My manager loves AI. I used it at work to make an educational podcast and it sucked. It made a lot of mistakes and I couldn’t get it to pronounce the name of my work correctly. I use it to help write emails at work or organize meeting notes. I feel bad about the environmental impact / drive a car still / do lots of things that are also probably bad for the environment.
I had a bite of someone else’s chocolate chop pancakes last Sunday and it was everything I needed.
1) Half full / half empty?! Depends on the day and / or my mood. Both are possible, I think. Some of the most talked about applications for it are not hard for me usually; I don’t feel challenged to write clearly most of the time. Your example of narration is an outlier, I think. The investment of time and energy required for the hope of a nuanced product makes the outcome questionable. Am I missing something?
2) Leave the yolk in! That’s where all the good stuff is! Mushrooms, sausage, and olives? Also my da favorite pizza! Pancakes okay too—even with chocolate chips though I wouldn’t go there. To each his / her / their own!
3) 👶🏻
4) Pass. I never found him that appealing. Maybe I sensed something? And it’s been said but he didn’t sound like BR really. Some strange admixture of sound waves….
AI because it’s already began and cannot be stopped. It’s like everything else. Some people will abuse it and use it for nefarious purposes. Why? Because people are like AI. They’re great and they’re awful.
Egg whites? Never. Chocolate chip pancakes and a side of thick cut applewood smoked bacon, please.
I was sort of anti-AI, until I got an AI summary of a client coaching session where I needed to eventually summarize that session for a class, and then I was a believer. It saved me about 30 minutes. When you have an energy limiting health condition, that’s actually a big deal.
I’ve never had chocolate chip pancakes. Should I take the AI lesson and apply it here? Crack open my mind a bit? I tend to find pancakes dry, but maybe the melted cc’s help out with that.
I really appreciate you bringing up the accessibility aspect here, Amy. Was just talking about this with a friend the other day. Most accessibility tools are clunky afterthoughts that tech companies roll out only after people stand up and say, "bro, not everyone on earth is a bro." AI tools weren't built for accessibility, but they do provide it in ways that feels a lot less clunky.
Now, for the pancakes. Honestly, if they're dry I don't think chocolate will help. It could be that you just don't like pancakes. That's ok, because there's more for the rest of us. But it could also be that your pancake source is no good. if that's the case, it's a damn tragedy. My advice: find the best pancake source in your area and give it a go. Hint: butter and syrup are your friends. If that doesn't do it, I'd say you're a french toast / waffle kind of griddle person
First thing I'd like to say is that thinking about the Apocalypse is _very_ future oriented. Paying your bills while everything is going to shit is believing that you'll survive the shit better if you're able to navigate the current shit. It's prosocial turtles the whole way down.
Which AI camp are you in? Hint: Both and neither are acceptable answers.
I'm pro workflow AI and anti generative AI. Which are which is the rub.
I appreciate the omelet choice for protein. Have you tried protein pancakes? They're not amazing, like real pancakes, but they're tasty and they (more or less) hit the spot, and they have protein.
Definitely in the cognitive dissonance stage. Can't be bothered to be pissed about stealing my words, since I've spent 15 years letting go the fact that pretty much all my work has been pirated many times.
But as ex-prof, with prof friends, I feel really bad about the potential damage and just nuisance factor of its use in academia.
Still contemplating using for audiobook narration since I haven't been able to justify the expense of getting last two books and last 3 novellas narrated, and I know that there are readers (listeners) who might appreciate.
And finally I have recently been inundated (at least one a day) with emails offering to market my books. And these are clearly AI generated because the are so good. Meaning, if a human had written they would have had to skim through all my books, done a deep dive into the themes, come up with things like key words. And if a human had done all that work, they might actually be someone worth checking out. But no human marketer would have bothered, the letters all from different people have large chunks that are identical, and 99% of the emails come from people who have no website or social media presence at all. Which is such an epic fail for a marketer. At first I found these amusing, now they are just a nuisance and I hate the idea of new authors getting sucked in and pay big sums for the promises of sales.
I feel for profs and college students right now. I read one story about how kids are using AI to write their papers. Note: "use" does a lot of work there, but the story implied that the students were prompting, copying, pasting, as opposed to working with the tool to augment their writing. Anyway, the story noted that profs had begun using AI to grade. Seems absurd to me. Have your AI grade my AI. I honestly, don't get that at all. But then again, I was always the student who signed up for a course *because* it had a writing requirement, so maybe I'm the weirdo.
1) I’m in the AI camp that uses Perplexity instead of Google because I stopped liking Google when I worked with Google on a project that we clearly did well, but they refused to pay us. Since then, I’ve found I could use a Google replacement fairly well and not give them my data to sell. Anyway, this generation of AI seems to work. It doesn’t seem like it’s going to take over the world before I leave this world.
B) Egg-white omelet? No, never. The egg is a good source of nutrients. Let’s not bastardize it just because someone tries to sell non-fat crap. I like French-toast or scrambled eggs more anyway.
3) I didn’t hear Burt Renolds. It seemed more like a mixture of about a dozen voices than Burt. Maybe it’s been too long since I watched me any Smokey.
5) You are never wrong. Neither am I when I disagree with you.
Perplexity feels like its come closer to the title of Google-killer than any other product I've seen. It probably accounts for about half the searches I used to do on Google. I just hope Perplexity doesn't screw it up with their ad business, but i suspect they will because ads and search are, ultimately, at odds with each other.
Thanks for your honesty on the subject! I’ve read a lot of hot takes, not enough nuance. I have mixed feelings, too.
I listened to your book with a built-in iPhone screen reader, which I prefer to your voice clone. I knew the robotic iPhone voice was going to be clunky, and it was. However, your voice clone sounded very close to your actual voice aesthetically, but it botches tone and jokes throughout because it doesn’t understand context beyond “these are negative words so say them angrily.” This has been my experience with generative AI. Big promises, big social costs, mild disappointment.
I wish the tech industry bet big on 3D printing instead, which seems like an amazing product that does pretty much exactly as advertised. Imagine Google and Meta and Microsoft competing to make 3D printed food and 3D printed circuit boards, instead of attempting to automate the arts and entertainment industry out of a job unsuccessfully.
On the 3D printing front, I don't think it's an either or choice. For me, AI and 3D printing are exciting technologies. What's concerning is that all of this tech is being developed by the biggest tech companies -- a break from historic development patterns. The concentration of power, when paired with the capabilities of both these technologies, is the frightening thing. With AI, Big Tech aims to control knowledge; with 3D printing the same folks would be aiming to control anything that's a physical product.
As for the voice, I should add a few more details. This voice was trained on podcast material. That's not ideal. For one thing, my mic still picked up some of Todd's voice, so it's not as clean as it should've been. The bigger issue, however, is that proper training (something I intend to do) would've required me reading specific passages to capture a full range of emotions. But the work doesn't stop there. After that process, you can (and should) edit the file for tone / emphasis / emotion / and pronunciation. You can do this word by word. It's tedious. I'd estimate that an AI version of my novel with my clone's voice would take me another few hours of recording, a couple more hours of fine-tuning the model, and then I'd basically have to take each word one at time to make it feel more, well, me. Basically, I'd be writing the novel again, but as an audio director. So I'd have more control than I'd likely have with an actor (time & budget wouldn't allow for notes on each word), but I'd likely end up doing A LOT more work. Go figure.
I agree that people with bad intentions will do bad things, no matter what the product is. I remember what searching on Google was like back when they promised not to be evil! IMHO, I see a lot more inherent utility with 3D printing. If all this time, energy and investment is gonna go somewhere in tech, I'd rather it be into an area – or yes, multiple areas – where the tech has utility. (That said, I'd rather the time, energy, and investment go to our public school system and public housing!) With AI, it feels like they put ChatGPT out there and asked us to figure out what it's good for, which is wild for a product that uses so much natural resources.
And if you had the time and energy to put in that amount of work, you could just, like, record an audio book! ^_^
But what does it really mean to sit it out? I get that you aren't using it for your work. Totally makes sense. But AI is a part of the movies and shows we watch. It plays a big role in de-aging for example. So you're seeing AI pop up all around you. Are you saying no to that stuff as a consumer? I don't mean, no to something that was 100% AI, but to something like a film where AI clearly played a role?
Good clarification. I meant I'm not seeking out AI for anything. Ignoring Google's AI search summary. Definitely not trying to optimize my life with AI. Etc
De-aging is an area where I sorta get it. But creating a movie or something entirely with AI? Nah.
The de-aging sorta excites me and sorta bums me out. Like, it's cool to see Han Solo looking young and vibrant in films that were shot 50 years later, but also I have a hard time suspending disbelief b/c I know Harrison Ford is old AF. We saw F1 and I liked Brad Pitt as an aging driver taking one more shot at glory, but I kinda wanted to see the same movie with an actor who was actually 40, as opposed to one who's been playing a 40-something for two decades.
I relate to this a lot. I am with the anti ai crowd for my own writing but teach business writing to undergrads and have had to learn about it figure it out as they are definitely already using it and will need to use it at work. But I still feel a general dread about where it’s all heading.
I would say some (although a smaller group) are trying to use it as a shortcut to do most of the work for them, a larger portion are just trying out using it to help in some ways, and then another smaller group don't want to use it much at all. I've also generally noticed my business major students seem to want to use it more than my English major students, which I guess makes sense.
1. Professionally, I have had to transfer my business journalism skills to a marketing context. Marketing enthusiastically embraces AI, so either I embrace AI too, or I begin working on a meaningful essay about John Henry lying dead before the steam-tunneler.
I imagine this will change, but right now, AI is great at delivering 80% of a finished product very quickly. If you’re a deeply mediocre person, you can’t distinguish between 80% and 100%. But if you can, you’ll know that AI works best when you put effort into how you use it. I am not working less because of AI. But the work I do is producing more. And, my success with using AI (ChatGPT in particular) very much draws on my editorial skillset.
Now, this is all on my professional side. Creatively, writing gives me great joy. I have no interest in outsourcing that to a machine. Personally, I find ChatGPT can be helpful in certain kinds of data-crunching, but it regularly hallucinates on me, so I use it with a very skeptical eye.
2. Egg-white omelet, or chocolate chip pancakes?
Shakespeare. If life is collapsing so badly as to drive that kind of choice, then I’m choosing Shakespeare. Read Station Eleven by Emily St. John Mandel (or watch the miniseries adaption of it) and you’ll understand. Deep down, people who choose the assless chaps method of surviving the apocalypse want that for themselves already, and that ain’t me.
3. 🧍🤖? Not yet. Not ever, really. Our brains process energy on chemical switches that can accommodate many different if/than computation cycles which interact in endless combination. Silicon is still only binary, and its speed still can’t match our depth, and probably never will. Binary computation is like painting with two colors. It can do wondrous things. But it’s still only two colors. It will only ever be two colors.
4. Burt Reynolds isn't a bad choice. Certainly better than JoJo Siwa.
I listened to both of those recordings, and again the 80% thing immediately came to mind. There were moments in each where the AI used a weird inflection or missed an inflection I felt should have been there. If producing audio versions of your work is your end goal, just read your stuff. You sound fine on your podcast already. It doesn’t matter if it doesn’t sound like the AI versions - it sounds like you, and that's what I show up for. Besides, it’s good for us to read our work aloud. Helps the editing process.
I experimented with NoteGPT, which takes text and generates an AI-narrated podcast around it. Very impressive until you hit that 80% wall, and when you do, you hit it hard. Listening to the AI podcast hosts try to laugh made me wonder when they were going to stand menacingly on my doorstep and ask me over and over for me to let them into my house.
5. I don’t think you’re wrong. This stuff is here. It’s tech developed in bad faith by tech bros who want to devalue human expression, but this ain’t the first time America has innovated by fucking people over. You and I both live on stolen land, after all. In the meantime, we gotta eat.
Bill, I'd be curious to see what you think of ElevenLabs. I got into a little more detail in my response to Geoffrey's question (above), but the long and the short of it is that what I shared was basically a raw output. You can (and should) go word by word, line by line adjusting the model's performance. To your point, that's not less work, but more. Probably a topic for another post, but doing an audiobook this way could be viewed as a time / cost saver that allows mediocre people to pump out mediocre material for mediocre audiences. I think that's primarily what we're seeing at this point. But the AI version can also increase the work that goes into making an audiobook, although the cost remains low and the labor shifts from the actor / editor / director back to the author. Why would I do this? Well, I like my voice fine when I'm talking naturally on a podcast. And I've always been one to read my work aloud while writing / editing. But when it comes to performing one of my stories, I've never been satisfied, no matter how may classes I take or practice I do.
I think solutions like ElevenLabs are quite interesting. The work it takes to get yourself dialed in on a single project strikes me as difficult ROI to compute relative to putting in the same hours on honing one's own delivery. Where I see this having more value is you then intend to scale your voice to lots of different things, in which case all that time dialing in the AI is really just a startup cost. What i hang up on is how the model still needs to be fine-tuned to understand your vocal nuance, especially when it comes to inflection. Their model seems to be charging early adopters to help develop the tech, and for me, I keep wondering, what if I just waited and got in on this tech when it's a lot more finely tuned so I don't have to spend so much effort getting it to sound like myself?
The other thing, again strictly for me, is audiobook narration. My use case would be for fiction, and the best narration is more like very subtle voice acting. That's nuance upon nuance, and I just don't see the tech getting there soon enough to justify paying for such a service. At least, not yet.
But the critical thing here is I struggle with simply imagining how to make use of Ai. Apart from any philosophical issues I might have with the tech, the reality is that as a user, I struggle to think of how to put the tech to my advantage. That's on me.
My manager loves AI. I used it at work to make an educational podcast and it sucked. It made a lot of mistakes and I couldn’t get it to pronounce the name of my work correctly. I use it to help write emails at work or organize meeting notes. I feel bad about the environmental impact / drive a car still / do lots of things that are also probably bad for the environment.
I had a bite of someone else’s chocolate chop pancakes last Sunday and it was everything I needed.
1) Half full / half empty?! Depends on the day and / or my mood. Both are possible, I think. Some of the most talked about applications for it are not hard for me usually; I don’t feel challenged to write clearly most of the time. Your example of narration is an outlier, I think. The investment of time and energy required for the hope of a nuanced product makes the outcome questionable. Am I missing something?
2) Leave the yolk in! That’s where all the good stuff is! Mushrooms, sausage, and olives? Also my da favorite pizza! Pancakes okay too—even with chocolate chips though I wouldn’t go there. To each his / her / their own!
3) 👶🏻
4) Pass. I never found him that appealing. Maybe I sensed something? And it’s been said but he didn’t sound like BR really. Some strange admixture of sound waves….
5) Never!
Wait, I can quit paying my bills and enjoy a full stack of chocolate chip pancakes? Awesome!
Yes! You are good to dine and dash. Enjoy!
This was uplifting. No terrible. Insightful. Stupid. Hepfu-- BRAIN EXPLODES!
i'm sorry / you're welcome
Could you have AI read your book with Burt Reynolds’s and Jackie Gleason’s voices like Smokey and the Bandit? Of course throw in some Sally Fields.
If those actors or their estates license their voices, yes.
AI because it’s already began and cannot be stopped. It’s like everything else. Some people will abuse it and use it for nefarious purposes. Why? Because people are like AI. They’re great and they’re awful.
Egg whites? Never. Chocolate chip pancakes and a side of thick cut applewood smoked bacon, please.
If I can’t have yolks, I gotta find pleasure somewhere!
You're making great breakfast choices
I was sort of anti-AI, until I got an AI summary of a client coaching session where I needed to eventually summarize that session for a class, and then I was a believer. It saved me about 30 minutes. When you have an energy limiting health condition, that’s actually a big deal.
I’ve never had chocolate chip pancakes. Should I take the AI lesson and apply it here? Crack open my mind a bit? I tend to find pancakes dry, but maybe the melted cc’s help out with that.
I really appreciate you bringing up the accessibility aspect here, Amy. Was just talking about this with a friend the other day. Most accessibility tools are clunky afterthoughts that tech companies roll out only after people stand up and say, "bro, not everyone on earth is a bro." AI tools weren't built for accessibility, but they do provide it in ways that feels a lot less clunky.
Now, for the pancakes. Honestly, if they're dry I don't think chocolate will help. It could be that you just don't like pancakes. That's ok, because there's more for the rest of us. But it could also be that your pancake source is no good. if that's the case, it's a damn tragedy. My advice: find the best pancake source in your area and give it a go. Hint: butter and syrup are your friends. If that doesn't do it, I'd say you're a french toast / waffle kind of griddle person
T
he AI voice is yours but is somehow lacking: it's slightly "regimented" seeming. It doesn't pause for breath and reflection, It doesn't"t live.
True. Although it's worth noting that this was an early output, so what you're hearing can get better. Can it feel alive? TBD.
I love AI and find it so helpful for so many things! Also I’m sorry but I don’t think Clone Michael sounds like you at all!
Ha! My clone sucks. I'm gonna fire him & replace him with me. Pretty sure I'll see a headline soon in Hacker news about how humans are taking AI jobs
First thing I'd like to say is that thinking about the Apocalypse is _very_ future oriented. Paying your bills while everything is going to shit is believing that you'll survive the shit better if you're able to navigate the current shit. It's prosocial turtles the whole way down.
Which AI camp are you in? Hint: Both and neither are acceptable answers.
I'm pro workflow AI and anti generative AI. Which are which is the rub.
Egg-white omelet, or chocolate chip pancakes?
Omelets. Baby needs his protein.
🧍🤖?
Wut?
Burt Reynolds?!
Meh. Yeah I said it.
Am I wrong, or am I wrong?
Right.
I appreciate the omelet choice for protein. Have you tried protein pancakes? They're not amazing, like real pancakes, but they're tasty and they (more or less) hit the spot, and they have protein.
Definitely in the cognitive dissonance stage. Can't be bothered to be pissed about stealing my words, since I've spent 15 years letting go the fact that pretty much all my work has been pirated many times.
But as ex-prof, with prof friends, I feel really bad about the potential damage and just nuisance factor of its use in academia.
Still contemplating using for audiobook narration since I haven't been able to justify the expense of getting last two books and last 3 novellas narrated, and I know that there are readers (listeners) who might appreciate.
And finally I have recently been inundated (at least one a day) with emails offering to market my books. And these are clearly AI generated because the are so good. Meaning, if a human had written they would have had to skim through all my books, done a deep dive into the themes, come up with things like key words. And if a human had done all that work, they might actually be someone worth checking out. But no human marketer would have bothered, the letters all from different people have large chunks that are identical, and 99% of the emails come from people who have no website or social media presence at all. Which is such an epic fail for a marketer. At first I found these amusing, now they are just a nuisance and I hate the idea of new authors getting sucked in and pay big sums for the promises of sales.
I feel for profs and college students right now. I read one story about how kids are using AI to write their papers. Note: "use" does a lot of work there, but the story implied that the students were prompting, copying, pasting, as opposed to working with the tool to augment their writing. Anyway, the story noted that profs had begun using AI to grade. Seems absurd to me. Have your AI grade my AI. I honestly, don't get that at all. But then again, I was always the student who signed up for a course *because* it had a writing requirement, so maybe I'm the weirdo.
1) I’m in the AI camp that uses Perplexity instead of Google because I stopped liking Google when I worked with Google on a project that we clearly did well, but they refused to pay us. Since then, I’ve found I could use a Google replacement fairly well and not give them my data to sell. Anyway, this generation of AI seems to work. It doesn’t seem like it’s going to take over the world before I leave this world.
B) Egg-white omelet? No, never. The egg is a good source of nutrients. Let’s not bastardize it just because someone tries to sell non-fat crap. I like French-toast or scrambled eggs more anyway.
3) I didn’t hear Burt Renolds. It seemed more like a mixture of about a dozen voices than Burt. Maybe it’s been too long since I watched me any Smokey.
5) You are never wrong. Neither am I when I disagree with you.
Perplexity feels like its come closer to the title of Google-killer than any other product I've seen. It probably accounts for about half the searches I used to do on Google. I just hope Perplexity doesn't screw it up with their ad business, but i suspect they will because ads and search are, ultimately, at odds with each other.
Thanks for your honesty on the subject! I’ve read a lot of hot takes, not enough nuance. I have mixed feelings, too.
I listened to your book with a built-in iPhone screen reader, which I prefer to your voice clone. I knew the robotic iPhone voice was going to be clunky, and it was. However, your voice clone sounded very close to your actual voice aesthetically, but it botches tone and jokes throughout because it doesn’t understand context beyond “these are negative words so say them angrily.” This has been my experience with generative AI. Big promises, big social costs, mild disappointment.
I wish the tech industry bet big on 3D printing instead, which seems like an amazing product that does pretty much exactly as advertised. Imagine Google and Meta and Microsoft competing to make 3D printed food and 3D printed circuit boards, instead of attempting to automate the arts and entertainment industry out of a job unsuccessfully.
On the 3D printing front, I don't think it's an either or choice. For me, AI and 3D printing are exciting technologies. What's concerning is that all of this tech is being developed by the biggest tech companies -- a break from historic development patterns. The concentration of power, when paired with the capabilities of both these technologies, is the frightening thing. With AI, Big Tech aims to control knowledge; with 3D printing the same folks would be aiming to control anything that's a physical product.
As for the voice, I should add a few more details. This voice was trained on podcast material. That's not ideal. For one thing, my mic still picked up some of Todd's voice, so it's not as clean as it should've been. The bigger issue, however, is that proper training (something I intend to do) would've required me reading specific passages to capture a full range of emotions. But the work doesn't stop there. After that process, you can (and should) edit the file for tone / emphasis / emotion / and pronunciation. You can do this word by word. It's tedious. I'd estimate that an AI version of my novel with my clone's voice would take me another few hours of recording, a couple more hours of fine-tuning the model, and then I'd basically have to take each word one at time to make it feel more, well, me. Basically, I'd be writing the novel again, but as an audio director. So I'd have more control than I'd likely have with an actor (time & budget wouldn't allow for notes on each word), but I'd likely end up doing A LOT more work. Go figure.
I agree that people with bad intentions will do bad things, no matter what the product is. I remember what searching on Google was like back when they promised not to be evil! IMHO, I see a lot more inherent utility with 3D printing. If all this time, energy and investment is gonna go somewhere in tech, I'd rather it be into an area – or yes, multiple areas – where the tech has utility. (That said, I'd rather the time, energy, and investment go to our public school system and public housing!) With AI, it feels like they put ChatGPT out there and asked us to figure out what it's good for, which is wild for a product that uses so much natural resources.
And if you had the time and energy to put in that amount of work, you could just, like, record an audio book! ^_^
I've been pro technology my entire life but I'm willingly sitting AI out. Zero interest.
But what does it really mean to sit it out? I get that you aren't using it for your work. Totally makes sense. But AI is a part of the movies and shows we watch. It plays a big role in de-aging for example. So you're seeing AI pop up all around you. Are you saying no to that stuff as a consumer? I don't mean, no to something that was 100% AI, but to something like a film where AI clearly played a role?
Good clarification. I meant I'm not seeking out AI for anything. Ignoring Google's AI search summary. Definitely not trying to optimize my life with AI. Etc
De-aging is an area where I sorta get it. But creating a movie or something entirely with AI? Nah.
The de-aging sorta excites me and sorta bums me out. Like, it's cool to see Han Solo looking young and vibrant in films that were shot 50 years later, but also I have a hard time suspending disbelief b/c I know Harrison Ford is old AF. We saw F1 and I liked Brad Pitt as an aging driver taking one more shot at glory, but I kinda wanted to see the same movie with an actor who was actually 40, as opposed to one who's been playing a 40-something for two decades.
Funny you mention Han... it seriously hurt my heart to see Harrison de-aged in the last Indiana Jones movie.
F1 is a good time but I agree about Pitt. He's too old to be playing these roles!
I relate to this a lot. I am with the anti ai crowd for my own writing but teach business writing to undergrads and have had to learn about it figure it out as they are definitely already using it and will need to use it at work. But I still feel a general dread about where it’s all heading.
Just curious, but are you finding that students are using AI to do it all for them, or using AI to help them?
I would say some (although a smaller group) are trying to use it as a shortcut to do most of the work for them, a larger portion are just trying out using it to help in some ways, and then another smaller group don't want to use it much at all. I've also generally noticed my business major students seem to want to use it more than my English major students, which I guess makes sense.
1. Professionally, I have had to transfer my business journalism skills to a marketing context. Marketing enthusiastically embraces AI, so either I embrace AI too, or I begin working on a meaningful essay about John Henry lying dead before the steam-tunneler.
I imagine this will change, but right now, AI is great at delivering 80% of a finished product very quickly. If you’re a deeply mediocre person, you can’t distinguish between 80% and 100%. But if you can, you’ll know that AI works best when you put effort into how you use it. I am not working less because of AI. But the work I do is producing more. And, my success with using AI (ChatGPT in particular) very much draws on my editorial skillset.
Now, this is all on my professional side. Creatively, writing gives me great joy. I have no interest in outsourcing that to a machine. Personally, I find ChatGPT can be helpful in certain kinds of data-crunching, but it regularly hallucinates on me, so I use it with a very skeptical eye.
2. Egg-white omelet, or chocolate chip pancakes?
Shakespeare. If life is collapsing so badly as to drive that kind of choice, then I’m choosing Shakespeare. Read Station Eleven by Emily St. John Mandel (or watch the miniseries adaption of it) and you’ll understand. Deep down, people who choose the assless chaps method of surviving the apocalypse want that for themselves already, and that ain’t me.
3. 🧍🤖? Not yet. Not ever, really. Our brains process energy on chemical switches that can accommodate many different if/than computation cycles which interact in endless combination. Silicon is still only binary, and its speed still can’t match our depth, and probably never will. Binary computation is like painting with two colors. It can do wondrous things. But it’s still only two colors. It will only ever be two colors.
4. Burt Reynolds isn't a bad choice. Certainly better than JoJo Siwa.
I listened to both of those recordings, and again the 80% thing immediately came to mind. There were moments in each where the AI used a weird inflection or missed an inflection I felt should have been there. If producing audio versions of your work is your end goal, just read your stuff. You sound fine on your podcast already. It doesn’t matter if it doesn’t sound like the AI versions - it sounds like you, and that's what I show up for. Besides, it’s good for us to read our work aloud. Helps the editing process.
I experimented with NoteGPT, which takes text and generates an AI-narrated podcast around it. Very impressive until you hit that 80% wall, and when you do, you hit it hard. Listening to the AI podcast hosts try to laugh made me wonder when they were going to stand menacingly on my doorstep and ask me over and over for me to let them into my house.
5. I don’t think you’re wrong. This stuff is here. It’s tech developed in bad faith by tech bros who want to devalue human expression, but this ain’t the first time America has innovated by fucking people over. You and I both live on stolen land, after all. In the meantime, we gotta eat.
Bill, I'd be curious to see what you think of ElevenLabs. I got into a little more detail in my response to Geoffrey's question (above), but the long and the short of it is that what I shared was basically a raw output. You can (and should) go word by word, line by line adjusting the model's performance. To your point, that's not less work, but more. Probably a topic for another post, but doing an audiobook this way could be viewed as a time / cost saver that allows mediocre people to pump out mediocre material for mediocre audiences. I think that's primarily what we're seeing at this point. But the AI version can also increase the work that goes into making an audiobook, although the cost remains low and the labor shifts from the actor / editor / director back to the author. Why would I do this? Well, I like my voice fine when I'm talking naturally on a podcast. And I've always been one to read my work aloud while writing / editing. But when it comes to performing one of my stories, I've never been satisfied, no matter how may classes I take or practice I do.
I think solutions like ElevenLabs are quite interesting. The work it takes to get yourself dialed in on a single project strikes me as difficult ROI to compute relative to putting in the same hours on honing one's own delivery. Where I see this having more value is you then intend to scale your voice to lots of different things, in which case all that time dialing in the AI is really just a startup cost. What i hang up on is how the model still needs to be fine-tuned to understand your vocal nuance, especially when it comes to inflection. Their model seems to be charging early adopters to help develop the tech, and for me, I keep wondering, what if I just waited and got in on this tech when it's a lot more finely tuned so I don't have to spend so much effort getting it to sound like myself?
The other thing, again strictly for me, is audiobook narration. My use case would be for fiction, and the best narration is more like very subtle voice acting. That's nuance upon nuance, and I just don't see the tech getting there soon enough to justify paying for such a service. At least, not yet.
But the critical thing here is I struggle with simply imagining how to make use of Ai. Apart from any philosophical issues I might have with the tech, the reality is that as a user, I struggle to think of how to put the tech to my advantage. That's on me.