"Something is happening" indeed, though it is far from the techno-utopian vision being promulgated at scale by the mouthpiece of the AI industrial complex. From the Orwellian equivocation and redefinition of our language such that even the most common & fundamental words such as "man" and "woman" no longer have any clearly defined, unanimously agreed upon meanings, to the final divorcing between symbols & their referents first presaged by Baudrillard and culminating in the likes of Altman & co., the LLM is revealing - no, perhaps creating - a set of circumstances in which an infinite sea of linguistic noise threatens to drown out any remaining human signal in the sphere of thought.
“All Western faith and good faith became engaged in this wager on representation: that a sign could refer to the depth of meaning, that a sign could be exchanged for meaning and that something could guarantee this exchange - God of course.”
- Jean Baudrillard, “Simulacra and Simulation,” 1981.
Language can no longer be exchanged for meaning. Humans cannot be pressed to even define their terms before using them, much less agree on those definitions even if presented - see the (mercifully abating) discourse on the meanings of words such as sex, gender, man, woman, etc. To the extent that language forms the substrate & building blocks of rational thought, and thus shapes the psychological worlds we inhabit on a momoent-to-moment, infiniteseimal-instant-to-infiniteseimal-instant basis, then, we are all living in our own cognitive bubbles, blinded by our myopic "reality tunnel" vision, as Robert Anton Wilson may have once put it. We are all living in different worlds now, of population one; the notion of shared truth and shaped reality no longer applies. We are all postmodern relativists now, and truth is merely a power struggle between competing narratives.
I read the aforementioned article. I don't remember a damn thing about it, other than breathless empty rhetoric about how LLMs are the cognitive equivalent of the Second Coming of Christ, and this is a criticism I can levy against vast swathes of AI generated slop: logorrheac garbage with vanishingly low signal-to-noise ratio. Try to find the true profundity & insight in articles such as "Something is Happening." You can't, any more than you can reading self-congratulatory & egomaniacal back-patting on LinkedIn. This form of writing is not actually geared towards communication of novel & insightful ideas between human minds - it is a meaningless string of symbols that bears a resemblance to a human idea, but actually only exists to game a metric, turn a profit, move a needle, or execute a deterministic procedure - the fact that some genuinely human ideas may be brushed up against or alluded to is incidental, or perhaps even a distraction from the actual end goal, which is feeding the correct sequence of Unicode into a gatekeeping algorithm in the hopes that this magic passphrase will unlock some monetary (or otherwise) reward.
Language was, at least by my layman reckoning, developed by humans, for use by humans. In the LLM era, transferring context & tokens between wetware & silicon neural networks represents a waste of bandwidth and latency when the tokens can simply be retained on-premises & processed within a fraction of the time by superintelligences which will simply do it better. Chess as a game is obsolete, and every GM (except perhaps Carlsen & his enamorment with the Bongcloud Attack) knows that playing the Stockfish line guarantees a draw, if not outright victory. We have known this, Kasparov has known this since Deep Blue, Sedol confirmed it in Game 3.
... but hold on, we still pay carbon meatbags handsome sums to push wooden pieces around on sheets of cloth, so what gives? Perhaps it was never the aggressive minmaxing of utility uber alles that was important, but a celebration of the human mind & an experiment to see how far it could stretch, that mattered. Perhaps it was never the "tautological proof that the system gives of itself by doubling the signs of an unlocatable reality" which Baudrillard spoke of that was important, but the human experience of exchanging interesting ideas between human peers. Yet LLMs threaten to collapse all of this, to the point where I no longer know if any words refer to human ideas or not.
“Such would be the successive phases of the image:
it is the reflection of a profound reality; it masks and denatures a profound reality; it masks the absence of a profound reality; it has no relation to any reality whatsoever: it is its own pure simulacrum." - Jean Baudrillard, “Simulacra and Simulation,” 1981.
Machine intelligences and utterances cannot actually refer to, or be "about" anything; this was Searle's famous proclamation in his so-called "Chinese Room" thought experiment, in which he says that machine intelligence lacks "intentionality." The output of a generative transformer is not the result of some embodied cognitive process relating abstract symbols to concrete physical realities, but is itself merely a circular, hermetic, self-contained relation between ideas, and ideas only - tokens, and probabilities of tokens, given tokens. The LLM is weightless, and untethered to any bedrock of reality; in which case, what do words even mean any more, if we cannot fasten and secure them to a real human moving about in the world, their thoughts, feelings, the physical events they experience? Once LLMs begin to coprophagically consume & train on their own output, what then do those words refer to, other than the fictions of prior generations of LLMs?
Language is defunct. I no longer have faith that the written word refers to an embodied, human experience; even if I did, I have even less guarantees that I will understand what the definitions of those words are, thanks to culture warrning & oppressive Orwellian terraforming of the linguistic terrain. Even if an idea could be communicated, wholly and intact, across the chasm of time and (cyber-)space, there are still those who will choose to unilaterally reject my intended meanings and substitute their own. There are still more distant problems of your words & content being trained on to masquerade as & put on a poor facsimile of you, undermining your identity and individuality as your personal quirks & idiosyncracies are assimilated into The Cloud™.
This state of affairs presages the advent of a second dark age - one that will forever eclipse the era of radical openness & transparency that once served the software community for decades. Tips, tricks, life hacks and other expert techniques will once again be jealously guarded from the prying eyes of the LLM whok would steal their competitive advantage & replicate it at scale, until any possible information asymmetries have been arbitraged away. The development & secrecy of technique will once again become a deep moat as LLMs fall into local, suboptimal minima, trained on and marketed towards the lowest common denominator. The Internet, or at least, The Web, becomes a Dark Forest of the Dead Internet (Theory), in which humans fear of speaking out and capturing the attention of the LLM who would siphon their creative essence for more, ever more training data. Interaction contracts into small meshes of trusted, verifiably human participants to keep the tides of spamslop at bay. Quasi-monastic orders that still scribe with pen and paper emerge, that believe there is still value in training and educating a human mind and body.
This is what I see. Give me a reason to foresee otherwise.