I know it's cliche, and probably done before, but in the spirit of this episode, I asked Chat GPT to write a letter responding to a letter threatening a defamation lawsuit, in the style of Ken White. The results I think are not too shabby; I bet with some more prompting we could get something even better:
Dear [Sender's Name],
I hope this letter finds you in a state of enlightenment, as I received your delightful missive dated [Date], brimming with the excitement of an impending defamation lawsuit. Your dedication to the fine art of legal threats is truly commendable, and I applaud your enthusiasm. However, I must take a moment to address some of the curious assertions and questionable legal theories that seem to animate your righteous indignation.
Let me begin by assuring you that I am no stranger to the laws surrounding defamation. Unlike certain other parties, I understand that truth is an absolute defense against such claims. Therefore, if my statements are, in fact, grounded in reality and uttered with the utmost sincerity, your defamation claims are as solid as a house of cards in a gentle breeze.
Now, while your impassioned diatribe hints at the presence of alleged defamatory statements, it disappoints me that you have failed to specify exactly what these statements entail. It seems that I am expected to play a game of "Defamation Detective" and piece together your accusations through vague insinuations and innuendos. Alas, my talents lie elsewhere, and I must implore you to provide some actual substance to support your claims. You know, like facts and evidence—the building blocks of any legitimate legal dispute.
In the spirit of fostering a cooperative and harmonious atmosphere, I suggest we explore alternatives to the tedious legal dance you propose. Have you considered taking a moment to reflect on the potential repercussions of your actions? Lawsuits tend to be expensive, drawn-out affairs that often leave both parties emotionally drained and financially depleted. Wouldn't it be more prudent to allocate those resources towards something more productive, like say, a therapy session to address any underlying insecurities?
Moreover, I must commend you on your tactical genius. Threatening a defamation lawsuit against someone who firmly believes in the virtues of free speech and public discourse is like attempting to douse a fire with gasoline. It tends to backfire spectacularly. I caution you to consider the potential damage to your own reputation should you choose to embark on this misguided crusade. You might find yourself cast as the unwitting villain in the play of public opinion.
In the spirit of generosity, I extend an olive branch. Let us engage in a civilized conversation, devoid of unnecessary legal posturing. I urge you to elucidate your grievances with specificity, that we may have an open dialogue and perhaps arrive at a mutually satisfactory resolution. After all, a constructive conversation has the power to enlighten and transform, whereas a lawsuit is often just a breeding ground for frustration and regret.
Thank you for gracing me with your legal threats. I eagerly await your response, with bated breath, as we embark on this thrilling journey through the labyrinthine realm of defamation law.
Ah yes, we haven't heard Ken discuss the pony menace in a while now. Maybe he's been silenced by Big Equine.
I did try asking ChatGPT to work "snort my taint" into it, but it couldn't figure out how to grammatically and contextually use it. I guess Popehat.com was not part of ChatGPT's training content.
You guys are killing it. It's refreshing that you have a good balance of experianced proffesional assessments, very thourough, and also some good sense of humor.
For your amusement my characterization of ChatGPT is that it’s simply a huge automated version of Family Feud where the question is “what’s the likeliest next word in this sentence”. When trained on fanfic you get the Kevin Roose events, when you train it on the internet in general and try to output legal briefs it might as well be fanfic. Now someday there will be a version whose only input is the Pacer library and might actually be a useful tool for lawyers, but in the meantime lawyers should stay far, far away from this as a work tool (other than writing bloviated marketing drivel)
Nah, nobody is underestimating it. Some people are, in fact, overestimating its potential to destroy humanity with its smarts and cunning.
Neither is likely to be right. We do this with every breathless announcement of discovery. We say electric cars will either save or wreck the economy. But, it turns out that something as unsexy as heat pumps have far more impact on our lives.
We worry about AI, while the slow creep of TikTok inserts itself into our lives in a way far more likely to be damaging.
"TV will kill radio!!!" Yet here we are listening to what is, in essence, radio.
"Computers will kill books!!!" And Amazon builds an empire on the foundation of selling books to people.
"Email will kill paper documents" Yet, I still have unending piles of bullshit paper on my desk.
Being able to draft a coherent legal brief to some minimally acceptable standard may be a hyper-specific task in one sense (after all, very few people do this for a living), but it requires a skill that is well beyond the capabilities of any AI we've seen so far: abstract reasoning.
To write a document that is anything other that a copy/paste of an existing document (with minor details like name/date filled in), one must apply an understanding of the law to the specific facts of a case. ChatGPT doesn't understand anything, it's just pumping out words one at a time using a sophisticated prediction algorithm. There are lots of entertaining examples of this out there, like its inability to count words or letters (e.g. if you ask "what is the fifth letter of this sentence?" it is likely to get it wrong), and its inability to recognize when a popular solution to a logic problem is invalidated by a small change of prompt (e.g. the Monty Hall problem but he shows you where the car is first).
Not that it couldn't ever produce a minimally acceptable brief - if the facts of one case are as such that parroting from a set of existing briefs using a one-word-at-a-time probabilistic guessing game algorithm will lead to something coherent, hooray. But you'd be nuts to depend on that from any AI currently in existence.
The problem is that LLM’s that mirror ChatGPT’s architecture preclude explainability. Other systems, primarily neurosymbolic ones like Writer’s, Elemental Cognition’s & Leela.ai’s, provide chain-of-reasoning explanations in addition to great generative capabilities. That’s what you need for legal reasoning
Ken and Josh, the folks over at "Opening Arguments" did a great breakdown on the chat GPT case and they kinda have a theory that they used Chat GPT to hide their initial fraud based on the timeline of events that I find persuasive.
Schwartz provided the export of his ChatGPT history today in response to the order to show cause. It backs up his story that he is stupid, but not initially malicious in his original motion for opposition (he doesn’t provide the same for the made up cases/case excerpts).
Additionally, I tried to recreate the idea that Schwartz isn’t a notary and I think they screwed that up too. There is a “license lookup” link on the New York notary website, but the link it takes you to only allows for limited types of licenses to search for, and notary public licenses aren’t one of the options. I think the host searched the “all licenses” selection and didn’t realize that “all licenses” didn’t include notary publics.
Is Torrez (of Opening Arguments) the guy who accused Ken of being a Nazi sympathiser and then got hit by MeToo allegations? Or is he just another guy who's been meTooed and I can't keep them straight? Ken was kinda oblique about all the history, but it seems like I remember a post he did about 5 months back wherein he talked about some podcaster who had made allegations against Ken being hit with "sex pest" charges.
Yeah I know there’s been allegations about him. Was curious if he was the same guy who leveled charges of being a nazi at Popehat. I have a hard time keeping these things straight.
I could go look it up but I’m too tired and lazy. Hopefully someone will chime in and either correct me or confirm.
I don’t want to cast aspersions here, but in fairness so many people have been accused of this it’s hard for an old Gen Xer to remember.
So there are a series of events where they first submitted a brief citing fake cases, get caught, submit them again with more panache, and then send the chat GPT results when it's clear they made up the cases. But ChatGPT users have pointed out you have to try really hard to get ChatGPT to say what the lawyers claim in said (they also doctored a notarized dose) and it reads like they are trying to cover their asses by blaming the whole thing on ChatGPT when they might have just fabricated the cases themselves.
Yes! I listened this morning. That was wild. Another interpretation could be that they doubled down, or one lawyer didn’t tell the other and kept prompt engineering ChatGPT to get what he wanted.
But the point that no reasonably competent lawyer, let alone one with 30y experience, would mistake what ChatCPT wrote for an 11th C decision is important. I hope one of the lawyers wasn’t having mental health issues that prevented him from perceiving that.
Thank goodness you've posted this week's episode! Now, I can hold out hope that some indictments might be forthcoming at any moment. (I presume that at least one of your subscribers is associated with the DoJ, in general, or the office of the Special Counsel, specifically, so they know when it's appropriate to act.)
I'm finding this ChatGPT case fascinating in that we are getting a look at human communication in a perfectly valid form, but entirely divorced from any value of epistemic truth. It's like saying 2+2=5, it's a valid equation, just not congruent with reality. It goes to how we define "truth" in the first place.
I love listening to you guys for both content (as a non-legal guy) and delivery.
Per delivery:
You two remind me of watching reruns of the "Burns and Allen Show" when I was a pre-teen in the 1960s. I didn't fully understand the humor being offered then until re-watching during grad school "Communications" in my twenties. Totally underappreciated humorous writing that depended on a literate audience.
I loved Gracie's sincere offering of her obtuse observations with either to husband George or their radio show announcer, Harry Von Zell. Harry had fantastic physical head-reaction shots to her lines.
George was a great writer and also presenter. But after years of being in Vaudeville, he knew what worked as an act. Therefore, he stepped aside from being the "funny guy" to let Gracie spread her comedic wings!
George Burns = Josh
Gracie Allen = Ken
As they would say back in the day (per seeing movies from those times) "WOWSZER"!!!
The mind-reading robot from Asimov's Susan Calvin stories? The best off-hand reference of the past week. Also, "spicy auto-correct".
There has been exactly one world-changing computing technology introduced into wide commercial use the past thirty years - the ability of any computer, anywhere, to talk to any other computer, as a peer. (Okay, arguably scale-on-demand compute services could be the second.)
ChatGPT (along with all the other so-called "artificial intelligence" technologies currently available) is the text-generation equivalent of Tesla auto-pilot - something that will not work in wide commercial deployment for decades (at least), that will be vastly over-hyped and breathlessly spoken about on cable news and (as Duncan Black refers to it) "that newspaper" regardless of its real-world record, and the basis of vast amounts of wealth for those ignorant enough or venal enough to capitalize on the extraordinary sums of money that will be thrown into bonfires in search of the next...Tesla auto-pilot.
I know it's cliche, and probably done before, but in the spirit of this episode, I asked Chat GPT to write a letter responding to a letter threatening a defamation lawsuit, in the style of Ken White. The results I think are not too shabby; I bet with some more prompting we could get something even better:
Dear [Sender's Name],
I hope this letter finds you in a state of enlightenment, as I received your delightful missive dated [Date], brimming with the excitement of an impending defamation lawsuit. Your dedication to the fine art of legal threats is truly commendable, and I applaud your enthusiasm. However, I must take a moment to address some of the curious assertions and questionable legal theories that seem to animate your righteous indignation.
Let me begin by assuring you that I am no stranger to the laws surrounding defamation. Unlike certain other parties, I understand that truth is an absolute defense against such claims. Therefore, if my statements are, in fact, grounded in reality and uttered with the utmost sincerity, your defamation claims are as solid as a house of cards in a gentle breeze.
Now, while your impassioned diatribe hints at the presence of alleged defamatory statements, it disappoints me that you have failed to specify exactly what these statements entail. It seems that I am expected to play a game of "Defamation Detective" and piece together your accusations through vague insinuations and innuendos. Alas, my talents lie elsewhere, and I must implore you to provide some actual substance to support your claims. You know, like facts and evidence—the building blocks of any legitimate legal dispute.
In the spirit of fostering a cooperative and harmonious atmosphere, I suggest we explore alternatives to the tedious legal dance you propose. Have you considered taking a moment to reflect on the potential repercussions of your actions? Lawsuits tend to be expensive, drawn-out affairs that often leave both parties emotionally drained and financially depleted. Wouldn't it be more prudent to allocate those resources towards something more productive, like say, a therapy session to address any underlying insecurities?
Moreover, I must commend you on your tactical genius. Threatening a defamation lawsuit against someone who firmly believes in the virtues of free speech and public discourse is like attempting to douse a fire with gasoline. It tends to backfire spectacularly. I caution you to consider the potential damage to your own reputation should you choose to embark on this misguided crusade. You might find yourself cast as the unwitting villain in the play of public opinion.
In the spirit of generosity, I extend an olive branch. Let us engage in a civilized conversation, devoid of unnecessary legal posturing. I urge you to elucidate your grievances with specificity, that we may have an open dialogue and perhaps arrive at a mutually satisfactory resolution. After all, a constructive conversation has the power to enlighten and transform, whereas a lawsuit is often just a breeding ground for frustration and regret.
Thank you for gracing me with your legal threats. I eagerly await your response, with bated breath, as we embark on this thrilling journey through the labyrinthine realm of defamation law.
Yours snarkily,
[Your Name]
"Thank you for gracing me with your legal threats."
Needs more pony.
Ah yes, we haven't heard Ken discuss the pony menace in a while now. Maybe he's been silenced by Big Equine.
I did try asking ChatGPT to work "snort my taint" into it, but it couldn't figure out how to grammatically and contextually use it. I guess Popehat.com was not part of ChatGPT's training content.
You guys are killing it. It's refreshing that you have a good balance of experianced proffesional assessments, very thourough, and also some good sense of humor.
Can’t believe you didn’t make the robot say RICO Hotline!
Open the courtroom doors, HAL.
For your amusement my characterization of ChatGPT is that it’s simply a huge automated version of Family Feud where the question is “what’s the likeliest next word in this sentence”. When trained on fanfic you get the Kevin Roose events, when you train it on the internet in general and try to output legal briefs it might as well be fanfic. Now someday there will be a version whose only input is the Pacer library and might actually be a useful tool for lawyers, but in the meantime lawyers should stay far, far away from this as a work tool (other than writing bloviated marketing drivel)
Nah, nobody is underestimating it. Some people are, in fact, overestimating its potential to destroy humanity with its smarts and cunning.
Neither is likely to be right. We do this with every breathless announcement of discovery. We say electric cars will either save or wreck the economy. But, it turns out that something as unsexy as heat pumps have far more impact on our lives.
We worry about AI, while the slow creep of TikTok inserts itself into our lives in a way far more likely to be damaging.
"TV will kill radio!!!" Yet here we are listening to what is, in essence, radio.
"Computers will kill books!!!" And Amazon builds an empire on the foundation of selling books to people.
"Email will kill paper documents" Yet, I still have unending piles of bullshit paper on my desk.
"Crypto!!!!" Need I say more on this?
And on and on.
Being able to draft a coherent legal brief to some minimally acceptable standard may be a hyper-specific task in one sense (after all, very few people do this for a living), but it requires a skill that is well beyond the capabilities of any AI we've seen so far: abstract reasoning.
To write a document that is anything other that a copy/paste of an existing document (with minor details like name/date filled in), one must apply an understanding of the law to the specific facts of a case. ChatGPT doesn't understand anything, it's just pumping out words one at a time using a sophisticated prediction algorithm. There are lots of entertaining examples of this out there, like its inability to count words or letters (e.g. if you ask "what is the fifth letter of this sentence?" it is likely to get it wrong), and its inability to recognize when a popular solution to a logic problem is invalidated by a small change of prompt (e.g. the Monty Hall problem but he shows you where the car is first).
Not that it couldn't ever produce a minimally acceptable brief - if the facts of one case are as such that parroting from a set of existing briefs using a one-word-at-a-time probabilistic guessing game algorithm will lead to something coherent, hooray. But you'd be nuts to depend on that from any AI currently in existence.
“Why you think that statement means I believe these AI's will replace anything is baffling.”
Well, it’s baffling to you because I don’t actually think that.
My point was about hubris. Specifically about the hubris of predicting an uncertain future. And maybe about humility.
The problem is that LLM’s that mirror ChatGPT’s architecture preclude explainability. Other systems, primarily neurosymbolic ones like Writer’s, Elemental Cognition’s & Leela.ai’s, provide chain-of-reasoning explanations in addition to great generative capabilities. That’s what you need for legal reasoning
Dear Ken and Josh:
Enjoyed the episode, love your work, please never use ChatGPT to write the accompanying text for an episode again k thx.
Love,
Rowyn
Ken and Josh, the folks over at "Opening Arguments" did a great breakdown on the chat GPT case and they kinda have a theory that they used Chat GPT to hide their initial fraud based on the timeline of events that I find persuasive.
Schwartz provided the export of his ChatGPT history today in response to the order to show cause. It backs up his story that he is stupid, but not initially malicious in his original motion for opposition (he doesn’t provide the same for the made up cases/case excerpts).
Additionally, I tried to recreate the idea that Schwartz isn’t a notary and I think they screwed that up too. There is a “license lookup” link on the New York notary website, but the link it takes you to only allows for limited types of licenses to search for, and notary public licenses aren’t one of the options. I think the host searched the “all licenses” selection and didn’t realize that “all licenses” didn’t include notary publics.
Is Torrez (of Opening Arguments) the guy who accused Ken of being a Nazi sympathiser and then got hit by MeToo allegations? Or is he just another guy who's been meTooed and I can't keep them straight? Ken was kinda oblique about all the history, but it seems like I remember a post he did about 5 months back wherein he talked about some podcaster who had made allegations against Ken being hit with "sex pest" charges.
Here’s the story I found: https://religionnews.com/2023/02/01/american-atheists-board-members-exit-dogged-by-misconduct-allegations/
Well that killed any interest I may have had in the Opening Arguments podcast....
All I know for sure is that Torrez sounds a lot like Paul F Tompkins.
Yeah I know there’s been allegations about him. Was curious if he was the same guy who leveled charges of being a nazi at Popehat. I have a hard time keeping these things straight.
I could go look it up but I’m too tired and lazy. Hopefully someone will chime in and either correct me or confirm.
I don’t want to cast aspersions here, but in fairness so many people have been accused of this it’s hard for an old Gen Xer to remember.
I haven’t listened yet, and haven’t heard this “initial fraud” thing? Is it easy to sum up, or should I just listen.
So there are a series of events where they first submitted a brief citing fake cases, get caught, submit them again with more panache, and then send the chat GPT results when it's clear they made up the cases. But ChatGPT users have pointed out you have to try really hard to get ChatGPT to say what the lawyers claim in said (they also doctored a notarized dose) and it reads like they are trying to cover their asses by blaming the whole thing on ChatGPT when they might have just fabricated the cases themselves.
Yes! I listened this morning. That was wild. Another interpretation could be that they doubled down, or one lawyer didn’t tell the other and kept prompt engineering ChatGPT to get what he wanted.
But the point that no reasonably competent lawyer, let alone one with 30y experience, would mistake what ChatCPT wrote for an 11th C decision is important. I hope one of the lawyers wasn’t having mental health issues that prevented him from perceiving that.
Just listen. It’s hilarious. I am about to start my 5th time through bc it’s so outrageous
Thank goodness you've posted this week's episode! Now, I can hold out hope that some indictments might be forthcoming at any moment. (I presume that at least one of your subscribers is associated with the DoJ, in general, or the office of the Special Counsel, specifically, so they know when it's appropriate to act.)
Word
Maybe better to have a "Chat RBG" to write legal briefs.
I gotta say, the lawyers indeed showed their briefs because they pantsed themselves with A.I.
I'm finding this ChatGPT case fascinating in that we are getting a look at human communication in a perfectly valid form, but entirely divorced from any value of epistemic truth. It's like saying 2+2=5, it's a valid equation, just not congruent with reality. It goes to how we define "truth" in the first place.
Funny enough, ChatGPT regularly gives wrong answers to mathematical questions of a similar difficulty level to "what is 2+2?"
I love listening to you guys for both content (as a non-legal guy) and delivery.
Per delivery:
You two remind me of watching reruns of the "Burns and Allen Show" when I was a pre-teen in the 1960s. I didn't fully understand the humor being offered then until re-watching during grad school "Communications" in my twenties. Totally underappreciated humorous writing that depended on a literate audience.
I loved Gracie's sincere offering of her obtuse observations with either to husband George or their radio show announcer, Harry Von Zell. Harry had fantastic physical head-reaction shots to her lines.
George was a great writer and also presenter. But after years of being in Vaudeville, he knew what worked as an act. Therefore, he stepped aside from being the "funny guy" to let Gracie spread her comedic wings!
George Burns = Josh
Gracie Allen = Ken
As they would say back in the day (per seeing movies from those times) "WOWSZER"!!!
The judge granted the extension request in handwriting. He just phrased it in a way to say that your client is in deep shit.
The legal jargon in legal works would be:
"Alta Cacas."
love listening to you guys, but 23 minutes is a pretty weak showing after a 2 week hiatus for a $6/month subscription.
I get the feeling they're expecting an indictment any time now, so they're pacing themselves so they can deliver a proper bonus episode.
I snortlaughed.
I see what you did there.
The mind-reading robot from Asimov's Susan Calvin stories? The best off-hand reference of the past week. Also, "spicy auto-correct".
There has been exactly one world-changing computing technology introduced into wide commercial use the past thirty years - the ability of any computer, anywhere, to talk to any other computer, as a peer. (Okay, arguably scale-on-demand compute services could be the second.)
ChatGPT (along with all the other so-called "artificial intelligence" technologies currently available) is the text-generation equivalent of Tesla auto-pilot - something that will not work in wide commercial deployment for decades (at least), that will be vastly over-hyped and breathlessly spoken about on cable news and (as Duncan Black refers to it) "that newspaper" regardless of its real-world record, and the basis of vast amounts of wealth for those ignorant enough or venal enough to capitalize on the extraordinary sums of money that will be thrown into bonfires in search of the next...Tesla auto-pilot.