#Intelligent Chatbots | Explore Tumblr posts and blogs

mostlysignssomeportents · 1 year ago

Text

How plausible sentence generators are changing the bullshit wars

This Friday (September 8) at 10hPT/17hUK, I'm livestreaming "How To Dismantle the Internet" with Intelligence Squared.

On September 12 at 7pm, I'll be at Toronto's Another Story Bookshop with my new book The Internet Con: How to Seize the Means of Computation.

In my latest Locus Magazine column, "Plausible Sentence Generators," I describe how I unwittingly came to use – and even be impressed by – an AI chatbot – and what this means for a specialized, highly salient form of writing, namely, "bullshit":

https://locusmag.com/2023/09/commentary-by-cory-doctorow-plausible-sentence-generators/

Here's what happened: I got stranded at JFK due to heavy weather and an air-traffic control tower fire that locked down every westbound flight on the east coast. The American Airlines agent told me to try going standby the next morning, and advised that if I booked a hotel and saved my taxi receipts, I would get reimbursed when I got home to LA.

But when I got home, the airline's reps told me they would absolutely not reimburse me, that this was their policy, and they didn't care that their representative had promised they'd make me whole. This was so frustrating that I decided to take the airline to small claims court: I'm no lawyer, but I know that a contract takes place when an offer is made and accepted, and so I had a contract, and AA was violating it, and stiffing me for over $400.

The problem was that I didn't know anything about filing a small claim. I've been ripped off by lots of large American businesses, but none had pissed me off enough to sue – until American broke its contract with me.

So I googled it. I found a website that gave step-by-step instructions, starting with sending a "final demand" letter to the airline's business office. They offered to help me write the letter, and so I clicked and I typed and I wrote a pretty stern legal letter.

Now, I'm not a lawyer, but I have worked for a campaigning law-firm for over 20 years, and I've spent the same amount of time writing about the sins of the rich and powerful. I've seen a lot of threats, both those received by our clients and sent to me.

I've been threatened by everyone from Gwyneth Paltrow to Ralph Lauren to the Sacklers. I've been threatened by lawyers representing the billionaire who owned NSOG roup, the notoroious cyber arms-dealer. I even got a series of vicious, baseless threats from lawyers representing LAX's private terminal.

So I know a thing or two about writing a legal threat! I gave it a good effort and then submitted the form, and got a message asking me to wait for a minute or two. A couple minutes later, the form returned a new version of my letter, expanded and augmented. Now, my letter was a little scary – but this version was bowel-looseningly terrifying.

I had unwittingly used a chatbot. The website had fed my letter to a Large Language Model, likely ChatGPT, with a prompt like, "Make this into an aggressive, bullying legal threat." The chatbot obliged.

I don't think much of LLMs. After you get past the initial party trick of getting something like, "instructions for removing a grilled-cheese sandwich from a VCR in the style of the King James Bible," the novelty wears thin:

https://www.emergentmind.com/posts/write-a-biblical-verse-in-the-style-of-the-king-james

Yes, science fiction magazines are inundated with LLM-written short stories, but the problem there isn't merely the overwhelming quantity of machine-generated stories – it's also that they suck. They're bad stories:

https://www.npr.org/2023/02/24/1159286436/ai-chatbot-chatgpt-magazine-clarkesworld-artificial-intelligence

LLMs generate naturalistic prose. This is an impressive technical feat, and the details are genuinely fascinating. This series by Ben Levinstein is a must-read peek under the hood:

https://benlevinstein.substack.com/p/how-to-think-about-large-language

But "naturalistic prose" isn't necessarily good prose. A lot of naturalistic language is awful. In particular, legal documents are fucking terrible. Lawyers affect a stilted, stylized language that is both officious and obfuscated.

The LLM I accidentally used to rewrite my legal threat transmuted my own prose into something that reads like it was written by a $600/hour paralegal working for a $1500/hour partner at a white-show law-firm. As such, it sends a signal: "The person who commissioned this letter is so angry at you that they are willing to spend $600 to get you to cough up the $400 you owe them. Moreover, they are so well-resourced that they can afford to pursue this claim beyond any rational economic basis."

Let's be clear here: these kinds of lawyer letters aren't good writing; they're a highly specific form of bad writing. The point of this letter isn't to parse the text, it's to send a signal. If the letter was well-written, it wouldn't send the right signal. For the letter to work, it has to read like it was written by someone whose prose-sense was irreparably damaged by a legal education.

Here's the thing: the fact that an LLM can manufacture this once-expensive signal for free means that the signal's meaning will shortly change, forever. Once companies realize that this kind of letter can be generated on demand, it will cease to mean, "You are dealing with a furious, vindictive rich person." It will come to mean, "You are dealing with someone who knows how to type 'generate legal threat' into a search box."

Legal threat letters are in a class of language formally called "bullshit":

https://press.princeton.edu/books/hardcover/9780691122946/on-bullshit

LLMs may not be good at generating science fiction short stories, but they're excellent at generating bullshit. For example, a university prof friend of mine admits that they and all their colleagues are now writing grad student recommendation letters by feeding a few bullet points to an LLM, which inflates them with bullshit, adding puffery to swell those bullet points into lengthy paragraphs.

Naturally, the next stage is that profs on the receiving end of these recommendation letters will ask another LLM to summarize them by reducing them to a few bullet points. This is next-level bullshit: a few easily-grasped points are turned into a florid sheet of nonsense, which is then reconverted into a few bullet-points again, though these may only be tangentially related to the original.

What comes next? The reference letter becomes a useless signal. It goes from being a thing that a prof has to really believe in you to produce, whose mere existence is thus significant, to a thing that can be produced with the click of a button, and then it signifies nothing.

We've been through this before. It used to be that sending a letter to your legislative representative meant a lot. Then, automated internet forms produced by activists like me made it far easier to send those letters and lawmakers stopped taking them so seriously. So we created automatic dialers to let you phone your lawmakers, this being another once-powerful signal. Lowering the cost of making the phone call inevitably made the phone call mean less.

Today, we are in a war over signals. The actors and writers who've trudged through the heat-dome up and down the sidewalks in front of the studios in my neighborhood are sending a very powerful signal. The fact that they're fighting to prevent their industry from being enshittified by plausible sentence generators that can produce bullshit on demand makes their fight especially important.

Chatbots are the nuclear weapons of the bullshit wars. Want to generate 2,000 words of nonsense about "the first time I ate an egg," to run overtop of an omelet recipe you're hoping to make the number one Google result? ChatGPT has you covered. Want to generate fake complaints or fake positive reviews? The Stochastic Parrot will produce 'em all day long.

As I wrote for Locus: "None of this prose is good, none of it is really socially useful, but there’s demand for it. Ironically, the more bullshit there is, the more bullshit filters there are, and this requires still more bullshit to overcome it."

Meanwhile, AA still hasn't answered my letter, and to be honest, I'm so sick of bullshit I can't be bothered to sue them anymore. I suppose that's what they were counting on.

If you'd like an essay-formatted version of this post to read or share, here's a link to it on pluralistic.net, my surveillance-free, ad-free, tracker-free blog:

https://pluralistic.net/2023/09/07/govern-yourself-accordingly/#robolawyers

Image: Cryteria (modified) https://commons.wikimedia.org/wiki/File:HAL9000.svg

CC BY 3.0

https://creativecommons.org/licenses/by/3.0/deed.en

#pluralistic #chatbots #plausible sentence generators #robot lawyers #robolawyers #ai #ml #machine learning #artificial intelligence #stochastic parrots #bullshit #bullshit generators #the bullshit wars #llms #large language models #writing #Ben Levinstein

2K notes · View notes

mostlysignssomeportents · 2 years ago

Text

The AI hype bubble is the new crypto hype bubble

Back in 2017 Long Island Ice Tea — known for its undistinguished, barely drinkable sugar-water — changed its name to “Long Blockchain Corp.” Its shares surged to a peak of 400% over their pre-announcement price. The company announced no specific integrations with any kind of blockchain, nor has it made any such integrations since.

If you’d like an essay-formatted version of this post to read or share, here’s a link to it on pluralistic.net, my surveillance-free, ad-free, tracker-free blog:

https://pluralistic.net/2023/03/09/autocomplete-worshippers/#the-real-ai-was-the-corporations-that-we-fought-along-the-way

LBCC was subsequently delisted from NASDAQ after settling with the SEC over fraudulent investor statements. Today, the company trades over the counter and its market cap is $36m, down from $138m.

https://cointelegraph.com/news/textbook-case-of-crypto-hype-how-iced-tea-company-went-blockchain-and-failed-despite-a-289-percent-stock-rise

The most remarkable thing about this incredibly stupid story is that LBCC wasn’t the peak of the blockchain bubble — rather, it was the start of blockchain’s final pump-and-dump. By the standards of 2022’s blockchain grifters, LBCC was small potatoes, a mere $138m sugar-water grift.

They didn’t have any NFTs, no wash trades, no ICO. They didn’t have a Superbowl ad. They didn’t steal billions from mom-and-pop investors while proclaiming themselves to be “Effective Altruists.” They didn’t channel hundreds of millions to election campaigns through straw donations and other forms of campaing finance frauds. They didn’t even open a crypto-themed hamburger restaurant where you couldn’t buy hamburgers with crypto:

https://robbreport.com/food-drink/dining/bored-hungry-restaurant-no-cryptocurrency-1234694556/

They were amateurs. Their attempt to “make fetch happen” only succeeded for a brief instant. By contrast, the superpredators of the crypto bubble were able to make fetch happen over an improbably long timescale, deploying the most powerful reality distortion fields since Pets.com.

Anything that can’t go on forever will eventually stop. We’re told that trillions of dollars’ worth of crypto has been wiped out over the past year, but these losses are nowhere to be seen in the real economy — because the “wealth” that was wiped out by the crypto bubble’s bursting never existed in the first place.

Like any Ponzi scheme, crypto was a way to separate normies from their savings through the pretense that they were “investing” in a vast enterprise — but the only real money (“fiat” in cryptospeak) in the system was the hardscrabble retirement savings of working people, which the bubble’s energetic inflaters swapped for illiquid, worthless shitcoins.

We’ve stopped believing in the illusory billions. Sam Bankman-Fried is under house arrest. But the people who gave him money — and the nimbler Ponzi artists who evaded arrest — are looking for new scams to separate the marks from their money.

Take Morganstanley, who spent 2021 and 2022 hyping cryptocurrency as a massive growth opportunity:

https://cointelegraph.com/news/morgan-stanley-launches-cryptocurrency-research-team

Today, Morganstanley wants you to know that AI is a $6 trillion opportunity.

They’re not alone. The CEOs of Endeavor, Buzzfeed, Microsoft, Spotify, Youtube, Snap, Sports Illustrated, and CAA are all out there, pumping up the AI bubble with every hour that god sends, declaring that the future is AI.

https://www.hollywoodreporter.com/business/business-news/wall-street-ai-stock-price-1235343279/

Google and Bing are locked in an arms-race to see whose search engine can attain the speediest, most profound enshittification via chatbot, replacing links to web-pages with florid paragraphs composed by fully automated, supremely confident liars:

https://pluralistic.net/2023/02/16/tweedledumber/#easily-spooked

Blockchain was a solution in search of a problem. So is AI. Yes, Buzzfeed will be able to reduce its wage-bill by automating its personality quiz vertical, and Spotify’s “AI DJ” will produce slightly less terrible playlists (at least, to the extent that Spotify doesn’t put its thumb on the scales by inserting tracks into the playlists whose only fitness factor is that someone paid to boost them).

But even if you add all of this up, double it, square it, and add a billion dollar confidence interval, it still doesn’t add up to what Bank Of America analysts called “a defining moment — like the internet in the ’90s.” For one thing, the most exciting part of the “internet in the ‘90s” was that it had incredibly low barriers to entry and wasn’t dominated by large companies — indeed, it had them running scared.

The AI bubble, by contrast, is being inflated by massive incumbents, whose excitement boils down to “This will let the biggest companies get much, much bigger and the rest of you can go fuck yourselves.” Some revolution.

AI has all the hallmarks of a classic pump-and-dump, starting with terminology. AI isn’t “artificial” and it’s not “intelligent.” “Machine learning” doesn’t learn. On this week’s Trashfuture podcast, they made an excellent (and profane and hilarious) case that ChatGPT is best understood as a sophisticated form of autocomplete — not our new robot overlord.

https://open.spotify.com/episode/4NHKMZZNKi0w9mOhPYIL4T

We all know that autocomplete is a decidedly mixed blessing. Like all statistical inference tools, autocomplete is profoundly conservative — it wants you to do the same thing tomorrow as you did yesterday (that’s why “sophisticated” ad retargeting ads show you ads for shoes in response to your search for shoes). If the word you type after “hey” is usually “hon” then the next time you type “hey,” autocomplete will be ready to fill in your typical following word — even if this time you want to type “hey stop texting me you freak”:

https://blog.lareviewofbooks.org/provocations/neophobic-conservative-ai-overlords-want-everything-stay/

And when autocomplete encounters a new input — when you try to type something you’ve never typed before — it tries to get you to finish your sentence with the statistically median thing that everyone would type next, on average. Usually that produces something utterly bland, but sometimes the results can be hilarious. Back in 2018, I started to text our babysitter with “hey are you free to sit” only to have Android finish the sentence with “on my face” (not something I’d ever typed!):

https://mashable.com/article/android-predictive-text-sit-on-my-face

Modern autocomplete can produce long passages of text in response to prompts, but it is every bit as unreliable as 2018 Android SMS autocomplete, as Alexander Hanff discovered when ChatGPT informed him that he was dead, even generating a plausible URL for a link to a nonexistent obit in The Guardian:

https://www.theregister.com/2023/03/02/chatgpt_considered_harmful/

Of course, the carnival barkers of the AI pump-and-dump insist that this is all a feature, not a bug. If autocomplete says stupid, wrong things with total confidence, that’s because “AI” is becoming more human, because humans also say stupid, wrong things with total confidence.

Exhibit A is the billionaire AI grifter Sam Altman, CEO if OpenAI — a company whose products are not open, nor are they artificial, nor are they intelligent. Altman celebrated the release of ChatGPT by tweeting “i am a stochastic parrot, and so r u.”

https://twitter.com/sama/status/1599471830255177728

This was a dig at the “stochastic parrots” paper, a comprehensive, measured roundup of criticisms of AI that led Google to fire Timnit Gebru, a respected AI researcher, for having the audacity to point out the Emperor’s New Clothes:

https://www.technologyreview.com/2020/12/04/1013294/google-ai-ethics-research-paper-forced-out-timnit-gebru/

Gebru’s co-author on the Parrots paper was Emily M Bender, a computational linguistics specialist at UW, who is one of the best-informed and most damning critics of AI hype. You can get a good sense of her position from Elizabeth Weil’s New York Magazine profile:

https://nymag.com/intelligencer/article/ai-artificial-intelligence-chatbots-emily-m-bender.html

Bender has made many important scholarly contributions to her field, but she is also famous for her rules of thumb, which caution her fellow scientists not to get high on their own supply:

Please do not conflate word form and meaning

Mind your own credulity

As Bender says, we’ve made “machines that can mindlessly generate text, but we haven’t learned how to stop imagining the mind behind it.” One potential tonic against this fallacy is to follow an Italian MP’s suggestion and replace “AI” with “SALAMI” (“Systematic Approaches to Learning Algorithms and Machine Inferences”). It’s a lot easier to keep a clear head when someone asks you, “Is this SALAMI intelligent? Can this SALAMI write a novel? Does this SALAMI deserve human rights?”

Bender’s most famous contribution is the “stochastic parrot,” a construct that “just probabilistically spits out words.” AI bros like Altman love the stochastic parrot, and are hellbent on reducing human beings to stochastic parrots, which will allow them to declare that their chatbots have feature-parity with human beings.

At the same time, Altman and Co are strangely afraid of their creations. It’s possible that this is just a shuck: “I have made something so powerful that it could destroy humanity! Luckily, I am a wise steward of this thing, so it’s fine. But boy, it sure is powerful!”

They’ve been playing this game for a long time. People like Elon Musk (an investor in OpenAI, who is hoping to convince the EU Commission and FTC that he can fire all of Twitter’s human moderators and replace them with chatbots without violating EU law or the FTC’s consent decree) keep warning us that AI will destroy us unless we tame it.

There’s a lot of credulous repetition of these claims, and not just by AI’s boosters. AI critics are also prone to engaging in what Lee Vinsel calls criti-hype: criticizing something by repeating its boosters’ claims without interrogating them to see if they’re true:

https://sts-news.medium.com/youre-doing-it-wrong-notes-on-criticism-and-technology-hype-18b08b4307e5

There are better ways to respond to Elon Musk warning us that AIs will emulsify the planet and use human beings for food than to shout, “Look at how irresponsible this wizard is being! He made a Frankenstein’s Monster that will kill us all!” Like, we could point out that of all the things Elon Musk is profoundly wrong about, he is most wrong about the philosophical meaning of Wachowksi movies:

https://www.theguardian.com/film/2020/may/18/lilly-wachowski-ivana-trump-elon-musk-twitter-red-pill-the-matrix-tweets

But even if we take the bros at their word when they proclaim themselves to be terrified of “existential risk” from AI, we can find better explanations by seeking out other phenomena that might be triggering their dread. As Charlie Stross points out, corporations are Slow AIs, autonomous artificial lifeforms that consistently do the wrong thing even when the people who nominally run them try to steer them in better directions:

https://media.ccc.de/v/34c3-9270-dude_you_broke_the_future

Imagine the existential horror of a ultra-rich manbaby who nominally leads a company, but can’t get it to follow: “everyone thinks I’m in charge, but I’m actually being driven by the Slow AI, serving as its sock puppet on some days, its golem on others.”

Ted Chiang nailed this back in 2017 (the same year of the Long Island Blockchain Company):

There’s a saying, popularized by Fredric Jameson, that it’s easier to imagine the end of the world than to imagine the end of capitalism. It’s no surprise that Silicon Valley capitalists don’t want to think about capitalism ending. What’s unexpected is that the way they envision the world ending is through a form of unchecked capitalism, disguised as a superintelligent AI. They have unconsciously created a devil in their own image, a boogeyman whose excesses are precisely their own.

https://www.buzzfeednews.com/article/tedchiang/the-real-danger-to-civilization-isnt-ai-its-runaway

Chiang is still writing some of the best critical work on “AI.” His February article in the New Yorker, “ChatGPT Is a Blurry JPEG of the Web,” was an instant classic:

[AI] hallucinations are compression artifacts, but — like the incorrect labels generated by the Xerox photocopier — they are plausible enough that identifying them requires comparing them against the originals, which in this case means either the Web or our own knowledge of the world.

https://www.newyorker.com/tech/annals-of-technology/chatgpt-is-a-blurry-jpeg-of-the-web

“AI” is practically purpose-built for inflating another hype-bubble, excelling as it does at producing party-tricks — plausible essays, weird images, voice impersonations. But as Princeton’s Matthew Salganik writes, there’s a world of difference between “cool” and “tool”:

https://freedom-to-tinker.com/2023/03/08/can-chatgpt-and-its-successors-go-from-cool-to-tool/

Nature can claim “conversational AI is a game-changer for science” but “there is a huge gap between writing funny instructions for removing food from home electronics and doing scientific research.” Salganik tried to get ChatGPT to help him with the most banal of scholarly tasks — aiding him in peer reviewing a colleague’s paper. The result? “ChatGPT didn’t help me do peer review at all; not one little bit.”

The criti-hype isn’t limited to ChatGPT, of course — there’s plenty of (justifiable) concern about image and voice generators and their impact on creative labor markets, but that concern is often expressed in ways that amplify the self-serving claims of the companies hoping to inflate the hype machine.

One of the best critical responses to the question of image- and voice-generators comes from Kirby Ferguson, whose final Everything Is a Remix video is a superb, visually stunning, brilliantly argued critique of these systems:

https://www.youtube.com/watch?v=rswxcDyotXA

One area where Ferguson shines is in thinking through the copyright question — is there any right to decide who can study the art you make? Except in some edge cases, these systems don’t store copies of the images they analyze, nor do they reproduce them:

https://pluralistic.net/2023/02/09/ai-monkeys-paw/#bullied-schoolkids

For creators, the important material question raised by these systems is economic, not creative: will our bosses use them to erode our wages? That is a very important question, and as far as our bosses are concerned, the answer is a resounding yes.

Markets value automation primarily because automation allows capitalists to pay workers less. The textile factory owners who purchased automatic looms weren’t interested in giving their workers raises and shorting working days. ‘ They wanted to fire their skilled workers and replace them with small children kidnapped out of orphanages and indentured for a decade, starved and beaten and forced to work, even after they were mangled by the machines. Fun fact: Oliver Twist was based on the bestselling memoir of Robert Blincoe, a child who survived his decade of forced labor:

https://www.gutenberg.org/files/59127/59127-h/59127-h.htm

Today, voice actors sitting down to record for games companies are forced to begin each session with “My name is ______ and I hereby grant irrevocable permission to train an AI with my voice and use it any way you see fit.”

https://www.vice.com/en/article/5d37za/voice-actors-sign-away-rights-to-artificial-intelligence

Let’s be clear here: there is — at present — no firmly established copyright over voiceprints. The “right” that voice actors are signing away as a non-negotiable condition of doing their jobs for giant, powerful monopolists doesn’t even exist. When a corporation makes a worker surrender this right, they are betting that this right will be created later in the name of “artists’ rights” — and that they will then be able to harvest this right and use it to fire the artists who fought so hard for it.

There are other approaches to this. We could support the US Copyright Office’s position that machine-generated works are not works of human creative authorship and are thus not eligible for copyright — so if corporations wanted to control their products, they’d have to hire humans to make them:

https://www.theverge.com/2022/2/21/22944335/us-copyright-office-reject-ai-generated-art-recent-entrance-to-paradise

Or we could create collective rights that belong to all artists and can’t be signed away to a corporation. That’s how the right to record other musicians’ songs work — and it’s why Taylor Swift was able to re-record the masters that were sold out from under her by evil private-equity bros::

https://doctorow.medium.com/united-we-stand-61e16ec707e2

Whatever we do as creative workers and as humans entitled to a decent life, we can’t afford drink the Blockchain Iced Tea. That means that we have to be technically competent, to understand how the stochastic parrot works, and to make sure our criticism doesn’t just repeat the marketing copy of the latest pump-and-dump.

Today (Mar 9), you can catch me in person in Austin at the UT School of Design and Creative Technologies, and remotely at U Manitoba’s Ethics of Emerging Tech Lecture.

Tomorrow (Mar 10), Rebecca Giblin and I kick off the SXSW reading series.

Image: Cryteria (modified) https://commons.wikimedia.org/wiki/File:HAL9000.svg

CC BY 3.0 https://creativecommons.org/licenses/by/3.0/deed.en

[Image ID: A graph depicting the Gartner hype cycle. A pair of HAL 9000's glowing red eyes are chasing each other down the slope from the Peak of Inflated Expectations to join another one that is at rest in the Trough of Disillusionment. It, in turn, sits atop a vast cairn of HAL 9000 eyes that are piled in a rough pyramid that extends below the graph to a distance of several times its height.]

2K notes · View notes