The Great Success of Artificial Intelligence

Many people have long doubted that AI can really simulate humans. But recent news made me a believer. Making up plausible-seeming but utterly false statements, and asserting them as true with no shame but with total assurance? That's about as human as you can get.

Comments

[Longtobefree]

So chatGPT is a democrat?

[Mr. Bumble]

Beat me to it.

[Á àß äẞç ãþÇđ âÞ¢Đæ ǎB€Ðëf ảhf]

Coulda been any politician, they all assert lies without evidence, although I'm inclined to give George Washington a pass since I think he really did prefer being a farmer over being President.

But they trained it on Democrat data, they get stuck with the label.

[Archibald Tuttle]

LOL. Thanks, that made my day.

[Rev. Arthur L. Kirkland]

I think it is good that guys like you get a moment of respite as you continue to accelerate toward defeat in the culture war, political irrelevance, and replacement.

Everyone deserves a laugh now and again, and relying on Greg Gutfeld, Jeff Foxworthy, and Jason Aldean for smiles must be a tough way to live.

[Jett's Pop]

Not lawyers?

[bevis the lumberjack]

With those qualifications, it’s a certainty that someday soon an AI will be president.

[Mr. Bumble]

No intelligence (artificial or otherwise) in the White House right now.

[Moderation4ever]

What AI is really showing is how unoriginal people are in their verbal, written and visual communications. Original work is difficult and so most people simple borrow from existing sources, something that AI does very well. I am in my late sixties, and it is difficult for me to be surprised by the storyline of a movie or TV show, because I have seen the storyline before. I few years ago I had my will rewritten. Now do I think my lawyer started with a clean sheet of paper? No, they took out a template and dropped my name into the various spots. Do code writer start a new program from scratch? Most writer I know have a library of code they just grab what is needed.

AI works well because so much of people's lives and work is routine. No originality needed, just scan the available work and pick out what is needed.

[taqmcg]

I'm not sure that this shows that originality is difficult. Rather it's that originality is generally undesirable. When I go to a web site I don't want a wildly 'original' organization: I want one that meets the expectations set by having gone to a thousand others. If I write a romance novel, my readers want to see the tropes they love. Usually originality should be carefully limited, typically we want only a single axis of change: we set the novel in an exotic local, or check out a web site discussing a new topic we've never needed or wanted to consider. I'm not a lawyer, but I'd daresay that original arguments are usually a poor substitute for winning ones however cliched.

ChatGPT recognizes the structures of discourse that we use, and populates those structures with characters, events and places based upon similarities between the prompts and its training set.

I asked ChatGPT to summarize my career. It required that I give it my occupation, but then filled in a structure for a typical short biography of a scientist with words that had some relationship to me -- the general field I was in was correct and -- but the details were wildly wrong (though it did get the year of my PhD correct). So in that sense ChatGPT was original... but it's the very originality of it's response that made it less useful.

[mydisplayname]

The guy who led the creation of the lunar excursion module (LEM) also believed that human behavior could be simulated: his 1972 patent (https://patents.google.com/patent/US3748750) still receives attention. But not as much as the LEM.

[Anon]

Shhh…. don’t make it angry.

[OakdaleFTL]

Considerable strides towards passing the Turing Test...doesn't equal true Artificial Intelligence.

[Dan Schiavetta]

Trump’s lies aren’t even “plausible sounding”.

[alexarafat]

AI’s ability to generate confident yet completely fabricated statements is both fascinating and a little unnerving—definitely a very human-like trait. It’s a reminder that while AI can mimic us in many ways, responsible development is crucial to ensure it enhances rather than misleads.

That’s why initiatives like https://humanizeai.now/ are important, focusing on making AI more transparent, ethical, and aligned with human values. AI’s success shouldn’t just be about replication but about meaningful, trustworthy interactions.