I mean really if we did elect an A.I. to office could it be any worse than what we have now? I'd be in favor of David Schweikert using A.I. at least there would be some intelligence.
Perplexity is, roughly speaking, the average number of words that can follow each word. This would tend to be much higher in, for instance, Proust, while being much lower in a book for children. Surprising words are good for getting attention, but repetition in political messaging is better for retention. So lower perplexity writing tends to be better for getting your message out.
Thanks so much for exploring this topic. I ran into my first AI generated product description the other day. I was shopping for new windows for my guest house. I went back to the vendor where I had purchased windows many years ago. I was intrigued by a glowing description of single hung windows with energy efficient “white glass”. Recently, I had seen cool photos in the NYT of a Paris loft apartment with a bank of white glass windows for light and privacy. Way cool. The bank of windows in my little guest house could look like a Parisian apartment! Such an upgrade from the current shabby chic look! The description was wordy and elaborate in describing how stylish the small single hung windows were. I called the local store with questions because they had some of the windows in stock but not all of the sizes I needed. I asked the quintessential somewhat sarcastic store clerk about the “white glass” windows. She said, “Oh … the Internet is weird. You better come in and look at them first.” There was no white glass like the Paris apartment. “White” glass was plain old “frosted” bathroom window glass. I bought one energy efficient frosted “white glass” bathroom window and six regular energy efficient windows. AI!
A.I. or staffer, does it matter? Candidates have to attend events, meet their potential constituents and fundraise (for hours and hours and hours). Surely, voters knew that many (or most) candidates and elected officials have staff that write their content. I can’t count the number of published OpEds I have written for others and no one will know who/want/when/where. You don’t write and tell unless you write for the big leagues (state of the union, state of the state). Policy platforms, priorities, website content, the simple invite to the coffee with the candidate…that is what strategists and staffers were paid to do. The candidate makes the decision where they want the ship to sail, but it’s the paid staff and volunteers that that are performing the below deck grunt work. The candidate needs to spend a majority of their time interacting with voters- not stuck at their laptop prevaricating on word choice. And, political speech often sounds stilted. Trust me, there aren’t that many creative ways to write about property values and taxation without repetition. Hence, the tortured phrases that sound unnatural: home investment and excise, real estate and tariff, housing toll, etc. (In my defense, I’ve never written “housing toll” in place of “property taxes” - that’s just the example that came to the top of my head while using thumbs to comment on my tiny phone.) If one has the opportunity to be brief and concise- no problem. But, when you have to explain in more detail, it’s difficult to find “natural feeling” synonyms so you aren’t left using “women’s reproductive health” or “public education” or “border policy” 50 times in one piece.
You know who writes almost all their own content and do so better than anyone else? Teachers that run for public office.
Is it me? This feels like a hit piece. The subject of the use of AI, deep fakes, and ChatGPT in politics is certainly worthy of investigative reporting. But, this article appears to be pushing an agenda (pun intended).
Your piece says, essentially, O’Callaghan’s staff used ChatGPT…but, actually, the software used to discern “real” content versus ChatGPT is extremely flawed so we can’t actually say that with any certainty.
Meanwhile, candidates and elected officials routinely outright lie and obfuscate data to bolster their positions. Candidates and elected officials engage in suspect campaign finance reporting and use their elected positions and our tax dollars for campaign purposes. Pick any day of the week and a reporter should be asking follow up questions from Tom Horne on everything that comes out of the Dept of Ed’s offices. Where exactly does Wendy Rogers live? Justine Wadsack? But, here we are, engaging in a deep dive on whether content on a candidate’s website was written by ChatGPT, a staffer, or the candidate? Is the content a lie? Does it not reflect the candidates proven positions on various policy? No? Then, how is this worthy of a pointed article?
As a voter, I don’t care who wrote or in what format I can learn about a candidate’s priorities or values. The important question is whether those priorities are authentic and truly reflect the candidate’s values so that I know whether they align with my priorities. That’s what is important. Whether a candidate for CD1 uses ChatGPT to illustrate and articulate their position on women’s rights is irrelevant.
What is relevant is Woods running her campaign largely on a “pro choice” platform when she still has “pro life” tweets in her Twitter timeline.
That is a story worthy of reporting, in my opinion. Don’t you think, Hank?
Nah, I think this was thoughtful and nuanced. If i wanted to write a hit piece, it could have been a lot harsher. And i think the reason we're even talking about where rogers or wadsack live is because people have already covered that. Same for Woods pro-live/choice background. I'm all for beating dead horses, but how many times do you want to read stories about things you already know? This is a brave new front of campaigning, and that's where we like to be.
Most voters don't take the time to read through online policy proposals, so maybe the bigger question is why campaigns are even including them, especially if they can't afford to hire staffers or consultants who know how to write them. If I was reading the policy proposal that was highlighted in this article (about pay equity), I'd be more turned off than inspired by this candidate because it's basically a bunch of prose that says nothing, which is the problem with chatGPT. I'd rather see typos or grammatical errors than take the time to read a long-winded proposal that tells me exactly zero about what a candidate proposes or how this position is unique to this candidate. This reminds me of the not-so-distant-past when every candidate used the same D.C. firms to compose fundraising emails with the same, cookie-cutter, armageddon language. It may have worked in the short-term, but in the long-run, it only served to annoy voters and did nothing to create an effective brand for the candidate.
I mean really if we did elect an A.I. to office could it be any worse than what we have now? I'd be in favor of David Schweikert using A.I. at least there would be some intelligence.
Perplexity is, roughly speaking, the average number of words that can follow each word. This would tend to be much higher in, for instance, Proust, while being much lower in a book for children. Surprising words are good for getting attention, but repetition in political messaging is better for retention. So lower perplexity writing tends to be better for getting your message out.
Thanks so much for exploring this topic. I ran into my first AI generated product description the other day. I was shopping for new windows for my guest house. I went back to the vendor where I had purchased windows many years ago. I was intrigued by a glowing description of single hung windows with energy efficient “white glass”. Recently, I had seen cool photos in the NYT of a Paris loft apartment with a bank of white glass windows for light and privacy. Way cool. The bank of windows in my little guest house could look like a Parisian apartment! Such an upgrade from the current shabby chic look! The description was wordy and elaborate in describing how stylish the small single hung windows were. I called the local store with questions because they had some of the windows in stock but not all of the sizes I needed. I asked the quintessential somewhat sarcastic store clerk about the “white glass” windows. She said, “Oh … the Internet is weird. You better come in and look at them first.” There was no white glass like the Paris apartment. “White” glass was plain old “frosted” bathroom window glass. I bought one energy efficient frosted “white glass” bathroom window and six regular energy efficient windows. AI!
A.I. or staffer, does it matter? Candidates have to attend events, meet their potential constituents and fundraise (for hours and hours and hours). Surely, voters knew that many (or most) candidates and elected officials have staff that write their content. I can’t count the number of published OpEds I have written for others and no one will know who/want/when/where. You don’t write and tell unless you write for the big leagues (state of the union, state of the state). Policy platforms, priorities, website content, the simple invite to the coffee with the candidate…that is what strategists and staffers were paid to do. The candidate makes the decision where they want the ship to sail, but it’s the paid staff and volunteers that that are performing the below deck grunt work. The candidate needs to spend a majority of their time interacting with voters- not stuck at their laptop prevaricating on word choice. And, political speech often sounds stilted. Trust me, there aren’t that many creative ways to write about property values and taxation without repetition. Hence, the tortured phrases that sound unnatural: home investment and excise, real estate and tariff, housing toll, etc. (In my defense, I’ve never written “housing toll” in place of “property taxes” - that’s just the example that came to the top of my head while using thumbs to comment on my tiny phone.) If one has the opportunity to be brief and concise- no problem. But, when you have to explain in more detail, it’s difficult to find “natural feeling” synonyms so you aren’t left using “women’s reproductive health” or “public education” or “border policy” 50 times in one piece.
You know who writes almost all their own content and do so better than anyone else? Teachers that run for public office.
Welcome to the Brave New World of AI.
Is it me? This feels like a hit piece. The subject of the use of AI, deep fakes, and ChatGPT in politics is certainly worthy of investigative reporting. But, this article appears to be pushing an agenda (pun intended).
Your piece says, essentially, O’Callaghan’s staff used ChatGPT…but, actually, the software used to discern “real” content versus ChatGPT is extremely flawed so we can’t actually say that with any certainty.
Meanwhile, candidates and elected officials routinely outright lie and obfuscate data to bolster their positions. Candidates and elected officials engage in suspect campaign finance reporting and use their elected positions and our tax dollars for campaign purposes. Pick any day of the week and a reporter should be asking follow up questions from Tom Horne on everything that comes out of the Dept of Ed’s offices. Where exactly does Wendy Rogers live? Justine Wadsack? But, here we are, engaging in a deep dive on whether content on a candidate’s website was written by ChatGPT, a staffer, or the candidate? Is the content a lie? Does it not reflect the candidates proven positions on various policy? No? Then, how is this worthy of a pointed article?
As a voter, I don’t care who wrote or in what format I can learn about a candidate’s priorities or values. The important question is whether those priorities are authentic and truly reflect the candidate’s values so that I know whether they align with my priorities. That’s what is important. Whether a candidate for CD1 uses ChatGPT to illustrate and articulate their position on women’s rights is irrelevant.
What is relevant is Woods running her campaign largely on a “pro choice” platform when she still has “pro life” tweets in her Twitter timeline.
That is a story worthy of reporting, in my opinion. Don’t you think, Hank?
Nah, I think this was thoughtful and nuanced. If i wanted to write a hit piece, it could have been a lot harsher. And i think the reason we're even talking about where rogers or wadsack live is because people have already covered that. Same for Woods pro-live/choice background. I'm all for beating dead horses, but how many times do you want to read stories about things you already know? This is a brave new front of campaigning, and that's where we like to be.
Fascinating series of articles about AI.
Most voters don't take the time to read through online policy proposals, so maybe the bigger question is why campaigns are even including them, especially if they can't afford to hire staffers or consultants who know how to write them. If I was reading the policy proposal that was highlighted in this article (about pay equity), I'd be more turned off than inspired by this candidate because it's basically a bunch of prose that says nothing, which is the problem with chatGPT. I'd rather see typos or grammatical errors than take the time to read a long-winded proposal that tells me exactly zero about what a candidate proposes or how this position is unique to this candidate. This reminds me of the not-so-distant-past when every candidate used the same D.C. firms to compose fundraising emails with the same, cookie-cutter, armageddon language. It may have worked in the short-term, but in the long-run, it only served to annoy voters and did nothing to create an effective brand for the candidate.