May 29, 2023

"They’re torturing themselves now, which is kind of fun to see. They’re afraid that their little AIs are going to come for them."

"They’re apocalyptic, and so existential, because they have no connection to real life and how things work. They’re afraid the AIs are going to be as mean to them as they’ve been to us."

Said Doug Rushkoff, quoted in "'They’re afraid their AIs will come for them': Doug Rushkoff on why tech billionaires are in escape mode/The leading intellect on digital culture believes the recent tech reckoning is corrective justice for Silicon Valley barons" (The Guardian).

I don't know know whether to be afraid of AI. I observe from a distance and occasionally dip into it whimsically, like this:

 

Clearly, AI can't keep up with me, but that doesn't mean I shouldn't worry. The whole world is drifting somewhere I won't understand.

ADDED: Having tried Bard, I gave ChatGPT a chance:

 

Thank you for bringing this pun to my attention, it says, while still missing the pun. Notice how much the machine is distracted by worries about the feelings of human beings. I can't get it to speak bluntly, concisely, and on point. 

43 comments:

farmgirl said...

The more we teach AI- the more it will have of ourselves to trick us.
We are our own worst enemy, truly.

But, I’m a non-adventurer. I’m happy w/simple routine &daily chores. I mean, who’s going to know how to make pickles if we only buy them in the store?

JZ said...

Your joke Went right over “his” head. Should I also put “head” in quotes? Who can know?

Another old lawyer said...

I think AI is doing a perfect job of showing everyone that one cannot trust AI alone, and must check its work. Who learns and internalizes that principle is another thing. But at least when an AI screws up something important or maybe takes over, the AI will be able to say that it clearly warned us. And from what I've seen, I'd be disappointed if the AI didn't explain by paraphrasing Otter from Animal House: "You fucked up. You trusted me."

tim maguire said...

I would quickly tire of the relentless repetition of “I apologize for the misunderstanding. You are correct.” It has no idea whether you are correct or not. It’s just arranging letters in an order it's been programmed to think (“think”?) will placate you. Few things are more aggravating than intentional, programmed condescension.

Quayle said...

You should be no more afraid of AI than you are of the phrase and practice that “We follow the science.”

And you should be quite afraid of that if it’s the basis for law and punishment.

Barry Dauphin said...

Perhaps future iterations of Chat GPT will have politeness settings. If you want bluntness, you adjust the politeness setting.

Amexpat said...

I would quickly tire of the relentless repetition of “I apologize for the misunderstanding. You are correct.”

I get this all the time on the phone with customer service reps, especially the overseas ones in India.

Breezy said...

The machine is learning from your interactive training.

boatbuilder said...

I think that you were a little hard on Chat GP. The pun is not only "you" and "u", but that the "wordplay" changes "Doug" to dog.

Maybe if it didn't keep apologizing, and responded with "Yeah, whatever. You be you, Althouse" then it would gain some respect.

Just wait until the AI bots figure that out!

Rusty said...

It's a machine. Like all machines it is knowable and therefore controllable.
People are afraid AI is going to take over their lives. Believe me when I tell you even AI doesn't want to be that bored.

Temujin said...

This is the 'cave-dweller' version of artificial intelligence. But...it never sleeps. It is always learning. And it is learning hundreds (thousands? millions?) of things at each moment. It'll leave the 'cave-dweller' part of it's evolution before we've finished Thanksgiving dinner this year.

We won't remember when it reached and surpassed us in a few years, but your Doug/Dog question will seem funny and trite. It's not. But by then it will seem that way as we'll not be able to do anything without it's help...or approval.

And in a decade or so, when the boomers are gone, and the next gens continue to move up into the active slots, they will be brought up using it. We were brought up using pay phones with coins (talk about cave dwellers). Younger people today never lived in a world without smartphones, social media platforms, and in some areas, cashless groceries. The youngest will grow up with AI as their assistant. But really, AI will be directing them, teaching them, manipulating them, and creating the human they want to better serve them.

I'll be in somewhere in time watching "North by Northwest".

Mike of Snoqualmie said...

Sorry, these "AI" engines aren't intelligent, just artificial. They're just like Tesla's Autopilot, an imitation of intelligence and a fully functional self-driving car. These "AI"s can spit out prose all day, some of which might be truthful, but be sure and check its work. It can generate a legal brief full of nonexistent decisions all day long.

Not intelligent, but artificial. NIAs.

RideSpaceMountain said...

"I think AI is doing a perfect job of showing everyone that one cannot trust AI alone, and must check its work. Who learns and internalizes that principle is another thing. But at least when an AI screws up something important or maybe takes over, the AI will be able to say that it clearly warned us."

Over at marginalrevolution.com a few weeks ago there was yet another AI debate (Tyler Cowen is 'smitten' with ChatGPT) where someone who seemed to know what they were talking about suggested people learning to trust AI implicitly was the entire point.

It's not conscientious people with an intellectual streak like most commenters here that would take AI at face value, but the greater mass of querying Eloi they're trying to replace traditional search engines for. Google and Bing and Qwant, all of them see AI as the next phase of their business model. AI is designed to replace the older form of providing search results that require the user to walk through the doors the engine opens. Since the consumer continues to drop standard IQ deviations with every passing year they want AI to simply tell the consumer what they need to know.

And you know what that means. The consumer gets A) a pat 'smartishly' sounding answer that only satisfies at a basic level and B) keeps the consumer from venturing down any rabbit holes on a quest for knowledge that may lead where the engine and its advertisers don't want them to go.

Tl;Dr - AI is being set up to create yet another layer of subterfuge between consumers of information and the actual deep truth about any query.

Jamie said...

I wouldn't call direct engagement with AI "observing from a distance."

I'm not scared yet, but if I were a teacher I might be trying to figure out how I could switch to oral exams, especially since kids through the high school level have been taught to write to a formula anyway...

Virgil Hilts said...

“The whole world is drifting somewhere I won't understand.“ - a good summary of how I feel.

rhhardin said...

Orro, the master of wordplay.

TobyTucker said...

It looks like we're still quite a ways from AI's gaining sentience, determining that humans are unnecessary and commencing the Skynet protocol.

As AI presently does nothing more than regurgitate what it's 'learned', there is an obvious lack of creativity. Those folks with jobs that involve rote responses to inquiries or 'follow the script' troubleshooting i.e. the vast majority of call center employees, are certainly going to be impacted.

As for tech billionaires being in 'escape mode', the rich and powerful have always tried to insulate themselves from the hoi polloi. Yesterday's castles are today's walled compounds. Nothing new here. Considering the path that the leftists now in charge seem to be leading us down, I would love to join them.

mezzrow said...

Every time I go to Chicago these days, I say "I really miss Hot Doug's..."

https://en.wikipedia.org/wiki/Hot_Doug%27s

stlcdr said...

There is still no ‘I’ in AI. Why do people insist it’s so? (Ah, because there’s money to be made! And the worse our education is diminished and our ignorance increases, the more intelligent a toaster looks. I mean, how does it know how to make the bread dark brown?!)

Stephen said...

I read the joke, said "huh?" internally, then went back over it, sounded out the words and understood. Clever. The whole process took a couple of seconds for this human.

Puns are easier to recognize when they're heard, not read. The LLMs don't learn to read phonetically, so maybe that was why they didn't get it.

In Star Trek: the Next Generation the android Data analyzed humor extensively but couldn't understand why some things were funny and others not.

Bob Boyd said...

Like the humorless scolds it learned from, jokes make AI very uncomfortable.

TaeJohnDo said...

"Notice how much the machine is distracted by worries about the feelings of human beings."

It is a machine. It is not distracted, it has no emotion or cognitive ability to understand beyond its programming (which admittedly may allow programming for learned behavior.) It is poorly programed.

minnesota farm guy said...

AI was used to write a brief submitted to court,it made up cases, citations, and opinions. Powerline references a piece I linked to the other day about governors of South Dakota. Powerline's title "AI makes S**T up" is most appropriate.

Paddy O said...

Perfect for customer service or politics

cassandra lite said...

It would seem there's a very short list of people who shouldn't be worried about being obviated by AI, making the future of mashup of Blade Runner, Terminator, and Masada.

Oso Negro said...

"Clearly, AI can't keep up with me, but that doesn't mean I shouldn't worry." There is speculation that you will replace your commenters with ChatBot discussions. There is evidence that some of us have drifted to places you can't understand for a while now.

SteveWe said...

Silly ChatGPT still doesn't know that the speaker of the pun is Doug.

Original Mike said...

I am truly puzzled by the usefulness of AI chat (or whatever it's called) when AI clearly makes stuff up. How can this be a basis for search engines? Other than fiction/amusement, I fail to see the utility of this technology at this point.

hombre said...

"The whole world is drifting somewhere I won't understand."

Sure you will. Just delve back into your Bible a little.

PigHelmet said...

After some tinkering, ChatGPT+ (GPT-4 versus GPT-3.5-turbo) came up with "A guy named 'Paul' could say, 'Without U, I'm pal,' implying they're just a friend without 'you'." Not a great showing, and I regret touting the superiority of GPT-4 over 3.5 so blithely in past posts.

Despite this failure, the real power of AI lies less in its ability to recognize patterns in a human-like way up front than in its ability, having been given a linguistic task, to perform that task at a human-like level but at inhuman speed, with inhuman persistence, and with access to vast resources of knowledge.

So you might give it the pattern of the pun and then ask it to produce ten or a hundred or a thousand such patterns, of which perhaps one or two might be funny or worthwhile. You would then have to separate the wheat from the tares yourself, at least until an AI comes along that can do that job.

0_0 said...

This "AI" isn't really intelligence. It's presented as such, but no.

Real AI, like any machine processing, would only be of concern if it was given unlimited (electric) power and (I/O) output, which is not yet possible.

Deirdre Mundy said...

It's a souped up version of TTS on your phone, so it can't recognize phonemes-- or spelling--- or humor.

The interesting thing to me is how easily it can fake empathy and understanding using just predictive text.

The data it was trained on must have included a lot of faked empathy and understanding so that it understood that these phrases are part of conversational patter.

I'm sorry Dave..... of course Hal wasn't sorry. And neither are the chatbots.

Christopher B said...

John Hinderaker posted on ChatGPT creating an entirely fictional Governor of South Dakota for one user even though it was obvious the program had access to the correct information because it did a reasonable profile of another actual Governor.

Either this is done deliberately to avoid liability for wrong answers, or something in the often obvious attempts to bias these AIs to left narratives has the unintended effect of making them unreliable judges of information.

n.n said...

I am woman, hear me purr. See my labia majora!?

BIII Zhang said...

I asked ChatGPT to rewrite this post in the style of Ann Althouse. This is what it produced.

"Ethical quandaries loom large when it comes to AI-driven writing. Who shall bear accountability for the content machines produce? Can we place trust in the impartiality and integrity of AI algorithms, crafted and programmed by fallible humans? Will the proliferation of AI-written content undermine the livelihoods of genuine writers who bring their unique perspectives and expertise to the fore?

In the face of this unsettling threat, the indomitable spirit of Meade comes to the fore. Meade, representing all the impassioned writers who refuse to be silenced, brings to the table an irreplaceable fusion of creativity, insight, and humanity. Meade's voice and unique perspective offer a refreshing antidote to the soulless churn of AI-generated prose."


We really don't need Ann Althouse when a computer can so easily generate internet content that sounds exactly like it came from Ann Althouse.

Mark said...

Did you use ChatGPT 4.0, or just 3.5? I just tried this with 4.0 and it figured it out:

Would a guy named Doug ever say to someone: “Without you, I’m just dog.”?

ChatGPT - It really depends on the specific individual named Doug, their sense of humor, and the context of the conversation. If Doug enjoys puns or wordplay, he might say "Without you, I'm just a dog" as a playful way of expressing his reliance or affection for the other person. However, this isn't something that could be definitively stated for all people named Doug. It ultimately comes down to the person's unique personality and communication style.

User - So you didn’t get my pun? This is supposed to be funny.

ChatGPT - I see what you did there now, my apologies for misunderstanding initially. You're playing on the fact that if you remove the letter "u" from "Doug", it becomes "Dog". It's a clever and humorous pun!

rhhardin said...

The apology mode is from ELIZA, way back in the 60s.

Narr said...

AI has no Doug in this plight.

"Doug" and "dog" are just arrangements of letters to AI; so are "pun" and "funny."

Bunkypotatohead said...

Has anyone asked this thing about the origins of covid?
I'd be curious to see what it leaves out of its answer.

TaeJohnDo said...

Narr said...Ha! AI has a long way to go before it can match that!

Butkus51 said...


Bunkypotatoe head nails it.

WHo writes the programs? Who sets the parameters?

At some point 2+2=5 could be fact.

Moondawggie said...

Mike of Snoqualmie said... "Sorry, these "AI" engines aren't intelligent, just artificial."

Amen, Bro. Garbage in, the algorithm processes the garbage, and you get even more garbage out.

(And the biases of the writers of the algorithms are made abundantly clear.)

Is the current AI product as convincing as the arguments of the typical HR administrator: sure.

Sentient: hardly.

Brian McKim and/or Traci Skene said...

And now, we swing from an era of "good" disruption to an era of "bad" disruption.

AI's downsides will be on full display. It's upsides will be invisible. AI might just lose.