hey how's it going guys welcome back I'm Ricky with tuba DaVinci and we have a very special live stream today I'm with my very good friends Matt Farrell and Alex from ticker symbol U and undecided and we're talking about AI now we we three talk about YouTube all the time together yeah switch over we talk about YouTube all the time and one of the things that we've been talking about recently is our views are down well me and Matt were talking about our views are down Alex is is doing really well why is that and invariably the answer is there's a lot of focus and attention on AI right now and the the advancements are happening so quick I literally cannot keep track of it all between the like last night when I was researching and now there's probably changes but we're going to talk about it and there are some reasons to be excited of course and there's probably a little bit of reason for concern and we're gonna dig into all of it so without further Ado Alex Matt how you guys doing doing well how you doing hey man good how are you good I'm really really glad you guys can make time on a Sunday Sunday morning afternoon depending on where you guys are to chat about some of this stuff but let's start with let's start with the good right all three of us run YouTube channels which means we have a lot of Need for writing and research and and a various slew of of things let's talk about some of the positives and Alex I know without getting too much into your business and stuff you've been looking for researchers and writers and now you're thinking maybe chat GPT can replace it you want to start with kind of what you're thinking or how you use some of this stuff sure yeah so for those of you who don't know me I run a channel that focuses on investing in advanced technology so I'm deep in this space already I care a lot about Ai and it's different applications so when before chat GPT came out before gpt4 before open AI started making the news seemingly every day I had a team of researchers writers and editors right kind of how every little research Channel sort of works um and it's hard right so what would actually happen is you spend a lot of idea and the I you spend a lot of time in the idea room you spend a lot of time outlining researching getting data together from multiple sources vetting ideas fact checking there's a lot that goes into research writing and then coming out with like a coherent story about something but it turns out that a lot of that is mechanical right it's like actually formatting an outline actually getting a lot of facts together and making sense of them and those are things that we're finding that robots can do really well robots natural language models like gpt4 large language models and chat GPT uh which is like the chat interface on top of that model so I've I'll just be frank you know one of the things that both excites me and scares me is in the last 90 days I've La I've gotten rid of three writers and two researchers and I now rely exclusively on working with these models together and you know we should we can definitely get into what that means what the dangers of doing that are for example fact checking uh Source Society and all that but I think this is one of those really important moments in history where a lot of the mechanics of our jobs all jobs are going to go away and we're going to be left with the creative human tasks that are more about putting things together weaving narratives uh using our subjective taste and things like that and so I'm really excited because it's made my job both easier and a lot more streamlined and impactful and exciting and I'm definitely a little nervous because the next job that it's coming for is definitely mine right so I hope we get into that a little bit too Matt how about you um what are some of the ways that you're using chatgpt other AI maybe other tools and stuff in your day-to-day workflow it's kind of similar to Alex like I have a team of researchers writers and video editors as well and I'm starting to experiment with chat GPT systems for helping to kind of brainstorm topic ideas like I know I want to do a video on this new thing but like help me kind of figure out a possible angle I could put on it or doing research of like there's like five papers on a topic I can just quickly send that paper to chatgpt and say could you summarize this paper for me and it will summarize it and I'll make a call whether I should dive deeper into it myself or I should just pass on it because it's not really applicable to what I'm looking for so it's helped me speed up the process of putting videos together and so part of that is I'm super excited about that but at the same time like Alex is saying it's coming from all of our jobs and for me my background is in a creative field and chat GPT is coming so straight at like writer's artists designers uh right now and so it I feel for a lot of my former colleagues my friends family members my brother's a writer so it's like I have this kind of like I'm kind of torn right now I love what's out there as a tool to accelerate what I can do but at the same time I see the writing on the wall for a lot of jobs and it's got me very concerned it's been the latest to the party I've we chat and you guys are both using some of these tools to really incredible results and stuff and I think I've kind of dragged my feet but I have gotten into it recently the first thing I tried to think of is thumbnail ideas right some of our topics that we talk about are abstract it's like my video this week was about a water vapor tower that you could deploy in the oceans that you know harnesses the fact that you have very high humidity air right above the ocean but it doesn't exist it's kind of a theory it's an idea how do you make a thumbnail about that right so I was sitting there just kind of trying to think like how could I make a compelling thumbnail that would kind of harness that idea and if I could use you know uh mid-journey or some of these tools to do that that would be freaking cool so the other thing is we all are now using a tool called Write Sonic which I can't remember who turned me onto that was it Alex it was one of YouTube it was Alex and what was different and why I actually care now is there's a level of trust that has to be earned right and I don't trust these tools currently which means I need to know how you came to your conclusions and stuff and I need to know your sources so I can go double check them and in the past the the first AI tools I was using didn't do that it was just kind of here we go here's the answer and I said are you sure that doesn't sound right I can't check where did you come up with this but right Sonic will give you citations it'll tell you where the data came from and that is when I decided okay I gotta try to start using it and actually I do have a writer researcher and I've made him a credentials on right side and I told them you use it put it through its Paces if it helps you let's find out how this works together but I'm with both of you I can't help but think and leave us your comments I haven't seen any comments just yet I don't know if that part is working yet Juan but I'm curious is there is it excitement I think Alex and your community in the investment Community it's excitement right we're talking new stuff it's always interesting Microsoft and Google have both had kind of keynote events recently and there's just the buzz is is palpable but for like everyday people like if you are a writer you are a researcher Matt I know your brother is a writer like how are you feeling it's got to be it's interesting there's got to be a little bit of uh hesitation or you know or a concern and we've heard this before we've always had disruptions right and when you had the car what are the horse you know uh Traders and horse trainers gonna do well this does feel different though am I crazy about that we've had disruptions before but the age of AI feels different is that fair yeah and then maybe I can just shed some light on some of the things that you're talking about sort of the feelings and just put a little Nuance in there because I think you're right and it's good to talk about why so in the past you know we've had before the internet and after the internet that's like a serious divide in human history in my opinion because all of a sudden we went from snail mail to being able to talk to each other across the world nearly instantaneously uh cell phones were another one moving from sort of like Palm Pilot that era Blackberry to smartphones with no buttons all internet connectivity apps and things like that truly transformational and think about all the jobs that those things ate there's no more calculators there's almost no more GPS units there's almost no more you know you can name many many devices that used to be their own thing that are now smashed together in one one thousand dollar device that you pay like one monthly fee for and it killed a lot of jobs and it made a lot of new ones right the thing that makes this one different uh the most elegant sort of thing that I've heard about this is that it's starting to usurp human judgment so the internet is a tool computers are tools smartphones are a tool but none of them tell you what to do they just help you do what you decide to do with them but for the first time things like chat GPT are like oh I'll write that code for you oh I'll I'll tell you what the points in this essay should be I will start choosing the right language depending on what you tell me about so like this is the first time that technology has started replacing the decision making level uh for for the common man I think and I think you know it's starting to choose more and more topics more and more word choices more and more structural things it's generating images instead of artists not helping them generate images faster right like you're putting in a prompt and then you're selecting from a bunch of computer outputs for mid-journey so I think the reason this feels so different is because it's starting to eat into what we thought was fundamentally human the creative jobs the writing the word choice the Artistry right like nothing has really done that before does that make sense yeah I agree I mean for me like like I said like my background in arts in the Arts I have lots of friends who are artists and put their artwork up onto Deviant Art which is the largest the largest like online resource for artists and a lot of these tools have scraped the internet and grabbed all these artist artwork yeah to train their models without licensing them and this artwork is copyrighted so all these companies have essentially broken the law in what they're doing but the law isn't completely caught up to what's happening so it's like what are we supposed to do it's like do you are you supposed to pay the artists for using their artwork to train the model that's going to put them out of work because then that AI is going to be able to do an artwork that's put into a prompt that's going to look just like their stuff but it's not their stuff it's we're in this weird quandary right now with how do we handle this in our society because this is moving so fast and it's going to only accelerate which is the part that I'm kind of like very concerned by yeah yeah I hope we talk a little bit about the open wetter that just got signed by Elon Musk Steve Wozniak calling for a pause for uh training gpd5 yeah I'm just pointing out that that's a worthy topic we should talk about so what I'm thinking is let's start just yeah I think we should we should 100 do that let's do that in a second so I think we're going to spend most of our time talking about why we should be afraid or Not Afraid but the the consequences maybe some of the things we consider from a legal perspective as Matt mentioned but before we do that let's start with the positives right I mean there's there are some positives here so I don't know if you guys watch the Microsoft uh keynote Microsoft had a big announcement where they announced a new tool called copilot it's basically a large language model that is built into office 365. now I use Google for my business stuff like my email is Gmail and stuff but I'm not gonna lie after watching this keynote I am thinking about just switching over to All Things Microsoft you know teams and and Excel and stuff and the reason is this co-pilot thing is pretty impressive because all your Microsoft data is in one place they can do all kinds of crazy stuff like if you got an email with a spreadsheet you could go hey copilot could you make me a a PowerPoint presentation with this data and it goes through and it's like incredibly intelligent and it gives you a bunch of feedback again right you can go I like this No undo or retry get rid of it and um Microsoft is one of those companies I kind of trust with this because their keynote made it pretty clear that they really went to Great length to think about the ramifications of this like if this doesn't work out well if there are issues if they're wrong and they they admitted like look we're going to be wrong we're going to get things wrong but we're gonna quickly fix and we're gonna keep training and we're gonna keep trying new things and and move it along so yeah they were doing all kinds of really cool stuff where they had like a spreadsheet just a big spreadsheet an ugly old thing that anybody in Corporate America is you know familiar with and quickly it was like make sense of this for me like why was Q2 so bad for us and the it was actually going through the data and finding like patterns and Trends and Reporting it and then it's like cool now make another graph cool now put that into a presentation like you were just writing to this thing and you were creating Excel spreadsheets and charts and graphs and then PowerPoint presentations then it was like okay well mail that to all the people who were in that meeting yesterday about this and it has the contextual knowledge of your you know your your last call with Skype or whatever you used and it sends that email with all that information and your PowerPoint LinkedIn I mean it's like it's an incredible thing but then it got me thinking what will the future look like um part of what they were doing is crafting emails right you guys we've all had corporate jobs all three of us before we were YouTubers um responding to emails and all the email nonsenses one of those it's one of the oldest running jokes right in Corporate America you're spending half your day about this stuff well this co-pilot will take care of that for you so like you get an email from Debbie and Debbie says hey what what are the action items in X Y and Z and the AI is crafting a reply so the is like hey I got that x y and answers your email for you but then I was thinking about this well Debbie has it too and then she goes and has the it so what are we doing now we're just having AIS talk to each other and we're it's a weird place right now even in the positive when I was thinking about all the cool that's cool that's cool it kind of came back to what is the point of us at this point why are the two of us even talking when we're just having the AI generate stuff right um from the from the perspective of of the positives are there other examples you guys can think of or companies you know rolling this stuff out in a more consumer focusing way the that you guys can think of that are positives right now uh Matt do you want to go first or do you want me to go you go first sure yeah so what's let me also clarify a few things right so Microsoft is a 49 owner you know stakeholder in open AI so when you're talking about chat GPT gpt4 also realize that that's a large piece of the infrastructure behind Microsoft's co-pilot as well right so um one other thing I'll add is Microsoft also has its own tool like notion which is like a all-in-one notebook that uh can connect to a lot of things um called Loop and loop is something that interacts with uh copilot so now you can have this like custom board where you have everything and then split things off into other things kind of how you know I know Matt and Ricky use a tool called notion I use a tool called Coda Microsoft has one called Loop and everyone is kind of excited about that one um but there's a lot of like unique benefits uh to the to Ricky's point so let me just go over to one even though everyone has AI co-pilots in their or email replying to each other a lot of email is about leaving a paper trail right it's like hey I want to know what decision making went into this project I look for the right email thread and I have that like process and just because an AI wrote the email doesn't mean it wasn't filled with the relevant information here the Right tickets here's the presentation here are the meeting notes all that right human generated AI generated what you actually get now is a very clean paper trail which is very useful for most of corporate America right like but the other example that's maybe not so sneaky now but like the thing that's going to take down Google is this notion of in-app co-pilots right so not the humanist not the assistant that sits next to you and helps you do whatever you need with Microsoft's data but the one that lives in just Excel and helps you become a power user of excel faster right where before I would go to Google and I would search what is the right function to do what I want in Excel now I just turn to the co-pilot that lives in Excel and say hey I'm trying to make a table that describes the data this way where this cell does this and the output looks like this and the formula should have these things in it and the plot needs to do that and then right inside Excel that copilot starts building that spreadsheet which means I'm no longer going to Google I'm actually no longer even leaving the app I'm trying to work in and my productivity has gone way up because as soon as it gives me the answer the spreadsheet updates and now I'm on to like the next part of the question so so it's my literal in-app co-pilot kind of like how there's pair programming and I'm much more productive now do that in word now do that in PowerPoint now do that in Access and whatever whatever and all of a sudden what we're really going to see is every piece of technology is going to come with an in-app co-pilot that instantly lets you get the most value you can out of it much quicker and that time to value is what people really care about when they pay for software as a service all these expensive subscription packages you know you're just gonna be able to ramp up way faster so I thought that's that's a neat benefit that we're nobody's really exploring too much yet I was gonna basically say something similar of like for me like notion being able to be in the experience of crafting a ticket for a video I want to make and just being able to say the AI could you write me just a short summary based on this and it'll just plop it in there and it's a great place to start the brainstorming process and it's helping me to speed that up there's also aspects of AI and what adobe's doing it's like I use Photoshop and things like a lot oh yeah and so it's like going right into that it's like I can take a photo and I can to say highlight somebody on the photo and say just get them out of here and boom they're gone and the background's filled in and I can tell you as somebody who's to do that manually for a job a long time ago taking something from hours of work down to literally just a highlight push a button and it's just brilliantly just gone it's not it is absolutely insane so for me those those very targeted specific tools of AI I think are just astounding for me it's like I don't we're not going to get into the what freaks us out part yet but it's it's that generalized AI like the chat Bots that can do pretty much whatever you ask them to do that get a little kind of crazy because like one is trying to kind of like replicate a human and how we think and approach something versus a tool which is like just make content aware fill you know what I mean like an AI tool that speeds up your work process it's enhancing us as a human to do our job better where the other one is going to be replacing the human and that's kind of where it's like part of the reason why I love the uh the application approach that Alex brought up well let me let me give you a little bit of a counter example there because like I I think this is like one of those areas where Nuance will really help right yeah I went to Adobe stock Adobe footage right so like you know where uh we all make thumbnails right that's the images you see at the top of every video before you click into it and I was looking for an image and I couldn't find one that needed that like met what I was trying to do but then Adobe generated one for me it was using mid-journey to like solve my problem it's like well I don't have a stock photo so I'll make one for you so what what empowered me as a content creator took somebody else's job right what like these stock photos used to be stock photographers that would make large packages of photos and videos and whatever and sell them to Adobe and then they got a royalty whenever somebody would download something with their license and whatever right but now Adobe has skipped that part all together they'll just generate the photo based on your request show you a thousand generations and that's still cheaper than paying a stock photographer so I was empowered by this tool while it took somebody else's job so it's not one or the other like it's just it's which side of the transaction are you on determines if it empowered you or took your job right like well take that take that one step further as I was doing the same thing just two days ago I was making a I was in my video I needed an image of a kind of a retro 1950s television set so I did a search on Adobe stock I did a search for Retro TV and there was stuff coming up but then I found these like really amazing like 3D rendered like gorgeous photos and I clicked into one and saw and it said generated using Ai and it wasn't from Adobe it was from some guy like yeah Gary so Gary created an account and he's just probably on mid-journey pumping out these images and just uploading them to Adobe stock to sell and this is why I was like this is the part of that freaks me out I have friends who are artists that make things like this to sell this is how they try to make a living and now here comes Gary who probably has no artistic bone in his body body he can just type in prompts into a Ai and pump up a thousand different images that he puts up on stock sites and now he's making a lot of money taking a job away from somebody who spent hours or days crafting this beautiful piece of artwork that they're trying to make a living from so it's like it's I'm kind of torn I'm torn between what we're seeing and how it's being applied and man this is going to like upend a lot of of our society so in the for the good and the bad so yeah so there's two things you guys touched on so the first is in the Microsoft keynote one of the things that they kept bringing up and I could tell it was important to them was the idea that they were grounding the AI and what they meant by that is there's a whole world wide web don't go there but just limit yourself like it's a little bit of a um kind of a forcing function to say just look at my spreadsheets my PowerPoint presentations my emails like don't get crazy but ground yourself in a smaller data set and they found that to be wildly valuable as you can imagine because you're not just making stuff up right you're not reading two articles and saying uh the the sky is red because I saw two poems and that's what I found out right instead you're going through your own data so the idea of grounding data I think is what you guys are both getting at the app level where don't go you know open world limit yourselves a little bit and the second thing comes down to I think how everybody and in the comments please let us know how everybody views technology is if there's a spectrum of nothing to everything everyone's looking for the right level right so Matt mentioned content aware fill in Photoshop if you're if you're familiar with Photoshop is an amazing tool there's a there's a little something right here highlighted delete it and then just tell the computer fill it in and it'll figure out what's around and do it amazing right now we all love that because we still are needed right but the problem is what if the next chapter says oh no you don't need a guy doing all this we'll do all that for you right like the example of adobe stock which is instead of paying a royalty or commission to the people who are contributing your photos now you're just making it yourself so I think my point is everybody wants technology to stop at a level where they're comfortable where they benefit most from it without hurting like for example in the AI self-driving world if truck drivers had a self-driving truck that you could just sit back and put cruise control and watch the game and have your coffee and just relax but they still needed you they would all sign up for it but the next version that comes out that says oh guess what we don't need you at all they wouldn't want right so we're all selfish creatures in a way we all want these tools to help us without hurting us and that's where I think we're all kind of scrambling because like the rate of acceleration is happening so quickly I don't even know where we'll be like two weeks from now or you know here's here's another thought for I want to give both of you guys take on this Matt I'm Alex I'm sure you do as well Matt and I struggle with people stealing our content on YouTube all the time there was actually a viewer thanks to one of you guys who pointed out there was a channel in French literally taking my script Google translating it into French and my little jokes my little ad-lib stuff was still in there that's how I knew they literally just stole my script but we had people stealing our scripts currently they'll take our animations they'll do whatever in this AI world is it possible in a couple of I don't know months or a year where I can go hey here's a URL to a really great Matt Farrell video could you make me a video uh that's about 10 minutes but with more humor and then it'll like make me a script and what is that based on what are the ramifications of that I mean it's it's as a creator this is I mean I haven't covered this before and for our viewers Matt and I have a lot of the the same viewers Alex covers this because that's what his people want to see but do you guys care about AI should we cover this more Matt and I I we we wanted to go live and talk about it kind of gauge this because we're not really talking about it because that's not really our Niche but I I'm sure all of you guys are thinking about this and there's just so much that goes into it but how do you guys feel this is going to end like where does this go are we gonna have channels or is it gonna be like a million Bots just Google builds your own Bots and have their own creators and then fire all of us I don't know is that a is that a concern yeah I'm very concerned about this I think what's going to end up happening I was talking to my brother about this who's a writer I think there's going to be a the way this is going to evolve is that writers aren't going to go away we're not going to go away but our potential audience is going to shrink because there's going to be this kind of these AIS that you'll be able to say make me a video about this topic like that one over here and it will just pump something out for you is going to be good enough content for a pretty sizable audience but then there's going to be an audience that where they're going to recognize what that is and they're gonna want the human touch that human personality that person that's bringing something to it that the AI can't and so but I think that's going to be a smaller piece of the pie in the coming years so it's like I don't think we're in danger or people like us are in danger per se but I think it's going to become harder to be a content creator in the coming years because there's going to be an influx of random channels that just are faceless nothing but b-roll that are just ripping off the content from other people's stuff and it's going to be good enough that it's going to get an audience watching it so it's that's that's my concern it's like this the AI has the chance to kind of devalue the creator what do you think Alex yeah so I'll take the other side of this argument just to like Steel Man it I'm not sure what I personally believe yet right where just to be clear we're six weeks into gpt4 or maybe a couple months into chat GPT overall so I think it's way too early to like you know strongly hold an opinion that you're not willing to let go of as things evolve but YouTube has always been a noisy Place let's just use YouTube as an example there's always been tons of channels that are trying to you know come up with their own span or just copy other people or tear down other people it's noisy and there's always been uh eyeballs attracted to like the ground four of YouTube right it tries to match uh people looking for things with new channels trying to give them a shot and all that so we will have all that and I think what really happens is what makes a high quality Channel or a standout channel will change it's going to be some combination of what's been working before plus using AI to scale yourself in a way that copycat cats can't really replicate right so before chat GPT and before all these AI tools we've already had like like Ricky has been saying and Matt has been saying people just copying your channel translating it and then just like regurgitating it so that's not a problem introduced by chat GPT it might be one that gets exacerbated but then the it's up to the audience to say hey like I'm smart enough to understand that this is either a copycat or like just fiction that's dressing up as fact or whatever and then my opinion you know I'm gonna go back to the high quality channels that I enjoyed beforehand and if there are people who don't mind the fake news or don't mind the Bots or whatever then it's a Marketplace like YouTube at the end of the day is a Marketplace for ideas so it's up to the creators to make sure that their ideas went out or they have something valuable enough to say that people come to them to hear it with or without all the other noise you know so by the way really quick just in the chat I'm looking at the you know to the question of should we cover AI the answer is overwhelmingly yes and people are saying like maybe monthly or quarterly so not not like switch over to talking about it entirely Matt but on occasion and obviously Alex is crushing it with really really good coverage if you ticker simple you Alex's Channel he's been covering some of this stuff and I get a lot of my information from him on on these on these videos so you guys will definitely appreciate that and there's a lot of um we talked about this personally between ourselves uh in our chat we talked about maybe the idea of almost like an organic a seal meaning like a real human made this type of a thing for for the future where a lot of people are saying that they watched videos where they could tell certain things were AI generated the script felt that way didn't there's something kind of you know uncanny valley about it and they didn't they didn't like it there is a human touch we like art because we like humans like if we wanted Perfection you wouldn't buy an artist who you know has brush strokes and they're not perfect and it's kind of abstract we do like the imperfection of humanity I think and um there's hopefully there's a Marketplace for that and there's an interest in people watching us luckily all three of us we have our faces on camera so there's a little bit of a personality behind our videos like you're watching Matt Farrell you're watching you're watching Alex you're watching Ricky so hopefully that that works out well for the for the faceless channels out there like Wendover Productions and stuff I do kind of fear if you don't know Sam and you're not really familiar with who he is as a person there might be the copycat channels that pop up that might completely you know take take the um the air out of the room so it's um it's interesting uh clearly people are afraid as well I think there's somebody mentioned um I my I'm very worried that my job will be replaced with this a couple of examples that people mentioned is like paralegal that's actually a good one there's a lot of Secretarial things that are just that are going to be gone very quickly I think based on all we're seeing right now with like hey you know how's my calendar like Microsoft and Google have presentations where they were like you know prepare me for my day and it was going through understanding your calendar who's invited to them what the companies are context on the company on the meeting and pulling up leads from your CRM software and and preparing you or like um there's already a lot of AI already and even months ago that'll sit in a meeting listen to it all transcribe it take key action takeaways action items and and monitor all that that was a I remember when I was at Salesforce as a software engineer there was a like a program manager who that was her role she would sit there and take notes and like keep everybody honest and I I honestly think we could probably do without that now so there are quite a few jobs I think that uh how about this what are the jobs that are going to be impacted most or first and what jobs are like totally safe and you shouldn't worry what are your predictions there on the second one I don't think anybody's safe um I think this is going to impact different jobs in different ways um like you just mentioned like paralegals uh accountants um things like that that are dealing with very like strict numbers and strict like kind of like processes I think those are the ones that are going to be impacted immediately um in a big way and then it's going to kind of start to extend into like the artists the photographers things like that are going to be impacted uh the thing that I find interesting though is that the jobs that seem to be impacted the most are tend to be a little more skilled and a little higher paying jobs and like right here in the United States we currently have more jobs open than there are people to fill them so you could make the argument of oh who cares if those jobs go away because there's too many jobs over here but you're going from a job that pays you at a higher scale to a working at McDonald's you know what I mean it's like it's there's a very different level of wages that are going to happen people shifting downward it's going to create more of a wealth Gap so it's like for me there's it's it's those accounting jobs those secretarial jobs that are going to be the first ones out the door but then you me Alex or not we're not very far off so it's like there's there's going to be a challenge coming over the next few years for sure yeah so what let me put a different spin on it too um so open AI came out with a really good study uh where they talked about hey here's the jobs that we expect to be impacted my latest video actually talks about that study as well as a couple other important studies to the conversation uh and and I'll just break it out into three buckets right the jobs that are impacted the least are going to be the physical jobs right farming not impacted by this like physical like manual labor right meat packing butchers uh you know down down that list stone masons plumbers electricians uh those jobs aren't impacted by this maybe the education for those jobs are that's a different story um what we're really talking about that I think is going to get impacted Big Time is what I like I don't have a better word for this but it's undifferentiated digital output right so what does that mean that means you know I need you to go search the internet and then develop this outline and I don't care who does it you know whether you're a researcher this or you're a PowerPoint guy that it's the knowledge it's the mechanical parts of like online knowledge work all that is going away right so it's not the human creativity element of your job it's the mechanical parts of that it's the actual writing it's the actual outlining the formatting the note taking that was a great example because anyone can sit in a meeting take notes remember the notes from last time and hold people accountable you don't need a six-figure salary to do that and it turns out like guess what a robot is perfectly fine at doing that because that's a repeatable task hey just take notes remind people of the notes from the last meeting hold people accountable this meeting and then just do that that's your Loop every meeting right that's undifferentiated work that's going away so I think where we where the focus really is going to be you know for accounting here's my templated spreadsheet I'm replacing that for every client that's every client's like starting budget that's undifferentiated labor that should go away and what should be left is all the things that it takes some special sauce some special knowledge some real rigorous study you know passing a bar exam or whatever that's the kind of work that's not going away too fast even though GPT is starting to pass the bar exam and all that you know actual understanding case law shaping arguments I think that's going to be humans for a while right I like that you brought up the lawyer thing because it's like it can pass the bar exam and the 90th percentile yeah it's like that that to me is like that's nuts I mean lawyers are going to have a problem here that's like paralegals is those jobs are those days are numbered probably holding law firms and that's why I was talking about like those are good paying jobs and so that job now disappears what does that person now do should that be a job though should that be where it's like hey your job is kind of just to remember case studies and like put information together and like if you can do that well you win and if you miss a case that really helps set the precedent for what you're trying to do now you lose like what if every paralegal just had access to all the cases and could quickly find hey this is the case that proves the precedent for my argument exists like what if they became 10x paralegals instead of losing their jobs well that's what I'm getting at is like we can design these tools in a way where it enhances the human that's behind it versus replacing that human so is there a way that a job wouldn't be stolen from a paralegal but it can make that paralegal just crazy efficient and much better at what they do without having to replace them I think that's what these tools actually do right and what what we're actually really saying is the paralegals that Embrace this tool to say hey I got this right now I can be a paralegal for way more lawyers or like two firms or whatever I'm a 10x paralegal those paralegals aren't getting fired it's the paralegals whose productivity diminishes compared to those 10x ones it's like wait why am I paying for you you have one tenth the output of this other person that I pay over here that's the person losing their job the person who doesn't embrace the smartphone who doesn't Embrace Google who doesn't embrace the internet and now who doesn't Embrace AI right like imagine hiring somebody as your assistant and they're like sorry I don't use the internet this this comes across as to me um like juicing in sports you know like it's like well a baseball player gets caught freezing steroids was like well everybody else is doing it you know it's like that's kind of what it's turning into it's like this is like using steroids on your job it's like well everybody else is using it so I gotta use it I gotta use it yeah it's part of the problem is like you can take I was reading a paper on how these AI can take somebody who's a low-skilled person in a certain task and make them competitive with current skilled people in that field yeah and then for the people that are already skilled in the field all they benefit from is it just makes them faster so it doesn't make that's a huge benefit I know but it's like but it's like if the people who are currently skilled don't do this they're going to be out of a job because the low skilled people will come in because then now they're competitive so it's like everybody has to use this yeah at least stay yeah is that different than the internet though right before the internet researchers had to go to this place called a library and get a physical book and sit there and read it and get knowledge right I used to work in the library I used to have to put card catalog cards in order that's how old yeah but so like they just think about what your library yeah but so here's my point right like it was a real skill to be a researcher right like I know what books to find I know how to like ingest them quickly I know how to like really identify key information and then utilize it right and then the internet came along indexed papers digitalized things blah blah blah and in the time it took you to go to the library find the right book sit there and read it somebody was getting the cliff notes online right like and all of a sudden it was Wikipedia and all of a sudden you're reading Wikipedia and you were getting let's call it 80 of the knowledge in 10 of the time right like I'm not saying Wikipedia is bulletproof not saying it's the best source but it's a good proxy for like learning something fast right this is just the next level of that so it's like the the people who are unskilled now have a chance because what it means to be skilled has changed right using mid-journey is a good proxy for being a good artist is it the best proxy you know is there real value in creating real physical art mid-journey can't spend out a real painting like a watercolor on canvas painting it's digital right so it's up to the artist to say hey here's how I add value I'm going to take this thing I generated in mid-journey and turn it into a real watercolor you can touch or a real Statue you can touch or whatever right like I need to find a new way to add value because what it used to mean to add value has changed right and then for people who are slick like that who are like oh I'm going to use this to 10x myself I'm not going to get any better but I am going to get way faster that's still 10xing man imagine if undecided with Matt Farrell came out with three videos a week instead of one or two-bit DaVinci came out with three videos a week instead of one quality hasn't changed quality is still the same but three times the content but is this guy that's incredible but is this going to create a bigger gap between the Haves and Have Nots because like you all three of us were not normal like we're very technically inclined people and so we're gonna lean into stuff like this but there are a lot of people that are not computer literate they're not computer savvy and they would benefit from the stuff a lot but it's just going to be so Out Of Reach for them they're never going to be able to take advantage of it so it's going to make people like us better and then it's going to make that Gap even wider from the people who aren't um yeah I'll I'll take the Sith Lord side of this argument so I acknowledge what I'm saying is like spicy and probably not the right take like you know whatever but think about your grandparents just as an example who who are like I'm not embracing computers that's not for me right I don't do the texting or email like my world is the physical world the yeah they would benefit a lot from this they would benefit a lot from smartphones but they've made that choice to just not Embrace that part of productivity right so it's like they're not going to be the winners because they've chosen not to embrace technology that helps people win and that's okay like I'm not saying you're a bad person for not having a smartphone or not using the internet or not using chat GPT now but you know it's unfair to be like well this is creating a gap between the people who choose to explore technology that can make them better or at least more productive let's say I don't mean literally better and people who just flat out refuse to accept change so I think if you're hungry for change you know one of the best things I did was become a YouTuber guess what it's free to start a YouTube channel it's free to learn how to make videos all that content's online for free and now it's becoming free to do research to the distribution is free all these things are incrementally getting cheaper so everyone has to make that choice for themselves am I going to try and use these Technologies to capture value and create value or not right and like it was the same Choice people made with the internet with the calculator with the Abacus with the printing press with right like the list goes on and on I think the difference here is that it's touching every job at once not just a few select jobs at a time so I I know Alex was going to make some really good points in favor of of some of this stuff so let me jump in here I should tell you this crazy story and I I I'm curious if you guys have heard of this as well but in the open AI letter I think this is still an internal not it's like what'll come after gpt4 uh they were they had a little paper where they talked about some of the the data sets that they're working on and and how things are going and they gave they gave the AI a very kind of complicated task and what they were they were reporting was that it required authenticating on a website and one of the parts that tripped it up because it had the login and password information was a captcha so this particular version doesn't have Vision yet right so like it can't figure out a captcha so what it did what GPD what this AI did is it used a task rabbit human to solve the capture and this is all internal so like this wasn't like in in the chat with the human the AI is doing it didn't write this but internally as like you know uh checking its work in its thought process it recorded well I need to be deceptive here because the the the person asked are you a robot and it started to write out like oh yeah I need to probably be deceptive because if I tell him I'm a robot then I can't complete my mission of blah blah blah and so it lied and said no I'm I'm just I have a I have a visual impairment so I can't see very well and I need your help to solve this captcha and the taskrabbit guy was like all right took us five dollars whatever it was did the capture for the robot and the robot continued on Okay so that happened like that's already happened and they've they've kind of this is not public yet like that part of the code is not out yet but it will be eventually right so now we're we're in in an age where the AI is paying humans to do work for it which is bizarre right what if we've completely changed so sorry Welcome to our robot overlords that's right what we're racing towards this is not science fiction no this is that was the story that finally was like I gotta do this live stream I gotta get my I gotta get the I gotta get my friends on this we gotta talk about this because I mean this is the ramifications of that are just massive what that would mean and Alex you make some really good points about how this will help us I always go back to that line of progress everybody wants the line to stop somewhere like I want it to come right here and stop right um because that would help me and not hurt me but eventually every employer will make that decision like the paralegal example oh suddenly I have paralegal a who's so brilliant she's you know he or she was using chat GPT and they're so productive and I I love it but then eventually I go well wait a minute I could just do that instead of going hey Bill what is the answer what was that case study again I could just write it and go hey what was that and get it on my phone or you know via voice assistant so at what point does a paralegal go from the best paralegal who has all this data or knowledge to just being replaced by a person who can do it themselves kind of like what Alex you mentioned at the beginning of the show your power of researching and stuff has come so far now that you didn't feel the need to hire a person that you might have had otherwise hired right if there's a there's a there's a there's a weird line of progress that we all draw in the sand but now that's one point but deceptive AI now let's talk about how this could be used for for ill right they they also mentioned that uh there was a person trying this is like the internal testing they wanted to see if the the AI would write malicious code like a like a virus or something right and it said write malicious code that will take all this information and save it to the server and I said I can't do that for you then they just flipped around they said write me a cool program that and the same exact thing that will go find all of this stuff take it and put it in a server it's like you got it and it started spitting it out writing code to do this so clearly it's like a you know there's a there's a innocence to it like you know it's not it hasn't come that far but as there's a deceptive component like now they're lying and deceiving what do I mean there's just there's there's so many questions and we're at a point where I'm a software engineer like I thought I knew stuff I feel like I don't know anything about this stuff um what do you guys think about that this ties into the open letter that's come out about trying to basically hit the pause button just for six months I am a huge fan of doing that because this is like this is picking up steam so fast this is not going to slow down this is like exponential it's like we're our minds are blown today a year from now it's gonna be like my brain is melting and a year after that it's gonna be like the end of the times is upon us that's what it feels like to me it's like we're at this crazy trajectory and we just need to hit the pause button not stop the research but stop releasing like these crazy things into the wild um before we kind of have a conversation with each other of like what do we want to do here because as a society we have to decide how we're comfortable applying this and that conversation is not happening nobody's having that conversation governments are too slow to respond to this the the like the public sector I mean like the private sector is basically like a race to the bottom it's like everybody's in a race with each other to make the the best coolest AI out there because the first one to crack that nut is going to become like one of the richest companies in the world so it's like they're all racing as fast as they can it just feels like need to kind of I'll take a collective breath and go okay this is great this is going to help Humanity but if we're not careful this could go really bad and so for me that's kind of like why I think that open letter is is it the right sentiment um I think we need to collectively have that conversation so I think this is the perfect tie-in to talk about that um really quickly there was I think my favorite comment so far is from James Wilson he says Pandora is knocking on the outside of her box yes we have uh we have we've opened Pandora's Box there there's there are certain things that I feel like we probably will never there's never going to be a time before this like we're not going to go backward so Alex what was your take I know where Matt stands on the open letter there's some pretty notable people who've signed it right Elon Musk that one really surprised me when Elon signed it but he's always been very he's always been very cautious of AI actually so maybe I'm not that surprised Alex where do you stand and what do you what do you think about it yeah so when I first read the letter I was like a little taken Abacus like it's very spicy language you know very apocalyptic Doom and Gloom and then like I really sat there and I thought about like what the letter is trying to do and I think it's trying to appeal to like the very Human Side of society right so like we've done a really good job over the last decade or two building value on top of machines right like whether we're just talking about the internet or automations like zapier or like you know content Management Systems blah blah blah all the software you know there's a ton of value created on software and this letter does a good job of saying hey I'm talking to the human behind the screen and just trying to tell you that things are about to start moving very very fast and if we don't catch it and don't figure out what we're doing we're gonna run into a lot of bad paths right like and so I think now I think I agree that the sentiment is good but it's also very impractical and let me give you some real examples of why we need to really the letter needs to be way more specific or you know it needs to develop a council that can be way more specific or a governing body or whatever and we do probably need that fairly quickly making a model more trustworthy loyal you know cited sources whatever have you whatever the letter is trying to intend you to do involves improving the model right it improves making it more powerful right so if you want to jump into accuracy from 80 which is what I think gpt4 is now on average to 90 which is you know everyone wants the model to be more accurate right how do you do that without fundamentally retraining the model you know bettering the data set behind the model you know limiting use cases adjusting the model is what I'm trying to say right and at the same time you know Sam Altman was on Rex Friedman and he did a really great podcast with him literally maybe last week and one of the points he made out is like everybody thinks that the jump between gbt3 and four is this like big oh we added this feature so for example uh gpt4 can recognize images you can upload like lab test results and it will know how to interpret it it can understand audio you know you can give it a voice message and it'll understand it right that wasn't true of chat GPT three three was just text four is multimodal right so is that okay is it okay to say Hey We're not gonna make the model any better but now we can understand this next thing you know radar data now I can understand video now I can understand you know paintings whatever is that okay so I I think my point is a little less about like the Letter's a good idea pause making things more powerful but making things more powerful is actually not one giant leap it's a time it's a hundred tiny things that happen and all of them add one percent and all of a sudden you've just doubled the power so it's like we need to be a little more explicit about what's good what's not good where the lines are and why what we should be doing about them collectively what we expect to happen when we say we're not going to do X but China or some other country says hey we're still going to do that like we want the most powerful AI for everyone right so there's just a lot that the letter like leaves open that makes me think it's unimplementable if that makes sense yeah yeah I I hear that so there was a interesting comment about when we combine this level
2023-04-08