If you really want to know, try offering us three or four new, anonymous essays and let us identify the author. Sort of like a blind wine tasting, so to speak. In any event I would say that none of the articles seemed so far removed from your normal points of view, modes of analysis and direct phrasing that I might have questioned seeing them under your byline after a brisk read through. I have read a huge percentage of your output for lo these 25 years (Happy Anniversary) and tend to read quite fast as I am familiar with you patterns of thought and argument development. And, as you say, GROK is only going to get better. I'd be pushing hard on "Royalties". Cheers!
I like to think I could tell a robot from real (and it is plainly apparent with older AIs), but not so sure it is currently or will remain that way. I love the idea of a test where we don't know the answer until afterwards.
Yes, please test. Also, it would have been hilarious is you had told Grok to write a column about the three columns and getting replaced by a robot. I was thinking you might have done so...
That was fun. A test of version 1! I might subscribe to a later version. You can't comment on everything so if I want to ponder what the Reynolds' thoughts are on something not previously discussed, I could fire it up. Or subscribe to more than one such robot.
When I was reading them, I could see your take on those subjects and they sounded like your themes. I thought they lacked your spark. You, being you, opted for "panache." Just so.
The AI writing was based on your writing. But your writing, which the AI learned from, was based on much wider reading over decades of time, including a broad knowledge of history and law among other things, plus critical thought. Who knows what Grok (or Claude, or Chat GPT, or whatever) would have come up with if you had said, read broadly on the following topics material published in the last five decades, plus world and American history, economics, jurisprudence, etc., and offer your critical analysis.
No, these AI columns don’t sound like you at all. I was a bit bored reading them. They did not bring me any new information or insights (“I know this already. Tell me something I don’t know”).
No matter what your feelings on AI and robots, the fact is that they are getting better and better every day, while humans are not. If they aren't superhuman yet, it just means we have to wait a while. It's like judo, the trick is to work *with* them, not fight against them.
To be blunt if your writing and thought construction were offered to me as the AI example was I would no longer be a paid subscriber. While the pieces read well they don’t live up to your standards of cadence and syntax. They “sound” more like a political stump speech than a thoughtfully written piece by a Yale trained lawyer. AI can not yet replace you. Sadly you will still have to put in the time and effort.
I'd like to be replaced by a robot. It would not only do a better job, it would outlive me. When I was writing on a daily basis, I wrote literally millions of words (which I think would be enough to create a permanent AI version of me). But the AI "me" would be up against the same problem I was, namely that people who agree with me would agree with me and people who disagree with me would disagree with me. Hey wait! Maybe the AIs can have a real debate and see whether one "side" can "convince" the other! That might matter, or would it?
Yeah, I have to say I didn't get any sense of your voice in these proposals. I *am* kind of impressed that it was at least willing to pitch you some ideas that were right of center and "anti-elite"; I'd have kind of expected this to be filtered out by the RLHF. Maybe AI *is* getting less woke.
There were also some real giveaways that it was AI; I was amused that it managed to use the "It's not X, it's Y" trope in all three proposals:
#1: This isn’t governance. It’s capture.
#2: This is not weakness. It is pattern recognition.
#3: Groupthink is not a bug; it became a feature.
There were some longer instances of this trope besides these, whose length served to disguise them slightly. It's a strong writing trick to use, but not when it turns into P-K4.
I won't disparage its use of em-dashes because I love them myself, and I think you do too. :-)
My real concern is that, despite the vanilla characterlessness of the AI's prose, it might still give you second thoughts about writing your own columns about these issues. On one hand, it's true that a column about, say, how democratic socialism always goes is almost trite these days; there's no shortage of them and it might be just as well if the AI warns you off. But I'd hate for you to get a case of writer's block because nothing you think of seems novel enough.
Of course, you might get better results from a more frontier model. I don't find Grok even as smart as the free ChatGPT, and it's been a while since I ponied up for a paid AI.
It seems like it does a nice job at reading your previous columns and impersonating you. But I would imagine that the more things change and time goes by, without constant updates from the actual Glenn Reynolds (i.e., new opinion columns in which you react to new events and potentially change your mind), it will become less good at playing GHR.
If you really want to know, try offering us three or four new, anonymous essays and let us identify the author. Sort of like a blind wine tasting, so to speak. In any event I would say that none of the articles seemed so far removed from your normal points of view, modes of analysis and direct phrasing that I might have questioned seeing them under your byline after a brisk read through. I have read a huge percentage of your output for lo these 25 years (Happy Anniversary) and tend to read quite fast as I am familiar with you patterns of thought and argument development. And, as you say, GROK is only going to get better. I'd be pushing hard on "Royalties". Cheers!
I like to think I could tell a robot from real (and it is plainly apparent with older AIs), but not so sure it is currently or will remain that way. I love the idea of a test where we don't know the answer until afterwards.
Yes, please test. Also, it would have been hilarious is you had told Grok to write a column about the three columns and getting replaced by a robot. I was thinking you might have done so...
I would have known immediately just like how I know every time a stupid AI essay is on my feed on Facebook. It sounds nothing like Glenn.
That was fun. A test of version 1! I might subscribe to a later version. You can't comment on everything so if I want to ponder what the Reynolds' thoughts are on something not previously discussed, I could fire it up. Or subscribe to more than one such robot.
When I was reading them, I could see your take on those subjects and they sounded like your themes. I thought they lacked your spark. You, being you, opted for "panache." Just so.
The AI writing was based on your writing. But your writing, which the AI learned from, was based on much wider reading over decades of time, including a broad knowledge of history and law among other things, plus critical thought. Who knows what Grok (or Claude, or Chat GPT, or whatever) would have come up with if you had said, read broadly on the following topics material published in the last five decades, plus world and American history, economics, jurisprudence, etc., and offer your critical analysis.
AI is not capable of writing with style. Not Glenn's and not really anyone's. It always reads like AI.
No, these AI columns don’t sound like you at all. I was a bit bored reading them. They did not bring me any new information or insights (“I know this already. Tell me something I don’t know”).
Your job is safe, for now…
The themes are definitely you … the content is only about 75% you. And it’s consistently not you in ways that make it boring.
But it’s also better than I expected.
No matter what your feelings on AI and robots, the fact is that they are getting better and better every day, while humans are not. If they aren't superhuman yet, it just means we have to wait a while. It's like judo, the trick is to work *with* them, not fight against them.
We figured it was going to happen. After all, you are married to a cyborg.
To be blunt if your writing and thought construction were offered to me as the AI example was I would no longer be a paid subscriber. While the pieces read well they don’t live up to your standards of cadence and syntax. They “sound” more like a political stump speech than a thoughtfully written piece by a Yale trained lawyer. AI can not yet replace you. Sadly you will still have to put in the time and effort.
AI is recursive by default.
Its just as in past; "GIGO"
Or in your case NotGI; NotGo.
Its flat and lacking your voice, gracious wit and original insight.
That can be copied but not made on its own.
A little boring and maybe a bit creepy. I don't think I like InstaAIpundit.
I'd like to be replaced by a robot. It would not only do a better job, it would outlive me. When I was writing on a daily basis, I wrote literally millions of words (which I think would be enough to create a permanent AI version of me). But the AI "me" would be up against the same problem I was, namely that people who agree with me would agree with me and people who disagree with me would disagree with me. Hey wait! Maybe the AIs can have a real debate and see whether one "side" can "convince" the other! That might matter, or would it?
I agree with the NY Post editor.
they're not up to your usual snuff imo. you tend to include both more witty turns of phrase and clever entendres.
Yeah, I have to say I didn't get any sense of your voice in these proposals. I *am* kind of impressed that it was at least willing to pitch you some ideas that were right of center and "anti-elite"; I'd have kind of expected this to be filtered out by the RLHF. Maybe AI *is* getting less woke.
There were also some real giveaways that it was AI; I was amused that it managed to use the "It's not X, it's Y" trope in all three proposals:
#1: This isn’t governance. It’s capture.
#2: This is not weakness. It is pattern recognition.
#3: Groupthink is not a bug; it became a feature.
There were some longer instances of this trope besides these, whose length served to disguise them slightly. It's a strong writing trick to use, but not when it turns into P-K4.
I won't disparage its use of em-dashes because I love them myself, and I think you do too. :-)
My real concern is that, despite the vanilla characterlessness of the AI's prose, it might still give you second thoughts about writing your own columns about these issues. On one hand, it's true that a column about, say, how democratic socialism always goes is almost trite these days; there's no shortage of them and it might be just as well if the AI warns you off. But I'd hate for you to get a case of writer's block because nothing you think of seems novel enough.
Of course, you might get better results from a more frontier model. I don't find Grok even as smart as the free ChatGPT, and it's been a while since I ponied up for a paid AI.
I agree with your editor.
Steely Dan reference missing.
It seems like it does a nice job at reading your previous columns and impersonating you. But I would imagine that the more things change and time goes by, without constant updates from the actual Glenn Reynolds (i.e., new opinion columns in which you react to new events and potentially change your mind), it will become less good at playing GHR.