That’s what I was thinking about. In theory there could be a cool idea here. But using AI to do it would just give shitty results. But you could actually do a little research and find a few short articles or stories about things your kid is interested in, and then feed them into a decent TTS and edit it together like a podcast. Then where they would normally take a pause for an ad read have a little shout out wishing your kid a happy birthday or wishing them luck on their exams or competition or whatever. I think it could actually be pretty neat, and give you something to chat about while you listen, and having the little personalized messages would feel like magic to the kid, because it’s kind of like getting a shout out on Radio/Television. It can be exciting and make you feel bigger than just your small social bubble. It’s exciting to have a little 30 seconds of fame.
But obviously Altman is just grasping at straws to shill his product, and hasn’t put this much thought into it.
Ever stumble across one of those totally AI YouTube video, with the professional sounding voice that pronounced words oddly, or emphasizes phrases in a weird way? They go on and on about some subject without really saying anything, making empty pronouncements with florid language?
Like that. The podcast equivalent of flavorless gummy bears.
Right? It’s not like the kid’s soccer game is part of the AI’s programming, it won’t know anything about it that you don’t tell it. It can’t go over who they’re playing against, or give their record so far, or provide any other new information about it, so what exactly is it going to say?
Then the birthday will still be “coming up” every day until the day itself arrives, so what is there to say about that that would be interesting and new on a daily drive?
That just leaves “some news,” at which point you might as well turn on the radio, or put on a real podcast.
Doesn’t matter so long as Altman just got geographic time stamps of where you your spouse your children and what activities they like to participate in at all times.
Likely all anyone puts in the calendar is “Simon’s basketball game” so it will at best reference basketball and Simon playing it and have some fake podcaster say, “And don’t forget about that basketball game at 8pm!”, and at worst invent/hallucinate tons of details it could never know.
One thing that worries me is even the c level executives offer the worst ideas when it comes to AI and their own AI products. Like, AI can manage to do a lot of useful stuff, but also a lot of terrible use cases… This being one. It takes 1 minute to glance at your week’s schedule. An AI podcast invites the worst kind of hallucinations. Whenever you give it a tiny bit of information and expect more back than you gave it like upscaling, you invite the worst hallucinations and inaccuracies.
The fact that they’re still struggling to make it sound relevant means they don’t even believe in it as much as they’re trying to sell it. Bubble is bound to pop soon with this sort of weird behavior… Seems like a year ago they were talking about the machine god being here by now, and now they’re like “hey did you know this thing can make a fake podcast?” Yeah I fucking knew that, it was a stupid use case 2 years ago and it’s a stupid use case now.
It’s funny, my job has been pushing AI hard for a couple years now… So I’ve been using it on them. I had it write my year end review 2 years ago, because I absolutely hate that I have to do that rather than my manager, and it was all bullshit, but I got a good raise. Last year I tried to be honest, and ended up half assing it, but because I was doing well and everyone knew it (so arguably I shouldn’t need a year end review in the first place) I got a promotion and another big raise. Now I’m using it to write my goals, another thing that I hate having to do myself, but the problem is, by now I don’t want to put any effort into it, so it’s half assing it, and what it put together sounds great, but there’s no way to measure it, therefore no way to achieve the goals… By the time I put the effort into giving it the data it needs to make something useful, I could have just done it myself…
I dont get it. What would th epodcat actually talk about?
That’s what I was thinking about. In theory there could be a cool idea here. But using AI to do it would just give shitty results. But you could actually do a little research and find a few short articles or stories about things your kid is interested in, and then feed them into a decent TTS and edit it together like a podcast. Then where they would normally take a pause for an ad read have a little shout out wishing your kid a happy birthday or wishing them luck on their exams or competition or whatever. I think it could actually be pretty neat, and give you something to chat about while you listen, and having the little personalized messages would feel like magic to the kid, because it’s kind of like getting a shout out on Radio/Television. It can be exciting and make you feel bigger than just your small social bubble. It’s exciting to have a little 30 seconds of fame.
But obviously Altman is just grasping at straws to shill his product, and hasn’t put this much thought into it.
Ever stumble across one of those totally AI YouTube video, with the professional sounding voice that pronounced words oddly, or emphasizes phrases in a weird way? They go on and on about some subject without really saying anything, making empty pronouncements with florid language?
Like that. The podcast equivalent of flavorless gummy bears.
Right? It’s not like the kid’s soccer game is part of the AI’s programming, it won’t know anything about it that you don’t tell it. It can’t go over who they’re playing against, or give their record so far, or provide any other new information about it, so what exactly is it going to say?
Then the birthday will still be “coming up” every day until the day itself arrives, so what is there to say about that that would be interesting and new on a daily drive?
That just leaves “some news,” at which point you might as well turn on the radio, or put on a real podcast.
Doesn’t matter so long as Altman just got geographic time stamps of where you your spouse your children and what activities they like to participate in at all times.
Garbage in garbage out
Likely all anyone puts in the calendar is “Simon’s basketball game” so it will at best reference basketball and Simon playing it and have some fake podcaster say, “And don’t forget about that basketball game at 8pm!”, and at worst invent/hallucinate tons of details it could never know.
One thing that worries me is even the c level executives offer the worst ideas when it comes to AI and their own AI products. Like, AI can manage to do a lot of useful stuff, but also a lot of terrible use cases… This being one. It takes 1 minute to glance at your week’s schedule. An AI podcast invites the worst kind of hallucinations. Whenever you give it a tiny bit of information and expect more back than you gave it like upscaling, you invite the worst hallucinations and inaccuracies.
The fact that they’re still struggling to make it sound relevant means they don’t even believe in it as much as they’re trying to sell it. Bubble is bound to pop soon with this sort of weird behavior… Seems like a year ago they were talking about the machine god being here by now, and now they’re like “hey did you know this thing can make a fake podcast?” Yeah I fucking knew that, it was a stupid use case 2 years ago and it’s a stupid use case now.
It’s funny, my job has been pushing AI hard for a couple years now… So I’ve been using it on them. I had it write my year end review 2 years ago, because I absolutely hate that I have to do that rather than my manager, and it was all bullshit, but I got a good raise. Last year I tried to be honest, and ended up half assing it, but because I was doing well and everyone knew it (so arguably I shouldn’t need a year end review in the first place) I got a promotion and another big raise. Now I’m using it to write my goals, another thing that I hate having to do myself, but the problem is, by now I don’t want to put any effort into it, so it’s half assing it, and what it put together sounds great, but there’s no way to measure it, therefore no way to achieve the goals… By the time I put the effort into giving it the data it needs to make something useful, I could have just done it myself…