The AI Model Maze

0:20 Hey, welcome to the AI Zeitgeist. I'm Austin and I'm Luke and we are kicking off the second

0:28 episode. Super excited and thanks for sticking around. This is gonna be a good one, I promise.

0:34 We wanted to kick it off first by talking about like just the different models that

0:39 are out there because you know it's one thing to like try to figure out how to use AI but

0:43 then you also have to figure out like which one you should be using. Yeah, because there's

0:48 like a billion of them. Exactly. It's ever-growing. Well, so there's like three big ones and then

0:54 a couple that are like kind of honorable mentions. So the big ones are Claude, Gemini, ChatGBT.

1:00 Those are kind of and right now Claude and ChatGBT are kind of at the frontier with Gemini

1:05 a little bit further behind and then there's also Grok. Grok is in their mix too but that's

1:10 kind of the whole shebang that you have to think about and I'll let you in on a little

1:16 secret, it really doesn't matter. That's really the thing. You kind of have to play around

1:23 with, you can just pick one and do the free version or pay like the 20 buck a month tier

1:34 and just mess with it and do projects but you can also, you can try using the free version

1:40 of all of them and ask the same prompt to each of them and kind of see which one you

1:45 feel like you jive the best with because they all kind of sort of have their own personality.

1:50 Yeah and the thing is that I would say probably the most intuitive one to use that's like

1:56 very straightforward and it's like you don't necessarily have to learn like a different

2:01 skill per se like is probably ChatGBT because it's like super straightforward and simple

2:06 like you can do pretty much everything in chat. You can have it code an app like inside of the

2:11 chat. Gemini, it's a little bit harder for do that. Clod, it kind of routes you towards codex

2:18 or no Clod code. Clod code. Yeah exactly but so it's like you know they're all there, they're

2:23 little nuances but honestly like I would say if you're trying to get into AI and you are just

2:29 now starting out, pick one of them and learn it and if you have the money to invest, spend $20 a

2:38 month just to make sure you don't get you know limited out, make sure you can kind of keep using

2:43 the frontier features but then you know just go for it and just you know it's as you learn one

2:48 of them you'll get to understand why you should use one here and the other one there. It's kind

2:55 of a thing you learn over time. Yeah and as you as you have more experience and stuff you can if

3:01 you're getting to a point where you don't feel like it's accomplishing what you want it to then

3:07 maybe try to switch over to a different one and then you'll see like oh okay well if I'm doing

3:11 this type of thing Clod's gonna be better or vice versa. Totally and the example that I ran into this

3:19 is I had an audio recording that I needed a transcript of and I dropped it into ChatGPT

3:25 and I asked for the transcript and it was really struggling and it was I had it on 5.6

3:30 soul high thinking so it should have been that's like the the frontier right now for the for

3:36 ChatGPT. It should have handled it you know but it couldn't it wasn't it wasn't able to do that but

3:42 drop the same audio recording into Gemini and because it's a little bit more

3:47 multimodal which just means it does audio video and text like natively it was able to like

3:53 instantly pull that transcript out of the audio and give it right to me and then I looked it over

3:59 and I was listening to the audio and it was spot on. So some models just do certain tasks better

4:06 and if you find the task is not working in one model it doesn't hurt to try the other one. Yeah

4:15 another thing with talking about spending money if you're going to spend money is something that

4:23 you might not realize is the context window. So every time you send a chat to like let's say to

4:29 Gemini you start the conversation by saying like hey I'm trying to work on this project

4:34 give me some advice so then it replies and then you send a thing like oh okay yeah let's go with

4:41 this route so when you send that it sends the beginning of the chat and Gemini's response

4:48 back so like by the time you finish your conversation in this one single chat it's

4:54 sending the entirety of your conversation every single time you send a new message and so that

5:01 builds up context the context window and that costs more tokens and so what I like to do to minimize

5:08 the kind of the cost of the context is I'll work on one feature in one chat and then once I get

5:15 that kind of like working good if I realize I'm going to something entirely different like a

5:20 different feature I'll start a totally new chat and so then your context window is wiped clean

5:26 and if you're using cursor or codex or whatever it looks at your code base to begin with so it's

5:34 you're working within the same project or folder it's going to know it's going to look at your code

5:40 when you start a new chat anyways and so you just you're just adding to this project so this is like

5:46 a fascinating thing like to think about because I think when you think about like when you when

5:51 you jump into chat GPT or Gemini and you're having a conversation back and forth it feels

5:56 like you're having a conversation right and we the way that humans experience a conversation is like

6:02 like I say one thing and then you say one thing and then I say one thing that's not how the LLM

6:09 experiences it every single time it hits the model it's hitting the model completely fresh

6:15 like it's it's waking the model up the model is thinking it's giving a response the model

6:20 shuts all the way down and forgets everything about the conversation right and the reason it

6:25 can have continuity across the messages is because it sends the previous messages with your

6:32 right message that you just sent and all of its responses to you so like that is the actual

6:38 mechanism that is the reason why all the messages have to go in one which it isn't intuitive to us

6:45 like no exactly that way right I remember when I when I first started working with the

6:51 the API for chat GPT I didn't realize that like like that's something you have to build into the

6:58 code is for it to send the whole context window because otherwise it's like you're talking to a

7:02 person at a party and each time you go back and forth you have to hand the other person a piece

7:06 of paper of everything that's previously been said yes conversation but it's like it's like

7:11 dory you know like is that the oh right yeah yeah yeah memory gets your memory gets wiped every like

7:16 three seconds right that's what it's like that's what it is it's like you're yeah it's like the

7:20 freaking you know uh in in uh oh yeah yeah men in black yeah the I forgot what that was called

7:26 the flasher thing exactly exactly it wipes it's like nebulizer yeah that's right that's right

7:34 exactly yeah um but yeah so it's important to like to not like to think about

7:42 you have to be very strategic when you're using AI to keep the cost down well it's so

7:48 yeah that's one part of it but the other part of it too is it's also a helpful technique to help it

7:55 be able to continue to reason well if it has too much context even if it like if you had all the

8:02 money in the world you still would want to limit your context window right yeah because if you're

8:08 if you're figuring out in the process of creating something with AI and you figured out like

8:15 halfway through oh this isn't really working out how I want it to the model is still getting all

8:21 of the previous bad thought process and so that can influence the outcome yeah if you take a left

8:30 hook and you are like realize that you need to go a different direction the previous context where

8:37 you were going the initial direction is like sticking and it's like it's like yeah it's like

8:45 there's like a lag and it's hard for the model to take a left turn like it's like giving the model

8:50 like inertia in that direction and if you try to change and pivot it's gonna resist that pivot

8:57 because the inertia is built in so the best way to do a pivot is like once you realize it's happening

9:03 like try to get the information out of the chat that's important you can ask for like a summary

9:10 you can have it summarized yeah exactly yeah pull it and then you start a new chat and say hey

9:15 because now your summary is just one turn and then you say hey we were going this way but now

9:19 we're going this way and then as you build like it becomes less and less reliant on that summary

9:25 and it's focused more on the new direction so there's just some practical ways you can think

9:29 about it but yeah the context thing is huge yeah but that's what you just said isn't as important

9:35 if you're using like github or something because it sees your whole code so you can just you can

9:39 you can start fresh because it's gonna see well what everything is right now 100 and that's the

9:45 thing is that as you build out systems whether it's something that someone's already been set

9:50 up for you or you're starting to build something out like i like to think of it like this sometimes

9:55 like i'll have even if you're not coding if you just have like a word document like or pdf save

10:01 to your device that you use your phone or your computer that details out a bunch of information

10:07 about you like some basic stuff like you can drop that into a new chat like let's say you're working

10:12 on a specific project it's okay so you have like a project brief like you could have that saved

10:17 and drop that into a new chat like if you're not coding on github like that's another way to do it

10:20 because that context that's like the base level context you want that to be like the in the chat

10:27 and then you can start layering on top of it right i want to differentiate because we're using the

10:33 word context in kind of like two different variations so like there's the context window

10:38 which is your chat history and then there's model context and so that's like information about your

10:45 or just anything the model would need to kind of answer like more intelligently what you're

10:55 specifically asking asking so there's just two different it can be a little confusing

11:00 yeah that's a good point so okay let's break that down a little bit so there's they're they're

11:05 technically measuring the same things like it's the same bucket like your context window also has

11:10 to fit the model context right right but like when we talk about context that you need to give

11:16 the model like that foundational layer like i was just talking about like that is the information

11:21 that the model needs in order to start working on what you're asking it to work on right outside

11:28 there's what you've been saying back and forth it's exactly thing and then the same thing the

11:32 things that you're saying back and forth also have to fit into that overall context window

11:38 and that is also something you have to think about too and these models have different

11:43 limits like i don't know exactly where they are i know jim and i was like out in the lead

11:47 had a million token context window i think the other ones are more or less caught up if they're

11:51 not there all the way they're like close like i think it's like 500 000 it's pretty typical

11:56 um some of them have a million as well so but it's like you have a lot of room but the other

12:02 to think about with the context window is it's like ram on a computer you don't want to burn

12:09 it like right to the top like you want to give it head room so it can actually think because

12:13 it needs like some right some room to actually process and you don't like we talked about before

12:18 you don't have irrelevant things that it has to think about because not only are you spending

12:23 tokens on sending the context but now it's using more tokens to process all the information that

12:30 doesn't need to anymore well and then there's like okay so we're going a little deep on this

12:34 but i think it's good then like there's another layer of this too is that they're the way that

12:39 the models are trained is they're trained on a bunch of documents and if you think about like

12:45 most documents have like an introduction and like maybe an abstract and then they have like

12:50 a conclusion and like maybe a findings so like if you think about that that means that the model

12:57 is getting a lot of its information from the first part of the document and the last part of

13:02 the document so that means that it's really focused on if you think about even in a context

13:08 thread that's probably a good thing like you want it to focus on your foundational stuff that you

13:13 initially said and then what did you say last turn what was the last thing you said in the last turn

13:17 that's that's a good good pattern to have but what that means is that you have lost in the middle and

13:22 that people say and it's basically means that your context that's in a long context a large

13:29 context window the context is kind of in the middle it gets a little bit blurry they i can't

13:34 see it as crisply and that's another reason why you want to make sure that you're thinking about

13:40 resetting your context window because if you let your context window go too long

13:45 it can get poisoned a poison context window and there could be two reasons it could be like

13:51 you pivoted and you took a left hook and now you're doing a different thing like that's going

13:55 to be hard for the ai to follow the other side of it is like if you just have a long context window

14:00 in general it may start to degrade because it can't remember all the fine details right right

14:07 all that said basically just always restart a new conversation is like the only thing you really need

14:14 to remember is like if you if you ever get in the habit of saving a project or a word document or

14:21 whatever get in the habit of restarting a conversation with the ai yeah i mean the thing

14:26 is like sometimes i'll be like chatting back and forth and i'll be like oh this one's kind of

14:30 getting a little long and like right when you when you get that feeling reset your context well you

14:34 you might even notice it's slowing down that happens i had that i had that a lot i don't know

14:39 if it happens as much anymore but like i had that a lot with um chat gpt i'd have these like long

14:44 conversations and this was before codex and stuff so i'd have like i'd be making i'd be having to

14:50 code and so it's sending all that stuff back and forth and eventually it would get to like

14:55 like almost it would be at like a standstill and it wouldn't do anything else and that was

15:00 before i realized that it was sending everything back and forth each time but yeah so okay one

15:08 thing that i wanted to talk about is this week i was at we talked about this last week i play i

15:16 think i i was watching our league director at our league so everyone signs in and then they sign

15:25 in on the app the app is like the the app that everyone uses to to score on the course but it's

15:32 not necessarily how they do everything else so like you when you're a club member you have a

15:36 bag tag and then you want to be in your division and like you there's three people per hole they

15:43 you want to stagger the whole start so that there's like you know multiple and everyone has

15:47 to lots of like things that go into like setting up and what they were doing is they just pinned

15:54 a paper they were like okay everyone signed in i got them all let's split them into the different

15:59 cards that they're going to start on we're going to shotgun start try to get this done in two hours

16:02 before it gets dark and then when people come back they have their bag tags based on how well

16:10 they shot like the person who shot the best gets the lowest number and so you okay the tags have

16:16 to be redistributed back out right it's very they're sitting there like looking at the it's

16:22 getting dark out they have flashlights they're trying to do all the math on the paper and figure

16:25 it all out right and i was like this is an app right yeah this is software right right and before

16:33 it was like that would cost you like you know two three grand to get somebody to build an app for

16:38 you that would do all those things unless you knew how to program right which you don't need to

16:43 know anymore exactly so i'm sitting there and i basically built out the application got it first

16:50 pass did a couple of iterations and i have an application that i showed the league director

16:56 and he was like oh my gosh why where is this i did this yesterday right so why are you just now

17:02 doing this exactly but the funny thing is like that's that's like uh i think gonna be a more

17:07 common pattern is that people are going to start using it for non-monetary purposes like i didn't

17:14 get any money out of it and i made it to make people's lives better and it is going to make

17:19 people's lives better but it's not something that's tracked in gdp for instance you know

17:26 like so it's just it's just a weird concept

17:29 to think about you know but yeah i i think the the way to think about like where we're going

17:37 with this is that it can be easier to think about creating value than it can be to think about

17:44 capturing the value yeah yeah you put yeah you get more it's unlikely that you're going to make money

17:51 off of it but it's very likely that you can you can use something and make something that's going

17:58 to make your life way easier that's probably the best way to think about it i mean there are ways

18:03 to make money off of ai but pretty much everyone's chasing them right it's gonna be arbitraged away

18:09 like super quick exactly i feel like the best way to think about like moving to this next era

18:16 is think about figuring out how to create value and eventually there will be a mechanism for you

18:22 understand like how to create the value in a way that also compensates you but that's not where you

18:29 should start right well and and right now since everything's so fresh you can like earn clout

18:36 amongst people who don't know how to use ai and you're like oh yeah i built this thing

18:40 and people are like oh my gosh that's so you're an amazing programmer right

18:46 well you know i know how to type in a chat window well okay so i was watching a youtube video today

18:52 and it was a programmer who had been in the industry for like 15 years and they were like

19:00 swearing off using ai to program they're like i'm just gonna go back to programming myself

19:06 and they were saying a lot of things that i agreed with and they were also saying a couple

19:10 things i disagreed with and the reason is i don't have a programmer background they were like i'm

19:17 right i'm missing the fact that i can understand the code and i'm getting in like i don't have a

19:21 doing the thing that i used to have a lot of enjoyment from and i'm like that's great like i

19:28 if you know how to code go for it like but but that that ship is sailing exactly like that's not

19:35 exactly every everybody even i've got a friend who works for like one of the big banks as a

19:41 programmer and he's like everyone uses ai to program now yeah and and i think the the the

19:48 is like it's not necessarily important

19:53 like he he was talking about it like i wish that like he was like i i'm trying to do this

20:00 thing so i understand what i'm doing so i'm going to do it just my way and do it slower

20:04 and like that's fine you get a good understanding of what you're doing but i think the the next

20:10 level like and he was saying like i'm not thinking anymore i'm just like managing these

20:14 agents to code and like they're coding and then i like see what their outputs are

20:18 but to me like i think the level that you have to be at is you have to start thinking about

20:24 how do you manage the work that's being done so you like have to go up the abstraction stack a

20:30 little bit start thinking about like i'm gonna do this thing and then i'm gonna start now that

20:40 i've bought some of my time back i'm gonna start thinking about the next thing i want to do

20:44 right yeah yeah and um you're more you're still driving the project and you're still like like

20:53 all the stuff that i've made still would have happened if i hadn't thought of them and like

21:00 there's there's features that i think of and i'm like oh well i can just instead of spending the

21:04 time painfully trying to get it to work i can just have an agent go off and do it while i can

21:13 like spin up another agent to do another thing that i have an idea for i mean that's exactly it

21:18 and you don't necessarily need to understand how each line of code works as long as you know

21:24 what you've built and what the features are and how they're supposed like how things are

21:28 like you understand the architecture of your app right and that that was like when i first started

21:35 using ai to do this stuff i started getting into things that i like i was able to do a lot more

21:42 like a lot more depth of things than i could before and at first i was like oh man but i'm

21:48 not going to understand how it's working like how am i going to fix something if it breaks

21:52 but then you just get the ai to fix the thing that breaks like just like just go full force

21:58 into it and just know that like okay well i'm going to get to a point where i don't know what

22:03 it's doing but i can just say i know what's broken and i know how to say what's broken totally and i

22:09 can say fix it and the other thing too is that like you can ask it to like say like explain to

22:13 me more or less what's happening yeah and it can give you like an explanation that you can actually

22:18 start to understand what the things are and obviously it's not like you can replicate 20

22:25 years of experience as a coder so you're not going to know all those things but it gives you the

22:30 ability to start to understand like more or less what's going on so you can do it okay well and you

22:36 hold on a second before you move on and you learn you learn more about the overall process of how

22:43 work just because just with chatting with the ai you're like oh i need to do this and i need to do

22:49 that stuff that you wouldn't have any idea about before and so i think i think you still learn

22:55 more about coding absolutely absolutely yeah i mean i think that like if you'd ask me like three

23:01 years ago a bunch of technical terms and what they meant i would have looked at you like you're

23:11 what's a front end what's a back end like what's an api like what's you know why do you need a

23:16 database here like what's what storage are you talking about cold hot you know whatever all

23:21 these terms like i don't have no and maybe you're listening you don't know what that is either

23:25 that's totally fine like right the point is as you use ai to do things and to build for you

23:33 all that stuff kind of just osmosis and you start to understand like what things are doing

23:38 things just make sense to you and then you like have a pattern yeah and you're like next time you

23:43 go for the app you're like oh i now i know that i need to have i need to do a smoke test and now i

23:49 need to do i need to do an end-to-end test at the end to make sure it's all working right so like

23:54 and so you you're like you're building the vocabulary on yeah you're building the vocabulary

24:00 but you're also like building upon like you're using ai to do something but like each time you

24:05 use the ai to do something you're going to learn something so that you can get it to do even better

24:08 things and so it's like this like stack of constant so the the the the learning is stacking

24:15 but then right you're also stacking the tools that you've built right like sometimes i'll use a tool

24:23 that i've made for one thing to do another thing and then like i'll or i'll build a tool that like

24:30 gives me overall efficiency so i can and it's like and then you you're staying on that platform

24:36 and you build another tool you go to the next level and it's like each project you do is giving

24:42 you learning right you're also getting the tools that are actually accelerating you too

24:48 right and it's it's super cool like the the grocery store sound thing that i have it it

24:55 connects to wi-fi because it's just it's a raspberry pi so it's an actual like linux computer

24:59 um it connects to wi-fi and so it's like opened up so many more like i suddenly have this box

25:07 that is connected to a network that i can have anywhere in the store and then i can see stuff

25:12 going on on it on my laptop and so like sometimes you have to check the continuity of a cable

25:17 and you have these um it's like tone generator you connect to the wires and a light comes on

25:24 if if there's a short in the wire or not and so there's um at the registers front registers they

25:31 a lot of times have uh bells where they press a button and it does this chime over the loudspeaker

25:36 if they need to get their manager's attention or something and that's hard to test because

25:42 sometimes where that chime unit is is like way in the back of the store upstairs and so i'll have

25:48 to like put my my camera on my phone in a place so it can see the device and i go downstairs and

25:55 i press the button and i go back upstairs and i re-watch the video to see if it had continuity

25:59 or not and so now i'm like oh man i can just have my device connect to it and i can see it

26:05 on my laptop as soon as i press the button downstairs like so it just like opens up and

26:11 and also there's not the barrier of not knowing how to set it up because i'm like oh i have this

26:18 idea and i can just ask ai to do it and it will help me walk through it and do it all for me

26:24 that's the thing if you have an idea and you're like i wonder if it can put it into chat

26:31 it probably probably can even just like and you'll so a lot of times too

26:38 it will i'm driving the project but then it will like do things that i'm like oh i didn't even

26:45 think about doing it that way or i wouldn't have thought of doing it that way and it's like better

26:49 and so it's it's like the perfect teammate yeah the code the co-thinker which is the scary part

26:56 because we're all co-thinking with it right yeah exactly hi exactly the zeitgeist not to be too

27:05 yes right okay so this is the question that i had that i wanted to ask you i was thinking about it

27:11 like so you started doing ai like what three-ish years ago like okay and for me it's like closer

27:19 to like two maybe two and a half years and i was just thinking about like how crazy

27:25 some of the things i'm doing right now i would be like completely shocked that i was even able to

27:32 remotely think about half of what i'm doing like two and a half years ago i just wanted to get

27:37 your take on that too um yeah that yeah i would agree with that um i had so i've had a really

27:49 interesting journey in like my career sure i used to be a very creative person like i did video

27:55 production graphic design web design like that was my identity was a creative person and then

28:02 very early on definitely within like the first year of using ai i was like i need to get a

28:11 different job like this is not sustainable doing anything creative isn't sustainable anymore even

28:16 though it wasn't like i read forums where people would be like um animators would be like oh it's

28:23 never gonna be able to animate like that you really need a human to do that and my whole thing

28:28 has always been like yeah not yet but maybe next week like and and now it does it generates

28:35 flawless looking video now and like stuff that you can like barely

28:40 like i don't know if you if anyone listening has heard of the uncanny valley

28:43 that's where like the more realistic something gets like um computer animation like the closer

28:50 it gets to being like a human the more we can tell that it is not human it's like a cartoon

28:56 character we don't get creeped out by it because it's such it's such it's a cartoon it's not the

29:04 actual like it doesn't it's not trying to look real but then if you see these robots that have

29:09 faces like there's just something we inherently feel is off about it and it creeps us out and

29:15 that's called the uncanny valley like like have you ever been to a wax museum and that's super

29:20 right yeah exactly exactly it's like or the polar express right

29:26 um that's that's like the number one uncanny valley example is polar express but like but

29:32 now the video that's being produced by ai looks it's like far beyond the uncanny valley like it

29:39 looks indistinguishable the only thing it's it lacks right now is kind of how people talk in

29:45 video or they do like some like weird things but like next week it's gonna be exactly you know

29:53 you're not gonna be able to tell well and so that's that's kind of like i i used to feel a

29:58 lot more anxious when it comes to like you know what i'm doing for a living like what what is it

30:04 where where should i deploy myself like what's the next thing you know because this is a major

30:08 disruption obviously and i used to feel a lot more anxious about it but i recently feel a little bit

30:14 more like sure of myself and i think this is the reason why it's it's not just because i'm starting

30:21 to learn the ai and i feel more comfortable in it it's also the fact that i think the the most

30:27 important thing is to to be engaged in doing things like in the real world and then be thinking about

30:36 okay how can i use ai to do this like maybe you're making an app for your local disc golf league or

30:43 you're building a tool that's going to help you understand where at your work the the audio is is

30:49 happening in the in the grocery store right you know you're if you are learning the ai and then

30:55 learning how to use it and the function that you're in that's all you have to do and then just

31:01 keep being flexible and seeing where things are moving and you may have to pivot a few times but

31:07 if you are learning the ai and then applying the ai like that's what you have to do yeah i think my

31:15 transition from being a creative person was that like i saw this wasn't sustainable and i was going

31:23 to be able to do anything i could already do that i spent years like so that was i was like a little

31:28 bit sulky because i had spent years i'd been doing this stuff since i was a teenager and like it like

31:34 i said it was my identity and i had built skill and so like i could do things that other people

31:39 couldn't do and then like now people were just doing stuff just by like chatting with a bot

31:47 and i'm like ah that kind of hurts but like so now i think my transition from kind of like

31:54 being against it is like now i've decided to lean into it and i'm like how can i

32:00 how can i do everything with ai in some aspect totally and i the the other like

32:07 it's not just the creatives too because pretty much anybody that does anything that touches

32:14 audio video text is affected so like for me as an economist like everything that i do

32:22 is text-based or maybe it's that it's numbers but you know it's it's it's it can be translated

32:28 into a markdown file so i'm screwed right so basically yeah basically the thing that i've

32:35 figured out is i can either stress out or i can figure out how to use the ai and i can

32:40 build tools and i can start to leverage it in my workflow because there needs to be a a human

32:47 doing the thing that's managing the ai like for the most part i think that's probably pretty

32:53 durable right now that's but maybe not next week maybe not next week but i think it's pretty

32:57 for a lot longer than it will be other things yeah well you you said um you talked about sam

33:05 altman okay and how he kind of like backstepped yeah a statement that he made yeah so yes so if

33:12 you if you don't know sam altman is the ceo of open ai which is chat gpt just so you know and

33:19 chat yes so back when chat gpt uh four was coming out he was like ai is going to replace everything

33:26 in a couple years like it's jobs are going to be gone but then he's walked it back and what's

33:33 funny is it's not because the model has not gotten to where he thought it was going to be

33:40 like he's like it's actually better than i thought it was going to be when i made this prediction

33:44 the reason that there's not going to be this jobpocalypse that he thought was going to happen

33:49 is because humans are just slower to change and adopt the technology so like the technology out

33:56 there is probably already good enough to take your job right it's basically what what what sam

34:02 is saying right but it's not and it's not going to in a rapid fashion right because it takes time

34:10 for companies to implement like all the things and for people to learn how to use it and all

34:15 stuff so that being said there is a moment like a window where if you get good with the tools and

34:22 you start to learn it like you're going to be ahead of the transition and when the transition

34:26 happens you're going to be able to catch that wave and you'll be fine well i also think like

34:31 in relation to mass job loss or whatever is for the most part people like people yeah and so like

34:41 maybe like the bottom line says that like we should replace all employees with ai but then like

34:48 the people managers not all managers some managers are just bad but like they're like well no we want

34:55 we want to invest in these people's lives and give them a livelihood and all that stuff but i also

35:00 think like you know when you go to the store when you're when you're talking to a professional you

35:05 know lawyer this that the other like sure you could hire a chat bot to do it for you and it

35:10 probably would do it better and cheaper but you also there's a premium that goes to having the

35:16 human element exactly so that's that's that's a durable like layer it it will be interesting too

35:22 if that like becomes a really big i don't know if commodity is the right word like is that is

35:30 that like that's like a scarcity that draws the value right yeah it would be it would be scarce

35:35 a scarce thing that people would be like oh well i'm definitely going to pay more to have a real

35:39 person do this and that the funny thing is is that the real person is probably going to do more

35:45 like interaction with you than they would have before because all the drudgery that that was

35:51 tying them down before is handed off to the ai so now they can completely focus on the

35:57 impersonal interaction well yeah i don't think anything's going to be a hundred percent human

36:03 anymore because humans are going to be using ai to take to get rid of the mundane things but it's

36:08 might free up more creativity it's not going to be a hundred percent ai but it's going to feel a

36:13 hundred percent or a hundred percent human it's going to feel a hundred percent human because

36:16 you're going to have way more human interaction than you ever had before because they're all

36:21 freed up to actually talk to right right right right right so yeah i think my perspective on ai

36:27 is it's a tool and you should lean into it and you should use it as much as you can

36:34 but use it as a tool and you will grow when you like when you start using a new tool and it makes

36:41 your job easier you get better at your job and you produce better things whatever that is and

36:47 then i would add to that that it's a tool for two main purposes the first tool is a tool to help you

36:53 learn things and the second way you can use it as a tool is a tool to help you build tools

36:59 so if you use those two ways like you're you're gonna just stair step your way up

37:06 additional capabilities right yeah all right that's been our episode for this week

37:14 and we will catch you guys next week all right yeah see you guys