{"video_id":"fp_cQB4PZ2zk3","title":"What are your thoughts on AI?","channel":"FP Exclusives","show":"FP Exclusives","published_at":"2023-09-01T22:53:00.017Z","duration_s":6217,"segments":[{"start_s":0.0,"end_s":3.7199999999999998,"text":"I am here with Dan. Hey, and what do you do for the company here?","speaker":null,"is_sponsor":0},{"start_s":4.24,"end_s":8.96,"text":"I'm the infrastructure and technical specialist these days. I pretty much just write documentation","speaker":null,"is_sponsor":0},{"start_s":9.36,"end_s":14.84,"text":"Sounds good. Sounds good. So I'm just asking everyone here at the company what the thoughts on AI is the current state of it","speaker":null,"is_sponsor":0},{"start_s":15.12,"end_s":19.76,"text":"So I'll ask you first. Sure thoughts on the currency of AI. It's pretty interesting","speaker":null,"is_sponsor":0},{"start_s":19.76,"end_s":26.76,"text":"there's a lot of cool stuff going on and I'm excited to see it kind of develop over the next I guess a year or a few months even","speaker":null,"is_sponsor":0},{"start_s":27.12,"end_s":31.24,"text":"So you talk kind of sound like you're for AI more than against AI","speaker":null,"is_sponsor":0},{"start_s":31.72,"end_s":35.32,"text":"Yeah, I would probably say my thoughts on it are a little bit mixed","speaker":null,"is_sponsor":0},{"start_s":35.32,"end_s":40.88,"text":"I think it's not maybe as terrifying as people think it is but yeah, certainly certainly for it","speaker":null,"is_sponsor":0},{"start_s":41.4,"end_s":46.84,"text":"Sounds good. Sounds good. So if you're for it, then what's one thing you're really against in terms of the state of AI at the moment?","speaker":null,"is_sponsor":0},{"start_s":46.84,"end_s":50.32,"text":"I think probably some of the protectionism. I mean","speaker":null,"is_sponsor":0},{"start_s":51.2,"end_s":53.92,"text":"Looking at some of the requirements for doing these trainings","speaker":null,"is_sponsor":0},{"start_s":54.4,"end_s":62.16,"text":"You're spending millions of dollars on like cloud infrastructure GPU time just to actually create a","speaker":null,"is_sponsor":0},{"start_s":62.72,"end_s":71.96,"text":"reasonable model and so People I think are willing to protect that rather than just open sourcing it and letting any other company who has access to these massive data centers","speaker":null,"is_sponsor":0},{"start_s":71.96,"end_s":82.0,"text":"Like get in there, right? Okay, okay So have you used any AI currently like that where it's like they have chat GBT the image random readers music","speaker":null,"is_sponsor":0},{"start_s":82.2,"end_s":87.4,"text":"Have you use any of it? Yeah, I've played with pretty much all of them and I actually run a couple at home","speaker":null,"is_sponsor":0},{"start_s":87.4,"end_s":94.36,"text":"So I run an LLM as well as stable diffusion for image generation and I run those locally","speaker":null,"is_sponsor":0},{"start_s":94.36,"end_s":102.34,"text":"I recently just bought a couple more 30 90s and you know upgraded some of them with VRAM so that I could try some of the larger models","speaker":null,"is_sponsor":0},{"start_s":102.72,"end_s":105.22,"text":"So they're like a chat GBT for","speaker":null,"is_sponsor":0},{"start_s":105.88,"end_s":109.18,"text":"But you can only talk to them for like I don't know","speaker":null,"is_sponsor":0},{"start_s":109.78,"end_s":117.26,"text":"Two to five minutes before the like token inference length gets too much and it takes like a minute to respond to you","speaker":null,"is_sponsor":0},{"start_s":117.54,"end_s":126.58,"text":"But initially they're they're really quite powerful And they're a little bit more entertaining because they're the uncensored varieties. So what do you mean by that?","speaker":null,"is_sponsor":0},{"start_s":126.62,"end_s":131.3,"text":"So, you know, you say something to chat GBT for and it goes as a large language model","speaker":null,"is_sponsor":0},{"start_s":131.3,"end_s":136.26,"text":"I can't respond to this the uncensored one the uncensored ones are kind of designed as","speaker":null,"is_sponsor":0},{"start_s":137.06,"end_s":142.98,"text":"training bed for censorship and so you can kind of get it to talk about whatever you want or","speaker":null,"is_sponsor":0},{"start_s":143.9,"end_s":146.42,"text":"You know, sometimes chat GBT for will","speaker":null,"is_sponsor":0},{"start_s":147.02,"end_s":150.58,"text":"Catch things that it thinks are offensive and it isn't","speaker":null,"is_sponsor":0},{"start_s":151.02,"end_s":153.78,"text":"I have some people have told me some stuff internally","speaker":null,"is_sponsor":0},{"start_s":154.54,"end_s":162.26,"text":"That they use chat GBT for to have like difficult conversations about difficult topics and practice talking to actual people","speaker":null,"is_sponsor":0},{"start_s":162.66,"end_s":168.38,"text":"about these difficult subjects and Initially GPT 3 was kind of okay to talk about these","speaker":null,"is_sponsor":0},{"start_s":169.18,"end_s":172.9,"text":"but GPT 4 just won't even broach these subjects that are kind of like","speaker":null,"is_sponsor":0},{"start_s":173.46,"end_s":178.66,"text":"You know, they're they're hard to talk about is it cuz like they're trying to market it. I guess I'm assuming or yeah","speaker":null,"is_sponsor":0},{"start_s":178.66,"end_s":182.82,"text":"I was probably a marketing thing. There's also a liability thing probably","speaker":null,"is_sponsor":0},{"start_s":183.86,"end_s":188.22,"text":"You know people like to post what they talk about and then of course you also have like","speaker":null,"is_sponsor":0},{"start_s":188.98,"end_s":196.98,"text":"Really negative sentiment and things like that and also you can ask it dangerous things and get it to tell you dangerous information","speaker":null,"is_sponsor":0},{"start_s":197.9,"end_s":207.06,"text":"and so it's really kind of Scary in that way and I think as these companies are large and have a lot of money invested in these it's a little bit","speaker":null,"is_sponsor":0},{"start_s":208.42,"end_s":214.58,"text":"Be a difficult for them to allow that kind of like unfiltered information because they are so","speaker":null,"is_sponsor":0},{"start_s":215.26,"end_s":221.34,"text":"Confident at all the time. They just will a hundred percent confidence in what they're saying","speaker":null,"is_sponsor":0},{"start_s":221.34,"end_s":225.86,"text":"Even if they don't know what they're saying and that I think can be really dangerous","speaker":null,"is_sponsor":0},{"start_s":225.86,"end_s":229.1,"text":"A lot of people don't understand that it's fake information a lot of the time","speaker":null,"is_sponsor":0},{"start_s":230.06,"end_s":234.18,"text":"Another cool thing about running them locally is that you get to like mess with their brain a little bit","speaker":null,"is_sponsor":0},{"start_s":234.18,"end_s":241.86,"text":"So you can adjust the level of inference and the level of like token following guys called config scale or sorry configuration scale and","speaker":null,"is_sponsor":0},{"start_s":242.54,"end_s":245.7,"text":"You can make it those just be like","speaker":null,"is_sponsor":0},{"start_s":246.66,"end_s":251.3,"text":"Kind of insane if you go too far you can get into something called hallucinations where","speaker":null,"is_sponsor":0},{"start_s":251.94,"end_s":256.78,"text":"The inference models like feedback into themselves and it just says like random garbage","speaker":null,"is_sponsor":0},{"start_s":258.14,"end_s":263.14,"text":"That's can be really quite funny and and interesting. Yeah, it's neat watching them tick like that","speaker":null,"is_sponsor":0},{"start_s":263.18,"end_s":271.1,"text":"Okay, so you use it on a person level, but do you think as a tech slash crave company? We should be using AI on any capacity","speaker":null,"is_sponsor":0},{"start_s":271.1,"end_s":278.06,"text":"I actually did some training for us. I did a fine tune of chat GPT for","speaker":null,"is_sponsor":0},{"start_s":279.34,"end_s":284.06,"text":"We were trying to get it to write our with a hell of a cult showcase scripts","speaker":null,"is_sponsor":0},{"start_s":284.3,"end_s":289.5,"text":"Which are pretty boring and basic videos where a brand will be like","speaker":null,"is_sponsor":0},{"start_s":290.02,"end_s":294.34,"text":"Write a sponsored segment about our product and here's a bunch of","speaker":null,"is_sponsor":0},{"start_s":294.7,"end_s":299.3,"text":"Specifications and then we kind of have to turn that into a script and it's not very creative and it's not very fun","speaker":null,"is_sponsor":0},{"start_s":300.3,"end_s":303.98,"text":"So I made a data set from about 50 of these scripts","speaker":null,"is_sponsor":0},{"start_s":305.14,"end_s":315.54,"text":"You know because you have to you have to set a Prompt and a response so the prompt was like here's all the data and specifications on the brand and the response is our script","speaker":null,"is_sponsor":0},{"start_s":315.74,"end_s":318.98,"text":"right and So I went this data set","speaker":null,"is_sponsor":0},{"start_s":318.98,"end_s":324.9,"text":"I did a quick fine tune and it wasn't performing very well because the minimum data that you need for one of these types of fine","speaker":null,"is_sponsor":0},{"start_s":324.9,"end_s":330.98,"text":"tunes is 500 Right, so 500 scripts our LTT's we have about 5,000","speaker":null,"is_sponsor":0},{"start_s":331.02,"end_s":334.9,"text":"so that would be a reasonable data set to feed into a fine tune, but","speaker":null,"is_sponsor":0},{"start_s":335.46,"end_s":340.3,"text":"you know LTT's are special and You know there they wouldn't really work","speaker":null,"is_sponsor":0},{"start_s":340.7,"end_s":342.98,"text":"But these showcase scripts are pretty","speaker":null,"is_sponsor":0},{"start_s":344.14,"end_s":348.14,"text":"Boiler play Unfortunately, I was told after I asked for the other","speaker":null,"is_sponsor":0},{"start_s":348.98,"end_s":353.26,"text":"450 scripts that we have only ever done 50 of them so that project","speaker":null,"is_sponsor":0},{"start_s":353.5,"end_s":357.86,"text":"Through It was a it was a waste of a couple days of data compilation","speaker":null,"is_sponsor":0},{"start_s":357.9,"end_s":364.18,"text":"But did you learn a lot like what to do what not to do kind of thing? Yeah. Yeah, I learned a lot about how fine tunes function and","speaker":null,"is_sponsor":0},{"start_s":365.1,"end_s":368.94,"text":"You know, there's there's also a lot of issues with model overfitting","speaker":null,"is_sponsor":0},{"start_s":368.98,"end_s":372.54,"text":"So you can you can train a model too much and then it can only do","speaker":null,"is_sponsor":0},{"start_s":373.1,"end_s":380.3,"text":"One thing and that's kind of a dangerous balance to and so you know if you feed in that 50 data set and you like","speaker":null,"is_sponsor":0},{"start_s":380.3,"end_s":386.5,"text":"You really try to make it Do that 50 data set then you'll just get garbage because it's over fit","speaker":null,"is_sponsor":0},{"start_s":386.5,"end_s":393.22,"text":"And if you try to go anywhere outside of that entire limited data set like if you train on an LLM on","speaker":null,"is_sponsor":0},{"start_s":394.26,"end_s":397.86,"text":"100,000 pictures of dogs and their only face on","speaker":null,"is_sponsor":0},{"start_s":398.5,"end_s":406.02,"text":"Then you ask it to do a side-on picture of a dog and it can't because it's over fit for like front-facing dogs","speaker":null,"is_sponsor":0},{"start_s":406.58,"end_s":412.06,"text":"Or if you ask it to make a cat it can't do that like we couldn't feed in 50 showcase scripts","speaker":null,"is_sponsor":0},{"start_s":412.3,"end_s":417.54,"text":"Or 5,000 showcase scripts and then have it make an LTT because it would just make showcase scripts","speaker":null,"is_sponsor":0},{"start_s":418.1,"end_s":422.38,"text":"Sorry, so we should Didn't use it for a company or","speaker":null,"is_sponsor":0},{"start_s":423.14,"end_s":427.38,"text":"Well, I mean this is kind of the problem and this is something that I've realized more","speaker":null,"is_sponsor":0},{"start_s":427.94,"end_s":434.82,"text":"playing with them at home and getting it to let it spend a lot of time with them that you wouldn't be able to do with like a","speaker":null,"is_sponsor":0},{"start_s":435.18,"end_s":442.28,"text":"Chat GPT-4 or one of these web things because I'm sending like a hundred thousand or two hundred thousand tokens through","speaker":null,"is_sponsor":0},{"start_s":442.46,"end_s":449.22,"text":"These platforms and that would cost me like a five hundred dollars a day to actually get to play with them","speaker":null,"is_sponsor":0},{"start_s":450.34,"end_s":455.66,"text":"But I think even more now. I'm more convinced that they are just tools","speaker":null,"is_sponsor":0},{"start_s":456.22,"end_s":460.26,"text":"I know some artists friends of mine have been concerned about them","speaker":null,"is_sponsor":0},{"start_s":460.98,"end_s":469.86,"text":"but what I've noticed is that Because these LLMs and the image diffusion models are trained on massive quantities of data","speaker":null,"is_sponsor":0},{"start_s":470.82,"end_s":476.66,"text":"the Massive quantities of data that they have are like the most generic pictures","speaker":null,"is_sponsor":0},{"start_s":477.18,"end_s":481.18,"text":"Right there the generic pictures. There are single sort of style of picture","speaker":null,"is_sponsor":0},{"start_s":481.94,"end_s":493.14,"text":"there are sort of a You know, it's it's one thing and so I think at least for artists what's gonna happen is maybe there'll be less of a demand for","speaker":null,"is_sponsor":0},{"start_s":494.02,"end_s":501.3,"text":"This basic art, you know, I want, you know a pin-up of a girl or I want a scene of a mountain that sort of thing","speaker":null,"is_sponsor":0},{"start_s":501.3,"end_s":508.82,"text":"You know, there's already a massive quantity of art out there that has those and you can use a diffusion model to create that kind of","speaker":null,"is_sponsor":0},{"start_s":508.82,"end_s":512.98,"text":"Either like a background or like I want a space station and then you have a space station","speaker":null,"is_sponsor":0},{"start_s":512.98,"end_s":517.66,"text":"but a student is to start getting to like weirder things or like more niche things or","speaker":null,"is_sponsor":0},{"start_s":518.7,"end_s":527.5,"text":"more unique things like I Don't know a mountain that's upside down, but in a space station that sort of thing a language model and","speaker":null,"is_sponsor":0},{"start_s":527.66,"end_s":536.14,"text":"and Diffusion model might like start to struggle with that because there's never been any data on that and it can only smush","speaker":null,"is_sponsor":0},{"start_s":536.5,"end_s":539.7,"text":"Different data points together so well and I do expect it to get better","speaker":null,"is_sponsor":0},{"start_s":540.54,"end_s":545.1,"text":"But it's also really difficult to prompt those That kind of creativity","speaker":null,"is_sponsor":0},{"start_s":545.1,"end_s":548.66,"text":"However, saying that you can use diffusion models as a tool","speaker":null,"is_sponsor":0},{"start_s":548.78,"end_s":554.38,"text":"So I know some artists who use them to create characters or like you want to do D&D art that sort of thing","speaker":null,"is_sponsor":0},{"start_s":554.86,"end_s":563.02,"text":"Or you need inspiration Or you would like, you know, 500 different dynamic poses that feature","speaker":null,"is_sponsor":0},{"start_s":563.5,"end_s":568.74,"text":"You know, like a ballet dancer or something like that. I want 500 ballet poses so that I can","speaker":null,"is_sponsor":0},{"start_s":569.66,"end_s":576.9,"text":"Get an immediate mood board as inspiration to do my own drawings or my own digital art or you know, even analog art, right?","speaker":null,"is_sponsor":0},{"start_s":576.9,"end_s":581.82,"text":"I think The same can kind of be said for the LLMs","speaker":null,"is_sponsor":0},{"start_s":582.62,"end_s":586.54,"text":"the large language models because they","speaker":null,"is_sponsor":0},{"start_s":586.54,"end_s":592.58,"text":"They spit out text, but you can't really have a conversation with them for very long","speaker":null,"is_sponsor":0},{"start_s":593.14,"end_s":595.62,"text":"Their brains don't work. I think that there's some","speaker":null,"is_sponsor":0},{"start_s":596.62,"end_s":600.22,"text":"There's some like papers out right now and there's also some","speaker":null,"is_sponsor":0},{"start_s":600.9,"end_s":606.58,"text":"Like prize pools if you can get it to go over like a hundred thousand token","speaker":null,"is_sponsor":0},{"start_s":607.62,"end_s":610.9,"text":"Inference length so like a history so you could have a conversation with it","speaker":null,"is_sponsor":0},{"start_s":610.9,"end_s":615.54,"text":"that would be about the length of Lord of the Rings and it would be able to remember and","speaker":null,"is_sponsor":0},{"start_s":616.42,"end_s":623.14,"text":"Make reference to anything in that entire block of text, right? That's kind of the that's the goal","speaker":null,"is_sponsor":0},{"start_s":623.14,"end_s":627.02,"text":"Right, you can you can actually have real conversations with them for a long time","speaker":null,"is_sponsor":0},{"start_s":628.1,"end_s":631.86,"text":"the The local models also support characters","speaker":null,"is_sponsor":0},{"start_s":631.86,"end_s":636.7,"text":"So I'm trying to create a D&D character for a friend of mine so that he can actually have a","speaker":null,"is_sponsor":0},{"start_s":637.18,"end_s":641.22,"text":"Conversation with the character and you do that with a bunch of different prompts and things like that","speaker":null,"is_sponsor":0},{"start_s":641.22,"end_s":644.54,"text":"There's already a bunch of like web stuff that you can do online","speaker":null,"is_sponsor":0},{"start_s":644.74,"end_s":650.3,"text":"Like a lot of people have already like taken this and go like D&D. We can do that on the internet hell","speaker":null,"is_sponsor":0},{"start_s":650.3,"end_s":655.02,"text":"Yeah, and so these are available like through web portals and stuff like that, but running them at home is fun","speaker":null,"is_sponsor":0},{"start_s":655.06,"end_s":659.94,"text":"So, yeah, that's a long-winded answer. I don't remember where the question was. I'm sorry. I'll go","speaker":null,"is_sponsor":0},{"start_s":662.3,"end_s":665.66,"text":"Should we use them should we use them? I think using them as a tool is fine","speaker":null,"is_sponsor":0},{"start_s":665.66,"end_s":669.5,"text":"I think there's no danger in them like replacing people","speaker":null,"is_sponsor":0},{"start_s":670.3,"end_s":675.58,"text":"But they're gonna work with people and there's no there's no escape from that now that genie is out of the bottle forever","speaker":null,"is_sponsor":0},{"start_s":675.94,"end_s":681.9,"text":"All right, all right We talked a lot, but is there anything that we haven't said that you like to say to the float planers?","speaker":null,"is_sponsor":0},{"start_s":681.9,"end_s":684.98,"text":"I was what is this video about? I'm sorry","speaker":null,"is_sponsor":0},{"start_s":684.98,"end_s":691.18,"text":"It's just like I just want to get people's thoughts on AI because you know We are a tech slash creative company, which I think aligns with both with AI","speaker":null,"is_sponsor":0},{"start_s":691.18,"end_s":694.9,"text":"So it's like I think people have some opinions and I think it's good to share. Yeah","speaker":null,"is_sponsor":0},{"start_s":695.82,"end_s":701.22,"text":"Try it at home. It's fun at home. It's a fun little exercise. It's really really simple to get going and","speaker":null,"is_sponsor":0},{"start_s":701.94,"end_s":705.3,"text":"It's it's a fun little break playground and you don't have to like","speaker":null,"is_sponsor":0},{"start_s":706.82,"end_s":713.78,"text":"Spend money on chat GPT. I mean chat GPT is still the best, but if you play with it at home, you're yours","speaker":null,"is_sponsor":0},{"start_s":714.42,"end_s":721.42,"text":"You can play with it as much as you want. Yeah. All right, Dan. Thanks for any of you. I'm still sweaty. It's hot","speaker":null,"is_sponsor":0},{"start_s":724.1,"end_s":726.1,"text":"Thank you","speaker":null,"is_sponsor":0}],"full_text":"I am here with Dan. Hey, and what do you do for the company here? I'm the infrastructure and technical specialist these days. I pretty much just write documentation Sounds good. Sounds good. So I'm just asking everyone here at the company what the thoughts on AI is the current state of it So I'll ask you first. Sure thoughts on the currency of AI. It's pretty interesting there's a lot of cool stuff going on and I'm excited to see it kind of develop over the next I guess a year or a few months even So you talk kind of sound like you're for AI more than against AI Yeah, I would probably say my thoughts on it are a little bit mixed I think it's not maybe as terrifying as people think it is but yeah, certainly certainly for it Sounds good. Sounds good. So if you're for it, then what's one thing you're really against in terms of the state of AI at the moment? I think probably some of the protectionism. I mean Looking at some of the requirements for doing these trainings You're spending millions of dollars on like cloud infrastructure GPU time just to actually create a reasonable model and so People I think are willing to protect that rather than just open sourcing it and letting any other company who has access to these massive data centers Like get in there, right? Okay, okay So have you used any AI currently like that where it's like they have chat GBT the image random readers music Have you use any of it? Yeah, I've played with pretty much all of them and I actually run a couple at home So I run an LLM as well as stable diffusion for image generation and I run those locally I recently just bought a couple more 30 90s and you know upgraded some of them with VRAM so that I could try some of the larger models So they're like a chat GBT for But you can only talk to them for like I don't know Two to five minutes before the like token inference length gets too much and it takes like a minute to respond to you But initially they're they're really quite powerful And they're a little bit more entertaining because they're the uncensored varieties. So what do you mean by that? So, you know, you say something to chat GBT for and it goes as a large language model I can't respond to this the uncensored one the uncensored ones are kind of designed as training bed for censorship and so you can kind of get it to talk about whatever you want or You know, sometimes chat GBT for will Catch things that it thinks are offensive and it isn't I have some people have told me some stuff internally That they use chat GBT for to have like difficult conversations about difficult topics and practice talking to actual people about these difficult subjects and Initially GPT 3 was kind of okay to talk about these but GPT 4 just won't even broach these subjects that are kind of like You know, they're they're hard to talk about is it cuz like they're trying to market it. I guess I'm assuming or yeah I was probably a marketing thing. There's also a liability thing probably You know people like to post what they talk about and then of course you also have like Really negative sentiment and things like that and also you can ask it dangerous things and get it to tell you dangerous information and so it's really kind of Scary in that way and I think as these companies are large and have a lot of money invested in these it's a little bit Be a difficult for them to allow that kind of like unfiltered information because they are so Confident at all the time. They just will a hundred percent confidence in what they're saying Even if they don't know what they're saying and that I think can be really dangerous A lot of people don't understand that it's fake information a lot of the time Another cool thing about running them locally is that you get to like mess with their brain a little bit So you can adjust the level of inference and the level of like token following guys called config scale or sorry configuration scale and You can make it those just be like Kind of insane if you go too far you can get into something called hallucinations where The inference models like feedback into themselves and it just says like random garbage That's can be really quite funny and and interesting. Yeah, it's neat watching them tick like that Okay, so you use it on a person level, but do you think as a tech slash crave company? We should be using AI on any capacity I actually did some training for us. I did a fine tune of chat GPT for We were trying to get it to write our with a hell of a cult showcase scripts Which are pretty boring and basic videos where a brand will be like Write a sponsored segment about our product and here's a bunch of Specifications and then we kind of have to turn that into a script and it's not very creative and it's not very fun So I made a data set from about 50 of these scripts You know because you have to you have to set a Prompt and a response so the prompt was like here's all the data and specifications on the brand and the response is our script right and So I went this data set I did a quick fine tune and it wasn't performing very well because the minimum data that you need for one of these types of fine tunes is 500 Right, so 500 scripts our LTT's we have about 5,000 so that would be a reasonable data set to feed into a fine tune, but you know LTT's are special and You know there they wouldn't really work But these showcase scripts are pretty Boiler play Unfortunately, I was told after I asked for the other 450 scripts that we have only ever done 50 of them so that project Through It was a it was a waste of a couple days of data compilation But did you learn a lot like what to do what not to do kind of thing? Yeah. Yeah, I learned a lot about how fine tunes function and You know, there's there's also a lot of issues with model overfitting So you can you can train a model too much and then it can only do One thing and that's kind of a dangerous balance to and so you know if you feed in that 50 data set and you like You really try to make it Do that 50 data set then you'll just get garbage because it's over fit And if you try to go anywhere outside of that entire limited data set like if you train on an LLM on 100,000 pictures of dogs and their only face on Then you ask it to do a side-on picture of a dog and it can't because it's over fit for like front-facing dogs Or if you ask it to make a cat it can't do that like we couldn't feed in 50 showcase scripts Or 5,000 showcase scripts and then have it make an LTT because it would just make showcase scripts Sorry, so we should Didn't use it for a company or Well, I mean this is kind of the problem and this is something that I've realized more playing with them at home and getting it to let it spend a lot of time with them that you wouldn't be able to do with like a Chat GPT-4 or one of these web things because I'm sending like a hundred thousand or two hundred thousand tokens through These platforms and that would cost me like a five hundred dollars a day to actually get to play with them But I think even more now. I'm more convinced that they are just tools I know some artists friends of mine have been concerned about them but what I've noticed is that Because these LLMs and the image diffusion models are trained on massive quantities of data the Massive quantities of data that they have are like the most generic pictures Right there the generic pictures. There are single sort of style of picture there are sort of a You know, it's it's one thing and so I think at least for artists what's gonna happen is maybe there'll be less of a demand for This basic art, you know, I want, you know a pin-up of a girl or I want a scene of a mountain that sort of thing You know, there's already a massive quantity of art out there that has those and you can use a diffusion model to create that kind of Either like a background or like I want a space station and then you have a space station but a student is to start getting to like weirder things or like more niche things or more unique things like I Don't know a mountain that's upside down, but in a space station that sort of thing a language model and and Diffusion model might like start to struggle with that because there's never been any data on that and it can only smush Different data points together so well and I do expect it to get better But it's also really difficult to prompt those That kind of creativity However, saying that you can use diffusion models as a tool So I know some artists who use them to create characters or like you want to do D&D art that sort of thing Or you need inspiration Or you would like, you know, 500 different dynamic poses that feature You know, like a ballet dancer or something like that. I want 500 ballet poses so that I can Get an immediate mood board as inspiration to do my own drawings or my own digital art or you know, even analog art, right? I think The same can kind of be said for the LLMs the large language models because they They spit out text, but you can't really have a conversation with them for very long Their brains don't work. I think that there's some There's some like papers out right now and there's also some Like prize pools if you can get it to go over like a hundred thousand token Inference length so like a history so you could have a conversation with it that would be about the length of Lord of the Rings and it would be able to remember and Make reference to anything in that entire block of text, right? That's kind of the that's the goal Right, you can you can actually have real conversations with them for a long time the The local models also support characters So I'm trying to create a D&D character for a friend of mine so that he can actually have a Conversation with the character and you do that with a bunch of different prompts and things like that There's already a bunch of like web stuff that you can do online Like a lot of people have already like taken this and go like D&D. We can do that on the internet hell Yeah, and so these are available like through web portals and stuff like that, but running them at home is fun So, yeah, that's a long-winded answer. I don't remember where the question was. I'm sorry. I'll go Should we use them should we use them? I think using them as a tool is fine I think there's no danger in them like replacing people But they're gonna work with people and there's no there's no escape from that now that genie is out of the bottle forever All right, all right We talked a lot, but is there anything that we haven't said that you like to say to the float planers? I was what is this video about? I'm sorry It's just like I just want to get people's thoughts on AI because you know We are a tech slash creative company, which I think aligns with both with AI So it's like I think people have some opinions and I think it's good to share. Yeah Try it at home. It's fun at home. It's a fun little exercise. It's really really simple to get going and It's it's a fun little break playground and you don't have to like Spend money on chat GPT. I mean chat GPT is still the best, but if you play with it at home, you're yours You can play with it as much as you want. Yeah. All right, Dan. Thanks for any of you. I'm still sweaty. It's hot Thank you"}