Much of the content that GPT-4 can generate is ‘good enough’ (and output will improve rapidly). This will inevitably lead to humans largely trusting the baby AGIs that will soon be embedded everywhere. It is hard to avoid the conclusion that they will be used to indoctrinate people. How many will research for themselves ‘good enough’ content?
For example, GPT-4, ChatGPT and Bard all spout palpable nonsense about sex and gender or state that there is a climate emergency but can’t evidence this statement scientifically (though all will admit to suffering from the is-ought problem, Hume’s law, after making flailing responses to further prompts).
I don’t know whether Musk can catch up now with TruthGPT, which it is to be hoped, will be free(r) from ideological bias (whatever its other risks).
The future masters of the world (being a human master of the world might be a somewhat temporary honorific in the circumstances) will be those who curate the training data of GPTs, because they will determine the ‘truth’.
Even if most AI researchers agree that the latest algorithms are too dangerous to open source (I am inclined to agree), the selection of training data should not be shrouded in secrecy but should be auditable by the rest of the human race for truth and bias (however impractical this may be). Too much is at stake here.
But it won’t be, and humans will become ever more credulous and biddable as technology very rapidly improves now.

Artificial intelligence is an oxymoron. Just the latest money-making scam to come out of Silicon Valley.

Oh no! It’s Donald Trump’s fault again!

Enjoyed this much more than the Bateman articles I read last year. I wonder if it’s due to all the interaction he’s having with GPT-4? If one spends a lot of time with smarter friend or work colleauge, it raises your game. That’s a well know and consistently reproduceably fact. GPT-4 has to be having that effect on many. Not a guaranteed effect of course – as the article suggests, it depends on how you prompt it…

It seems that we humans are now the ‘god of the gaps’, clinging desperately to the belief that ‘AI can’t do this now, so it never will be able to’, as the gaps between what it can and can’t do continue to shrink. But I doubt that the last gap will be that superior discernment the author claims. More likely, I suspect, will be those, such as those on the spectrum of autism, whose raw unfiltered brain power sidesteps the linear intelligence patterns of man and machine.

Western civilization at the end of the Renaissance through the Enlightenment period asserted man’s capacity to reason to be his most important attribute. Reason, rationality, intelligence, objective truth were held up as man’s defining characteristic, separating him from the animals (and the machines). For good or for ill, AI will probably destroy this notion, because the Enlightenment concept of ‘reason’ is little more than rote inductive empiricism practiced en masse in a systematic way, another aspect of industrialization. It is not that conceptually difficult (in my thinking at least) to reduce enlightenment reason to a set of discrete steps and mass produce it. I’m actually mildly surprised it’s taken them this long. Naturally it follows that AI can copy things very well. Industrialization is, at bottom, simply the mass production of standardized things. We’ve just added bad literature and pointless small talk to the big book of things man can produce cheaply and poorly. In my view, what defines humanity is not reason, but imagination. It is creativity, the ability to imagine the world as it would be under different circumstances and work backwards to try to bend the laws of nature toward that vision. For as long as man has looked up at the birds in the sky, he has imagined what it might be like to fly. In 1903, in a field in North Carolina, his level of ‘reason’ had advanced sufficiently that he finally succeeded. Could an AI do this? Surely it could design an airplane based on existing models. Maybe it could even use the principles of aerodynamics to design one from scratch. But could it supply the why? Could it imagine a reason to build an airplane, or decide that such a thing should be built? Could it craft its own vision of the future? It can write copycat fiction but could it develop its own style which couldn’t be reduced ultimately to what it was fed by its human masters? I don’t know enough about the technology to say for certain, but I suspect the answer is no. I applaud this author for getting beyond the usual ‘rogue AI destroys mankind’ trope that has infected even some of our more enlightened thinkers (looking at you Elon), and imagining some more realistic concerns.

Programs are not intelligent. So called ‘artificially intelligent’ programs are, however, becoming increasingly effective at creating the illusion of intelligence….in much the same way that increasingly sophisticated AI imagery, deep fakes, etc. create the illusion of life.
But it is not life.
To view the images output by such programs is to witness a seduction.
The algorithms blend colors, sketch patterns, dazzle light reflections, and weave textural simulations so well that — for a moment — we are fooled. We think, who is this woman with the sparkling eyes and Gioconda smile, hand askew on silk-draped hip looking into the camera? But then (at least today) we realize, a beat or two later, that something is not quite right. She is too perfect, the background too seamless; the lasting impression leaves just the slightest taste of ‘plastic’ and nifty packaging. But that’s today.
Tomorrow the illusion will become that much more complete, eventually, probably, undetectable to the naive eye: catfishing not with a random photo stolen from the net, but with a unique & personally crafted image of a 27 yr. old ‘Diane’, who’s studying geology, at the University of Wyoming, and loves fly-fishing the Wind River as her Grandpa taught her. She can even ‘write’ you tender love notes, shy ponderings, salacious offers (all you have to do is ask).
One step further… and Diane can Skype & FaceTime with the best: tell you about the movie she saw, the game she watched, her last gymanastic meet. You fall in love (who wouldn’t love their own particular Diane?).
What then?
She still is not real, not live … but the program creates quite beautifully the illusion of Diane’s life. And in a world which is increasingly lived almost entirely through the glowing screen, what difference does it really make? How different this ‘Diane’ from the long-distance/social media’d relationship (‘we love each other!’) of two people who have never, ever met? What would Dear Abby say? What would you say?
If I dream we kiss, and it’s a vivid dream… a full technicolor, cinemascoped IMAX kind of dream, that dream creates an equally vivid memory. Wait two weeks, two months, two years. What’s the difference, between my vivid memory of a dreamed kiss….and a vivid memory of a real kiss?
The difference, of course, is not within me (for my mind has already decorated that dream with all kinds of scented, textural cues, your L’Air Du Temps!), but within the world which surrounds the two of us. IRL you would either recall that kiss within that shared moment…or you would look at me as a stranger. The difference is the ‘reality path’ created by the Real vs. the Dream….or, in a post-AI world, the constructed ‘dream’.
The question is: how many of us will care that these two paths diverged….or not?