Bullsh!t Machines, Episode 01: AI for idiots, by idiots. A beginner's look at artificial intelligence.

[00:00] It's AI for idiots. Bye, idiots. [00:04] Welcome to Bullshit Machines. I'm Natasha Hoskins. I'm Dina Burke. This is a series on the basics of AI brought to you by Voice Club. Expect accessible intel on artificial intelligence. [00:16] What was your... [00:18] First magical AI experience. I didn't have it until Easter of this year, so I was late. I was late to the game. Late bloomer. At family Easter with my husband's family. Cute. In Pennsylvania. In an AI bot searching for... [00:33] Barbie as Margot Robbie taking communion. [00:38] in a Catholic church. And I was desperate to get this image so that I could share it on Twitter and say, [00:46] This Barbie's body is the body of Christ. [00:51] It's the dumbest shit in the world. That's my magical AI moment. It's so dumb. And did you get your magical? It wasn't right. We didn't get there. It really struggled with the cracker component, the communion cracker. Okay. So I was using ChatGPT, chatbot. Okay. Okay. [01:11] Famous. She is the Beyonce of the chatbots. A-lister. A-list. So I have spent a lot of money on lawyers in my life in the context of business, not like in my personal life. Yeah, she's litigious. I've never been sued or anything like that. But in business have just...

[01:34] You spend so much. Yeah, totally. Every time they're answering an email, you're spending $300. So I – [01:39] really bristle at that expenditure. I really don't like it. And so one of the first times I was like, okay, this is incredible, was putting the prompt in for ChatGPT to write a contract for me. And... [01:54] It obviously did it in seconds. Yeah. And... [01:57] I am not a lawyer. Yeah, RIP to my husband who just spent – [02:02] A pretty penny becoming a lawyer. Lawyers are mad. I understand why. That's some power. But it spit it out and – [02:09] I used it. And so it felt... [02:13] Like not only a huge time save, but also a huge money saver as well. And, [02:19] I will just say, if anyone is listening to this or watching this and [02:23] They receive a contract from me. [02:25] To not interrogate it too much. To just go with the flow. Be chill about it. [02:33] *outro music* [02:39] We are the co-founders of Voice Club, which is a community of people who love to learn about emerging technology together in community. And we have done that with other organizations. [02:53] I hate to use the word here because it might be a trigger. Trigger warning. Trigger warning. Crypto and Web3. We started this community about a year and a half ago. Really...

[03:04] as a call for... [03:06] context for folks who are wondering how [03:09] these new technologies were impacting their lives and careers and finances. And so we built Boys Club really to address that, where we wanted to learn together and have conversations about what was going on. [03:20] The same moment is happening again with AI where we're all looking around and kind of going like, what the fuck? Yeah. And what's this going to mean for my future and the things that I care about? [03:32] And yeah, it feels like there is a train just like flying down the tracks and [03:37] And you're either on it or you're off it. It's the feeling. And I... [03:43] had that same experience with crypto on Web3. [03:46] And an even more amplified way that feels like what's happening in AI. I want to be on the train, please. I want to be on the train. And I want to pull all my friends onto it too. And to interrogate that person. [04:00] technology and also interrogate the feeling around, am I going to get left behind if I don't understand this or am I [04:09] not, not, [04:10] interested in it or learning about it or a part of the movement. So to be clear, we're not AI experts. We're not, but we have listened to an unprecedented amount of bad AI podcasts. Yeah. But the intention of this podcast is to learn in community and we hope that you [04:25] You're on this journey with us. Yes. So let's dig into it a little bit.

[04:35] Why... [04:37] Is everyone talking about it right now? Okay. Yes. [04:40] What's happened? I was like, what is she going to ask me? I got so nervous. Okay, so... [04:44] The reason I think that it's feeling right here... [04:48] is because there's now this application of ChatGPT. When ChatGPT launched, it launched. [04:55] was the fastest growing application on the internet. That is insane. [05:00] I think a million people were using it within two days and, [05:04] 100 million people within two months. That's why... [05:07] you feel like you can't escape it. We were at Soho House yesterday and there was a table next to us [05:13] talking about how they work in real estate and they're going to be using this tooling to like aggregate listings and do all this stuff. [05:20] My understanding is that there is two main things that happened, which has caused a [05:25] major inflection point. The first one is that there have been material advancements and breakthroughs in [05:33] machine learning algorithm and development. Machine learning is a fancy [05:39] complex word and term to define, but basically it's the idea that [05:45] You give your computer some instruction and it takes that instruction. And over time with less and less human intervention, it's, [05:53] the machine is able to learn and get smarter and produce more reliable or accurate or persuasive results. Cool. We have been using machine learning tools

[06:06] algorithms for some time. We are on Spotify. [06:10] That is a machine learning algorithm that's working to provide better and better suggestions to keep you on Spotify for longer. And depending on what time of day it is and other music you've listened to and other people in your neighborhood and all of these sort of inputs... [06:24] produces a prediction that... [06:26] the machine thinks you will like. Yeah. So you're saying what's happened is those algorithms have gotten smarter. We've gotten better at training them and technology has developed around them. Exactly. So that's number one, breakthroughs in machine learning. And number two is the... [06:41] data sets have gotten really big. So basically there's these things called large language models, LLM. You'll hear that thrown around a lot. [06:52] And LLMs are... [06:54] built [06:56] with [06:57] Huge amount of data inputs. And the data that they're using is all of the text that's on the internet. All your dumb tweets, all your stupid stuff that you post on Instagram, all that stuff. It's basically fodder for these LLMs. And we as a species have just produced so much stuff. Yeah. Okay. And every day are producing more and more stuff. And so... So you're saying that got to like a tipping point. And that's what all of these LLMs are being built with. And so it's really these two things that have caused a very material change in... [07:26] the AI industry. Then to your point, the user interface of ChatGPT has captured the [07:34] imagination of everyone and has entered the zeitgeist in a way because it's come to life for us open ai has been doing this since 2015 they've

[07:41] They've been fucking around for years. And... [07:45] I actually probably think that some of these advancements that I just mentioned, like they may have happened years ago, but it's only because we have now this user interface that... [07:55] with both chatbots and generative AI images and video. So you talk about these two [08:01] big moments in the tech, the data sets, and the machine learning algorithms getting better. And then for me, I think there's two... [08:11] consumer experiences that... [08:14] have created this pervasiveness of this technology. [08:17] ChatGPT and chatbots becoming really useful. And I think we can talk about why we think they... [08:23] are different than a Google. And then the second is the visual stuff, where you hear this term a lot, generative AI. And essentially what that is speaking to is the idea [08:34] text to image generation. So my example of Barbie taking communion, [08:40] That is an example of generative AI where you use a tool – [08:45] that [08:47] is able to take your inputs and through stable diffusion, which is another buzzword that you'll hear all the time, and math and coding and all of these things, produce a new generative image based on these inputs. And we've all been duped by the like Pope, Drip, Balenciaga moment recently because these... [09:09] tools have gotten so good at producing images and videos that look very real. We've gone from like keeping up with the Kardashians to like Hulu Kardashians in terms of

[09:20] pervasiveness culturally. Totally. Do you agree? I do, yeah. I think a picture is worth a thousand words. And I think that viral images of the Pope dressed in Balenciaga. That's the pinnacle moment. It was like as dumb and as stupid as that is. And I'm sure all the AI researchers are like the very serious academic people. They're just like people of God. But yeah. [09:41] It worked. [09:41] It worked and it brought it to life for people in a way that I think is remarkable. It feels like a toy right now, but we can extrapolate out from here into a near future where... [09:55] It has very... [09:57] serious implications on all of our lives. My experience has been I'm learning these things and sort of [10:04] silos and then [10:06] fitting them together into where they all sort of as a puzzle start to make sense in this broader technology. [10:12] has been important to... [10:14] have a feeling of understanding it. Yeah. There's... [10:18] something that happens [10:20] when you're learning something new, especially with technology, where there's so much jargon that's used. Yeah. NLP, LLM. [10:28] neural networks, like deep learning. It's just jargon, jargon, jargon. And it's really easy to [10:34] approach one of those words or read one of those words or hear one of those words and [10:39] think [10:40] I do not understand this at all. And so... [10:45] This is just not a place for me. [10:47] This is not for me. It's impenetrable. I'm not smart enough or in touch enough to be able to understand this industry and then set the whole industry or the whole thing aside. And I think.

[10:57] That is a shame. And at some point that word will make sense or it won't and it's fine, but it doesn't mean that the whole thing happens. [11:05] I need to disregard entirely. It doesn't mean that you can't participate. It does not mean that I can't participate yet. [11:10] Something I kept coming up against that I couldn't reconcile was – [11:15] how is this tool different than Google? And is it just a better... [11:19] organization of search and, uh, [11:24] What I have found has been... [11:27] Google is crawling the internet. [11:31] going through all of these things [11:33] millions and [11:34] billions of websites and finding information and indexing it and then surfacing relevant information. [11:42] So if I Google search... [11:45] Gwyneth Paltrow smoothie, goop smoothie. It's going to go through the internet. It's going to find all these goop recipes and it's probably going to surface the one that is- [11:54] Best SEO podcast. [11:55] optimized and that's what will happen. Yeah. [11:59] If I go to ChatGPT or another chat bot... [12:04] And I write, create a smoothie recipe in the spirit of Gwyneth, in the vibe. Can I do it? Let's do it. In the vibe of goop. [12:15] It will not search the internet and find a recipe that exists on Goop. What it will do is... [12:24] look through all of its data, this crazy amount of data that we've talked about,

[12:28] and generate [12:31] A... [12:31] New recipe. [12:33] Something that [12:34] is based on the predictions of what other recipes on Goop have been to produce a new recipe for me to use. And that's the main difference. It's like producing something... [12:45] And that you can get into a whole philosophical conversation of whether that's real creative generation or not. But it's not a link to something. It's a compilation of this data into a predictive set. So spirulina... [12:59] We didn't get spirulina. We did get chia seeds. So yeah. Okay. Smoothie recipe inspired by goop that incorporates healthy ingredients for refreshing and nutritious. It's selling it. Oh, refreshing and nutritious drink. Goop inspired green goddess smoothie. Okay. [13:14] Avocado, spinach leaves, cucumber, pineapple chunks, lime juice, chia seeds, unsweetened almond milk, optional one tablespoon honey, or a natural sweetener of your choice. Ice keeps as desired. A second... [13:26] AI in action moment was today. [13:29] In our weekly AI newsletter, [13:32] Sam on our team wrote an incredible line. [13:36] That goes, maybe she's born with it. [13:38] Maybe it's Mid Journey, which is an unbelievable line. Honestly, give her a Cannes Lion. An Emmy. [13:45] Okay. Mid Journey, for those who don't know, is the leading – [13:49] a generative tool for images [13:51] It's an incredible line because you're seeing all of these AI generated images. And so what I did today is made a carousel for Instagram of all of these different images

[14:01] versions of Maybelline inspired because the original line is maybe she's born with it. Maybe it's Maybelline. If you don't know that. [14:08] Why are you listening to this podcast? All of these different AI generated images that are inspired by... [14:14] ads, Maybelline ads. What was your experience like in the mid-journey Discord? It was insane. So mid-journey, [14:22] All happens in Discord. [14:24] Which is such a bummer. Such a bummer. Discord is a – The worst. Like Slack for gamers and crypto people. It's a necessary evil IMO. Mid-journey happens in Discord and you backslash imagine and then you put in your prompt. [14:40] for it to generate an image. It's a chaotic space where all of these different people are putting in their different inputs and they're... [14:48] They're working, massaging their different images to get closer to what they wanted. And I saw a lot of people trying to get the Pepe image. [14:58] I know he's like his own person, or his own character, but to me, he's just the PepeCoin guy. And you're all in this together. You're all in it together, which is sort of weird. [15:08] For so many reasons. But one of the reasons is, and this also gets into some ethical issues around AI, but I really wanted a diverse set of women in these Maybelline areas. [15:18] images and it was just giving me a lot of white women and then a lot of blonde women. And so then I had, I was getting more specific, trying to get more diverse set of images and [15:28] I was struck by, this feels like, [15:30] private work, creative process. And I'm doing this in a very public way where other people are just seeing my prompts. And I was actually learning a lot about how to prompt by being in there and seeing, okay, what makes a good prompt? What makes a bad prompt? What makes

[15:45] or not bad. There's no bad prompts, but it's more specific prompt. Right. And what's crazy is you could write Kendall Jenner, [15:54] eating a taco, [15:56] in the streets of New York city inspired by, in the style of Terry Richardson, problematic figure. Sorry, but just can't find. And yeah, [16:05] It would generate something that looks aesthetically a lot like a Terry Richardson image. And that is just wild. And the more specific or prompt in many cases, the better or the more reliable that output would be. Yes, exactly. I went into one of those Discord channels once and I was like, oh my God, get me out of here. Unsubscribe as quickly as possible. Just the level of chaos. But I was struck by... [16:30] Yes. One, I think it's a really interesting sort of like creative exercise to be in a [16:36] shouting amongst a group of people to try and get some result from the computer, especially in creative pursuit. Like it's just fascinating. But the second thing was so many images of... [16:47] woodland nymph women. Oh my God. [16:51] Or sexy mermaid women or... Pretty woman in nature, like a lot. The other thing that I see a lot is I was trying to learn about stable diffusion as a concept. [17:03] How'd it go? [17:06] We're not defining it on this podcast. Let's skip over that one for now. I went to Twitter. [17:13] the bowels of Twitter, which usually can be a tool for learning. And I wrote in Stable Diffusion in...

[17:20] Search. Mm-hmm. [17:22] And [17:23] I'm not kidding you. Every single top search was an Asian woman with just huge boobs. Like huge. And I was just like, what – [17:33] I was like, this is a bad prompt. Yeah. I'm doing something wrong. But also what is happening? Like what is honestly happening? It speaks to exactly why I want to do this podcast, which is that what you found with the big-breasted women and what I found with like the woodland nymphs is that's lacking creativity. Yes. Yes. [17:56] It's so... Basic. Basic. Yeah. And just, we're not even at first base of what's possible with these generative AI images tools. We have not left the home base. Yeah. The possibilities are endless and that's almost what makes it so hard to use and to understand how to leverage as a tool in our lives and in our professional lives or in our creative lives because it's like unbounded, right? Yeah. And it's possibilities. But I think also just speaking to the sheer... [18:26] lack of imagination that I'm seeing in what's being created by men. We need to bring some more voices in and some more perspectives in so that we can get [18:38] this. [18:39] creative [18:40] use case of AI to be just a lot more interesting because the possibilities are literally endless. There's a lot of very like scary rhetoric. I have one. Revenge porn. [18:49] it's a cousin to the conversation around the woodland nymphs. Yeah. Which is, I'm going to...

[18:56] to create an AI porn with this woman that I hate and use her face. And you're already seeing that with celebrities. That's the tiniest tip of the iceberg of what's possible with... [19:09] these tools. [19:10] on both a celebrity... [19:11] level, but also on an individual level and personal level as well, as these tools get more and more powerful. And... [19:18] You can just imagine... [19:20] Especially as they become more convincing how damaging that can be for people. And that is terrifying. [19:27] What are some things that you're concerned about? [19:30] I'm going to do a coin. [19:32] Like a good side, bad side or shadow side. Okay. So... [19:36] Something that my new hero, Sam Allman. I know three things about this man. And I'm like, oh, I found he has no equity. And I was like, he's a hero. Yeah. Based from Chicago, so salt of the earth. Middle America man. One of the things, if you were to ask him, why are we doing this and why build this? There's a lot of very scary rhetoric around what could happen as this technology develops. [20:04] One of the things I've heard is... [20:08] health and [20:10] medical advancement. So the ability to develop research for pain and suffering [20:17] cancer, diabetes, other terminal illnesses. [20:20] With a tool that can take massive amounts of data and make sense of it and look for patterns and find new solutions, we are...

[20:28] going to be able to leapfrog [20:32] in [20:33] development around medical needs and medical solutions. So that's incredibly helpful and a reason in and of itself to continue doing this type of research. What's the shadow side of that? [20:45] The shadow side, these tools are being trained by human beings. [20:49] to make decisions and make predictions. And [20:53] When... [20:54] a human being is doing that work, we all have bias. And what that means is these [21:01] tools are being trained in such a way that that bias is going to be written into the way that these predictions are produced. And so when you think about the medical field and [21:12] the type of research that has been done and the types of communities that are often prioritized [21:17] It's not usually possible. [21:19] disenfranchised communities or poor communities? And are we going to see a world where all of that research begins to [21:27] solve for white wealthy individuals and not for [21:33] humanity at large. [21:35] There's a lot that we can continue to say about the ethics and the inherent bias that's being written into politics. [21:41] these algorithms and these large language models, but... [21:45] That's one that I think about in relationship, especially when you start talking about. [21:50] health and well-being and developments around that. [21:53] Yeah. [21:54] I think if I'm really looking at it, looking at AI, [21:57] And looking at all the problems that people are talking about.

[22:00] my honest feeling is, [22:02] is I'm just not scared. I can't conjure the feeling that... [22:08] I'm not going to have a job. Although I will say that that's what everyone thinks. [22:13] It's a phenomenon. Okay. Where, where... [22:16] we're having this big conversation about the future of work and whether or not we're all going to have jobs and will this [22:22] affect blue-collar workers or white-collar workers? And where is the value of AI going to accrue? Is it going to accrue evenly? Or is it all going to go to Sam Allman's brother or cousin or something? The tech elite. Not him because he has no equity. Not him because he has no equity. Everyone... [22:40] is in that conversation now. And everyone thinks that their job will be the one job that won't be affected. That's not how I'm thinking about it. Okay. It's not like, oh, my... [22:49] But my job won't. It's [22:51] that [22:52] I have a real belief in the humanity's ability to adapt to technology. [22:58] And... [23:00] that [23:01] For many centuries, technological developments have happened and people have had there's been this rhetoric that we're all going to none of us are going to have jobs anymore. We're all going to lose our jobs and the economy is going to go into a downfall and it's going to be the end of human race. And that happened with factories. [23:19] And then that happened with the internet and... [23:22] our ability as human beings to [23:24] take this incredible technology and improve our lives, do more, learn how to use it and make it

[23:33] make the work more efficient and effective [23:36] we've been able to do in these major inflection points around technology prior to this. And [23:42] I think my most honest belief is that that is what will happen here, is that [23:46] My hope is that it will make us more creative and... [23:50] make us have new eyes to see the world be less visible [23:55] bound by things that we couldn't do before. And that will produce a lot of good in the world and a lot of new work and new fulfillment in that work through this tech. [24:06] That's my hope. [24:07] I think that's a deeply privileged position to look at this technology because [24:12] sitting here as a white woman, I can say, oh, it's all going to be okay. Like they're all going to figure it out because in many ways – [24:18] Thank you. [24:19] I have always been taken care of. [24:22] I think there are very big questions to be asked and there we need to be looking at it and people much smarter than me and much more in the development of this technology need to be very present to that and need to be sounding the alarm. [24:35] sitting here today, what is my POV on it? It's exciting. And there's some really [24:40] amazing things that I think will come out of this. Yeah, without stepping over whenever there is change, there is loss in some way. Yeah. [24:46] And that will happen. [24:48] My feeling is that [24:51] there's [24:53] unbounded potential in [24:57] the education... [24:59] application. [25:01] Right now I could... [25:03] Learn what AI is. Learn what AI is.

[25:07] I could ask ChatGPT to... [25:10] I don't know. [25:12] give me beginner's lessons for Mandarin. Okay. And I could have a personalized tutor. That's really powerful. And I think that applied... [25:22] at a global scale, [25:24] could result in better education than any of us have had, but for everyone. Yeah. And that's remarkable to think about in terms of what it could unlock for human potential and value. And so I am really excited by that. The shadow side of that is... [25:41] We're going to have to learn how to learn with these things and how to use them not as crutches, but as things. [25:47] enhancements to our own human thinking. And I think about that a lot as it pertains to my kids. [25:55] And how... [25:57] They're going to be growing up in a world that is AI native in many ways that I can't even anticipate what that's going to look like. Yeah. I can't even imagine it. So I don't feel like I have the tools to teach them what. [26:11] what they need to know in order to be able to properly cope with that. My six-year-old [26:16] It's a constant struggle around screen time. [26:20] Right now, and he's just playing Minecraft. And it's like little blocky things. And it's not... [26:26] responding to him [26:28] knowing him and his preferences and what he likes and what he doesn't like and what he... [26:33] thinks is cool and [26:35] If that feature were turned on within Minecraft, like I would never be able to get him off it. So, and that's because he doesn't have yet the coping strategies and I haven't.

[26:46] taught him how to deal with that and how to create that distinction between him and the computer. That's just an ambient feeling that I have around it, where it feels like this [26:56] big parenting problem that I'm going to have to be solving for. And I don't, [27:01] know what the problem, how that's going to be shaped yet. It's like a little bit of a amorphous anxiety, which I love. [27:09] I love just a low-key, always-on anxiety. Just here. It's here all the time. It is here. Yeah. The last thing that I'll say around this is – and then I'll shut the fuck up about it. Great. Great. [27:22] Is – [27:23] We are in a different way using these technologies every day. [27:28] And using other types of tech that have made our lives so much better. Like I use Google maps. I use spell check. I use Google. I use all of these things every single day to, uh, [27:38] make me a better writer to help me be more efficient and effective and [27:42] thoughtful in the way that I show up in the world, there are 100% downsides. But I think if you're looking at this and you're feeling scared or you're feeling [27:50] Like there's this big, huge, massive thing that is happening and you don't know what to do with it. [27:57] My best guess of what's going to happen is it's going to settle into some new reality for us all that makes our lives better. And that's a really like hopeful, like kumbaya way of looking at it. [28:08] But I just want to think about what could go right. [28:10] You know? Nice. That's the invitation. Welcome to the right. I like ending on that. Thank you. We're going to be doing more of these. We're going to be inviting on people who are much smarter than us to talk about-

[28:20] experts to talk about [28:22] how AI is going to show up in different things that we're interested in, like creative fields and, uh, [28:29] our social lives and our work work. So, [28:34] Expect that. This is a series, so we're going to be not doing this forever, but doing this for a set amount. [28:41] of unknown episodes at this point. And the next episode that you can expect though, is going to be an overview of the, [28:50] The players. So we've talked a lot about chat GPT and open AI. We talked a little bit about mid journey. [28:57] Don't be scared or worried about what these things are or if you don't know what they are, totally fine. Doesn't matter. We're going to dig into them in another episode. [29:04] where you can get a lay of the land of things that you could use if you wanted to. The players, there's some very interesting stories and drama that's fun to dig into as well. [29:14] So expect that. And thanks for listening. Like, subscribe. [29:18] Follow. Rate. [29:20] Comment. [29:21] at us in the nicest way.