consciousness with his locus classicus "I think, therefore I am," that then suggests one of two realities:

Sentient or not is up for grabs, but what is evident from the collective fraternization between humans and LLMs is that this whole shindig is shaping up to be one of the largest attributions of personhood to a tool in human history.

Eliza & Parry escape the lab

In 1966, Weizenbaum, an MIT computer scientist, deployed ELIZA: a simple chatbot, designed to regurgitate what a user inputs within it to a generic response. For instance:

Instead of recognizing its simplicity and its more tangible use cases, users grew to confide in ELIZA, especially through a particular script within the program called DOCTOR, which simulated a psychotherapist. This willingness, knowingly or otherwise, to emotionally attach to artificially intelligent chatbots that exhibit any semblance of "care" through mere strings of words has come to be called the ELIZA effect.

Similarly, PARRY, a chatbot developed by psychiatrist Kenneth Mark Colby in 1972, set out to mimic a patient exhibiting symptoms of paranoid schizophrenia. It was able to confuse some clinical psychiatrists as to who was speaking: a human patient with schizophrenia or PARRY.

What Parry's ability to bypass clinical discernment and the ELIZA effect goes to show is that even nascent AI systems were able to get the better of humans anthropomorphically and expose how quickly they infer minds from language.

Enter status quo.

Memory mirage & the dubious dub

Much has Moore-ishly changed since the archaic chatbots of the 60s and 70s. ELIZA and her counterparts are now ubiquitous and wear tools around their belts that make their illusion of personhood that much more convincing.

Memory, for starters. A confidant is what begins to be seen as opposed to a machine when one's profession, family, likes, dislikes, proclivities, and tendencies are regurgitated by AI. Not due to memory begetting sentience, but because it has long been used as an indicator of personhood.

Nomenclature, for seconds. The feature of being able to dub and name chatbots, akin to the way pets and loved ones are named, increases the anthropomorphism of Artificial Intelligence. Names are not usually assigned to a tool; they're assigned to a social presence.

Continuity, for thirds. The daily occurrence and, frankly, reliance on AI inevitably begin to resemble a relationship between the AI and the user, by virtue of repeated points of access and communication. Continuity serves as the main lever in one key emotion arising in the user: familiarity. Humans come back again and again to the familiar and safe, because, well...it's just that: familiar and safe.

The amalgamation of these attributes is eerily human-esque, as per the ELIZA effect, offering any given user of AI enough markers of personhood to infer sentience from lines of code.

Impressive lines of code. But lines of code nonetheless.

Homo sapiens' final say

Ironically, the argument of this piece, that language naturally invites the inference of a speaker, was proved while its research was carried out.[1]

Modern LLMs are terrific with lingo. Language has historically been humanity's most reliable indicator of mind. If something spoke coherently, recalled prior interactions, exhibited a degree of consistency, and responded proportionally to emotional stimuli, there was almost certainly someone there. After having served humans remarkably well for the bulk of history, language suddenly stopped being foolproof.

To slightly combat and proof against the fool, it is well-advised that this is remembered vividly as interactions with AI grow more frequent: memory, names, continuity, familiarity, and companionship are no longer attributes reserved solely for humans; they now belong to machines too.