AI just replaced us with Devin... seriously? Dr Chuck!

we're going to get tired of dunking on AI where someone like you or someone like me will listen to some foolish statement whether it's Devin or the CEO of Nvidia we're like I'm going to spend a couple hours so I can dunk on this knucklehead right these people you know they're wrong you know they're wrong and I just don't think it's worth my time or your time to dunk on every knucklehead that's going to show up and say something outrageous so I think it's more important to to identify when it is that these people are saying something outrageous and it's just outrageous and teach everybody to detect outrageous statements so let's look at some of the statements let's look at the Nvidia CEO statement here's a multiple choice question why did the Nvidia CEO say that programmers are obsolete a because programmers are obsolete B because he mistakenly believes programmers are obsolete C because his stock crisis way hyperinflated and by by saying something crazy and telling people that buying CPUs instead of humans will get his stock price even higher so which of those two things is highly likely and the answer is C I like C yeah we all like C we you and I see this stuff like Tech Bros are always saying stuff now let's take the Devon thing here's another multiple choice clip the Devon announcement why was it because people are widely using Devon to write new software and replace people a uh B uh because Devon has gone and written a million new things be without even being asked to write a new thing or because that person is in between Venture Capital rounds and needs to say something outrageous to their Venture capitalists so that they get their next venture capital round and the answer is see right and so we have to just realize and the Devon announcement was such a perfect Tech bro pitch to venture capitalist that was total crap and fluff I mean absolute garbage and and so all these Tech Bros they want to make some money they want to fund their next thing they want to up their stock price they something they say something that is completely outrageous that they know is absolutely not true they do a demo of something that they know doesn't scale and yet then foolish people flow money to them and and and so I think that the idea of dunking on these folks by giving a counter example it's just a waste of time because it's like whack-a-mole every person who wants more venture capital is going to say something outrageous and they're just going to keep doing it hype cycle right it's just a hype cycle and it's just it's just the thing where I I get it they want the money they just every time they pop up and there's a bunch of them but I just want your viewers to start thinking about like why are these people saying what they're saying why are these human beings yeah who are CEOs or running startups why are they saying these things don't just assume that they're there for truth right they're not truth they're not they're motivation has nothing to do with informing you their motivation has to do with their own company [Music] interests everyone David Bumble back with Dr Chuck Dr Chuck it's fantastic to have you back on the channel it's good to be back so Dr Chuck we were talking offline very recently Devon got released new AI software engineer you and I were talking previously about your trip to India and you I think this is perfect timing perhaps you can tell us about you know the sky falling people are worried end of software engineering jobs please give us your take and your advice so I was I was in a world Whirlwind tour of India for 10 days six universities three companies and I was talking about you need to be better programmers and we talked about master programmer in previous and previous sessions and I was the course is done and I'm like you need to be a master programmer and then the inevitable question that came up at the end of the talk and sometimes these are rooms of 2,000 people in them the inevitable question that came up was like wait why are you telling me to be a better programmer when I don't even we don't we're being told right now that we're it's pointless to be a programmer and so the thing that was driving them a week and a half ago was the quote from the envidia CEO that basically said programmers were Obsolete and I knew this question was coming so I had six times more like 10 times to answer it and so I I I was able to sort of like take a large audience and give one answer and then in the next 24 hours give a different answer and then watch what answers like got the right reaction from the crowd and by the end of it I think I came up with an answer that sort of stunned them like really kind of hit home and the first few answers are things you and I have talked about before like how you know a mid-tier programmer know can it'll come back with mid-tier programmer it has no awareness of its mistakes you know I've been noticing my students like I'm a savant helping my students on their homework assignments because I've seen everything once and what I'm starting to see things I've never seen before and it's kind of weird because like I can usually snap and see what's wrong with it and this one woman who was a really good programmer and her solution to the actual assignment was one of the best I'd seen um but she had something like deep in a configuration file later took me forever to find it a deep in a configuration file and I'm like where did this come from and so you start imagining that people who are sort of depending on these Technologies are going to bake into their code base something wrong and that's going to go deeper and deeper and deeper and deeper and then later we're going to have to unwind it some somebody who knows what's going on is going to have to like wait wait that's slow or that didn't work or it blows up once in five times or whatever and someone's going to have to hunt that down right uh and the initial programmer you know wasn't just a beginning program so I talked about that that was one of my sets of questions right about these little flaws that that come out of it but then I went to uh one University the lovely professional University and they took me on a tour of their Aerospace area and they showed me some drone work that they're doing and the thing that was going through my head was like are we still working on drones are we still like doing drone research I mean you just go down to the store and say give me a drone and and these guys were doing research drones and it was their whole College thing and and and I'm thinking to myself and I went back to the talk and and I thought now couldn't we ask chat GPT or some AI to kind of just build a drone and then I came up with the following idea and that is I don't know if you're old enough David but I remember when we started seeing kind of like such and so University in Zurich is trying to build these quadcopter things that use Wi-Fi and and inertial things and and they're in these giant padded rooms and they would go up and crash and explode and you know of course now we just go buy a drone right but there was a time there was a time where the concept of this quadcopter with inertial checking and Wi-Fi and all this stuff and collision avoidance and a camera and all that stuff that was just crazy futuristic research yeah and so this is this is what I said I said let's just imagine for once that we can take all these wonderful Nvidia CPUs as the Nvidia guy says is the solution to everything take that bunch of Nvidia CPUs right take a bunch of computers take a whatever and we pack them into like a couple of rolling suitcases and we get in a time travel machine and we go back to 2000 and then what we did the internet existed in 2000 and we turn on the Nvidia processors and we run llama for a while or whatever and we sit there in 2000 and we let r give everything that Humanity knows to llama in 2000 and it runs for a while the lights dim and all those things it runs for a while and llama says okay I'm ready and then in 2000 you say I'd like to build a drone that flies automatically and has an you know that's got a joystick L's going to say drone what's a drone but if you do it that today it'll be like oh and then there's a neral things and then there's this and there's that and and so the fallacy of all of it has to do with there are certain things that are being done faster and faster and faster and these Nvidia CPUs definitely do a thing faster but they just am always kind of say that somehow the creativity is growing and the answer is the creativity is like zero and like it's gone up 50% and it's still zero right that there just is no futur looking thing and so I'm standing there in a lab talking to a student who's struggling with something in the future and there's just nothing that AI can help with whatever the problem is AI is a concept that comes from 1947 which is the first time I saw this concept it's a pro it's a thing called the memic which a fell named vanav our Bush immediately after World War II they had done a lot of sort of knowledge mapping and finding ways to use knowledge uh to their advantage and the thing that they the the MX is a fictitious device that basically had to do with you you go into a library and you put everything on microfilm and then you have a machine that you can watch people looking things up in microfilm and what you do is you track where all the people go you make connections between documents they call them threads so that they'd have all these little threads of who read this one then who read that one and then who read the other thing etc etc etc making connections between documents and capturing then the associations we're making in our minds as we're processing information and then gaining Knowledge from those associations right and so the the machine did not write the Books the machine did not read the books the machine watched us reading the books and intended to get knowledge out of that and you see that same thing from 1947 to present of you know the evolution of search engines the first search engines they read the whole internet and they stuck it in a big database and then they looked at all the words and tried to make sense of those words in like Alta Vista and they were terrible and then Google came out with page Rank and said the words aren't so important but the links are and that's called page Rank and that's how Google got better but really Google did not get better because of page rank all page rank gave Google was five documents to show us and then when we clicked on the third document and when millions of people clicked on the third document they moved the third document up to the first document again go back to the notion of threads it's our interaction with knowledge that then came something where Google is telling us codifying capturing the knowledge that we're generating as human beings and then giving it back to us right and we're like y wow you're awesome I assume that means you're creative and the answer is they still can't design a drone if the Drone hadn't already been designed right and you just keep going on and on and on and then now you look at like all these large language models and they're magnificent in a way the future of I mean Google's in some trouble because the future of search is not links they've been going a long time on links in websites and then our Behavior but if our Behavior stops asking Google questions and stops clicking on Google things Google's going to be in a very bad place so what's cool is now this to me is absolutely the next generation of search engine in that we're going back to uh software that can begin to understand the actual nature of the documents and make the connections between the documents which means this is like search engine times a million because we we look at the documents and it's looking at the language it's making all the connections and it's going to be even more amazing when we start doing stuff with it and I think that's the the problem is is that everyone wants to claim that it's a done deal and it just it solves a problem today and the answer is just like pag Rank and and search engines this the problem will be solved by us using it right and so that okay this notion that we don't need to be there anymore and we don't need to click and we don't need to try things and we don't need to explore these things to make them smarter is is a terrible mistake that's basically what I told him is like you know they're they're not futur looking they're not creative and so we look at them and look at their speed and then we impute somehow creativity on top of that and the answer is their creativity is not growing anywhere near as fast as their speed okay so if I understand correctly you're saying AI is great at looking at what we as humans did speeding that process up but can't invent new yeah it really can't anticipate what our next step is going to be and okay you know and I look at the software that I'm about to write and I look at the software that I have written the stuff that I've written that I've worked on over 10 years it took a path it found a place and it's like yeah yeah that was the right idea I but as I was going forwards through that I was never sure what the right idea was and there's a lot of bad ideas that just kind of fell away but looking backwards you can see the right idea and so the the problem is is that there's so many there's right ideas and so many wrong ideas going forwards that it's difficult for AI to figure out that forward path and so I would love to have some help on the work that I want to do going forward but I just I just can't imagine that the kind of problems that I'm solving they're just unsolved problems right and so if it's an unsolved problem I just don't see how way I can even come close to it like you can say hey build a crawler of course everybody builds crawlers and it's not hard and so yes a I can is a savant at building crawlers but that doesn't mean it's a savant at the kinds of things that good programmers have to do going forward when you hire someone to build a website do you just want it to go back through all the websites of history and find you like somebody else's website and say oh there's your website like I changed the name and the answer is no you want there's something you want something that's cool or uh you know think of when you go to Disney and you got these little wristbands and you hit the wristband that's a different application and once it's been figured out by human beings large language models can gain from whatever it is we document about that but still you know and and they can look at API documentation so much more rapidly than we can and that's really cool and they can generate you know we don't it you know instead of me sitting in looking at some sample Json and writing some Java code for a couple of hours that I can just say hey write some Java code for this Json and that kind of stuff works great it's just it doesn't replace the programmer when you're in India what was the reception to what you said because I love what you said like this says hidden bugs I think this is what you've said before when you try to use AI to write code I code could have full of vulnerabilities and stuff and then you sp the Drone example what did your did your audience counter any of your arguments or did they agree with you so I would say that the best way to think about it is some of my early stuff about code bugs and whatever they just didn't buy it they I mean I I I didn't like win the room at that point I mean they're like oh you're a nice guy you came from far away so we'll listen to your question the only one the only thing where I felt like uh you know mic drop after you know mic drop after it was all done uh yeah that was when I talked about time travel right I talked about the fact that it's really hard to imagine not that it's not capable of looking forward until we force ourselves into a prior context where we know what happened but we can easily see that the AI could never have predicted what was going to happen so that felt like a mic drop moment where you know I felt like a mic drop moment where the room was kind of stunned I mean meaning they got the point when other times all the other kind of Slam all the other dunks on AI that I did in other talks and matter of fact that's when I contacted you and says let's talk and I realized that I had sort of mic dropped the whole room with just an answer to the AI question yeah it's good that you've got multiple arguments against it right because the the problem is always the hype cycle there's so much noise and that's what I really appreciate by talking to someone like you Dr Chuck because you've been doing this for a long time there have been many hype Cycles in in your lifetime right so you know separate the noise from the truth well the other the other story that I I told them and this didn't go over too well and that is I asked him to guess what year it was that I first heard the following statement that from a very respected computer scientist who was not lying had no reason to lie said that computers are getting faster at such a rapid Pace that within two years programmers are not going to have to write software anymore and so I saw that in a quote in a document and you have to tell me what year it was that I first read the end of the end of the end of the life of programmers no I I I have no idea but I'm assuming it was quite a while ago yes it was 1978 oh no wow okay okay I didn't think it was that far back no no it was literally 1978 when I first heard that the end of pro prediction of the end of programming and it was a very short short time what was going on in 1978 was computers were getting faster very rapidly and again this is the problem of speed increasing means that they're capable of doing things that we like them to do faster but then we sort of like put creativity on the same curve is speed and that's why I say you know in 1978 computers had a creativity of zero and in 2000 computers had a creativity of zero and in 2024 computers have a creativity of zero and all they're doing is like finding the trails that of the information that we've laid down for them and looking at those Trails better than we can look at those Trails but their creativity and uh and future look is been zero since 1978 and no matter how much we increase that creativity if it starts at zero it's just not going up and so that's the problem is that we just give too much credit for doing cool things faster and we just assume that creativity is coming along and it's really not so AI is going to be an assistant to programmers but not a replacement yeah no question about it I mean there is just no question about AI I mean because it's a language model which is a whole new technology on the scene it is going to make searching obsolete I mean what what really is going to happen is searching is going to be obsolete because of AI and the key thing is is that because AI can look at all the documents make reasonable connections with the documents before we tell it what documents are important it can it can construct a sense of the importance of information where searching could never do that search engine could never really be informed about the meaning within documents whereas the large language models can look at documents and extract meaning and then make connections that we didn't show them and that's why I say Google is the thing that's most worrisome because Google has never produced a business model for anything other than search and advertising and I think you know just like alavista went on the dust when page rank came out I think if Google can't figure this out then Google's going to be on the Dust I seriously I love that I think it's this is why Microsoft I'm assuming from what I've heard have in has invested so much money into open AI because there a way they can beat Google I I completely agree with that I completely agree with that and and here's the other thing I don't think anybody knows what it's exactly going to look like that is the Google beater right because we're seeing such simple stuff it's like we're we're playing with a debugger interface to chat GPT right now right that that's still not the interface that replaces search and maybe Devon is kind of a new thing but I think they're they're barking up the wrong tree right I mean you know I think maybe github's co-pilot is a smarter thing right where where it's like dude I think you're typing something right now and man I got a feeling I can help you right and you know that and again who's GitHub Microsoft right and and and GitHub co-pilot where it's like dude you don't even type the search request I'm just watching what you're doing and I can help and that's going to be cool that's not creative that's not forward-looking that's just simultaneously while you're doing something you know it's like you got augmented reality and these things are popping up from your like augmented reality assistant and it's like Dude Looks Like You're parsing some Json in Java I know a lot about that you want me to help you know and that's going to be awesome and and and you know the problem there is like I had this problem where I was fixing a the window uh switch in my car because the wires got messed up in my race car right and it took me 3 weeks before I found the pig t the little plug right and so here I am searching and I'm doing all this stuff and I'm doing this yeah and turned out I went to this website called findpigtails.com and I searched and I tried that and then they said if you're having trouble click on this little chat icon down at the bottom and I clicked on the chat icon and it was not artificial intelligence it was a human being and I said I having a problem with that I bought this thing it's the wrong thing I bought another thing and he said well just send a picture like in the chat and I sent a picture in the chat and this human being like within 5 Seconds says that's a that's a p2704 and here's a link to it in our website and I'm like whoa it's right now all their AI I sent pictures to their Ai and it didn't do it um and this person in less than 10 seconds recognized it and told me what to do and so I can imagine that person is like my little helper right he's just like a little thing that pops up and helps me out just you know if if the darn browser had been watching my searches for two weeks trying to solve this dang problem when I say 2009 Ford Mustang door switch pigtail and just like dude looks like you're having some trouble here I'm a helper over here you're I'm your assistant I'm just watching you struggle you know if you just let me know I'll help you on this right and if that that thing then was as smart as that human being turned out to be I I think that's how we're going to write software right it's like I see you struggling you know I I see that you're you're reading this python code and your cursor is stopped on this one line for a while are you struggling a little bit do you want me to is there anything I can help you here and then it'll help you but that doesn't mean it's smarter than you or when it's done helping you it's not going to know what the next line is and so I think it's I think it's a gloriously exciting time you know for programmers it's a gloriously exciting time but it is not the time to stop learning software development I'm glad you said that I have to be sunic cold for one second or or you make make a make a really bad joke it's it's as if Microsoft paperclip has now finally become useful right you know I I think about Microsoft paperclip probably far too often right and and and and and it's things like uh hype Cycles there's all kinds of things that talk about hype and and like you travel a lot you have a lot of busy things you you're trying to figure out when it is that you can even go on vacation and one of the things that the paper like clip like creature was supposed to do is to look at your like whole life and look at all your calendars and watch you perhaps while you're searching for a flight say David uh I see you're searching for a flight on April 9th you do know right that your your brother-in-law's birthday's on April 9th and you've already agreed to D there okay so so that again that's the paperclip right just like I'm here to help you I'm your little administrative assistant and I mean the paperclip was just bad and that's not saying that all little chats in the lower right hand corner of a website are good no no no no no no some are terrible way worse than humans the one I liked was actually a real human and it was impressive and surprising how smart that human was and so yeah that it is paperclip I have to agree with you that you know the idea of paperclip is coming back around but you said I just wanted to back you said that it's a very exciting time to get into software development so if you were talking to your your younger self or you know these people that you spoke to you still recommending to become a developer right oh absolutely and I'm recommending that you become the the master master developer where you instead of trying to learn everything you try to learn a depth of understanding and the one way that I described it um when I was describing the the master programmer experience is that the difference between a regular programmer and Master programmer is the fact that and when when I'm looking at syntax on a screen I see an entire three-dimensional ever evolving set of pictures behind the screen of these things connected to that thing the memor is working the operating systems working the networks working the compilers working the runtime environments working and so when I write one line of code there's like all these things that are simultaneously happening behind it and the problem is if you just think of code AS syntax and you don't understand code as a kind of living creature that operates in some kind of complex environment well you you don't know how to do that and so my suggestion would be don't spend a year learning react don't spend a year learning angular don't spend a year learning C++ don't whatever it is AI will help you with that but if you have no understanding what computers are or what they do then AI can't help you and so this is where the thing we talked about in the past of the programmer is learn fewer things but learn them better and take the time to learn them deeply so that when you're being advised kind of from the side you're you know how to take that advice and you know the meaning of that advice and you know you can separate good advice from bad advice because you see everything uh in a deeper way and you know looking at your expertise in you know networking you when you're looking at the syntax of a Cisco configuration command you're not just going like that's characters you're like oh yeah that's like a line wire and things are moving back and forth and you see protocols happening you know when you're typing a thing and that's the difference between sort of a a master Network configurator person and just a person that's typing stuff and doesn't understand that that's really a an abstraction of a very complex very deep picture that the best people know the picture so uh so David the the main purpose of my trip to India was to talk about the path to the master programmer and I think this is kind of a hard cell and I think I'm really really passionate about it um I have for the past 10 years been giving people their introductory programming courses online through corsera and elsewhere by teaching python to three million people one million of which are in India and lately I've been thinking more about like not like how you get used to programming but instead how do you get a job and so that's where the master programmer comes in the master programmer comes into what I value in an employee and I kind of I kind of summarized it this way and that is if you imagine you're a senior in computer science right now and you're going to go into a job market and you look at all the possible technologies that you should know there's probably 25 to 30 of things that you might touch in the first three years of your job there's no way a college can teach you all 25 or 30 of those things without you waiting until you're 35 years old and then the problem is is as soon as you graduate the thing you're going to work on isn't even one of those 30 things that we taught you so we just have to accept the fact that college can't really prepare you for the job and I started thinking about like people that I've known who are older and have been doing this for a long time and none of their education really prepared them for the exact skills and then I kind of realize that the most valuable thing that I see in a programmer that I like adore them because of this feature and that is any programmer that I like can walk into a new situation and say hi uh we're using angular I apologize we started this project a while back we're using angular and we're using a this and we're using Ruby in the back end and the people that are good they like okay tell me where to check it out I'll go read the Ruby documentation I've never done Ruby and I've never done angular I'll go read the dular documentation and what happens is after a month not only is the good person up to speed on angular and Ruby if that's what it has to be they're one of the strongest people on the team with angular rby because they're coming from a deep understanding of how computers and software work and so it's so it's like if you if you're not ready to learn a new thing in four weeks you have not been well prepared with your education and our job as Educators is to prepare people to learn new things and master them in four weeks so then I'm looking at a curriculum that says not I'm going to try to teach you everything you're ever going to see in the marketplace but I'm try to say what are the foundational Concepts that you need to know and that's when all of a sudden C becomes important C becomes important because C is the way you explain all the other languages and so if a person really is good at C and understands pointers and how memory management works and how operating system works and and how networks work you can literally say oh angular that's that's four weeks and Ruby that's four weeks so that's where the path to the master programmer is it's trying to come up with the fewest possible courses that can lead a person to this moment where we drop them into a job situation four weeks four weeks later they know whatever it is they need to know and they become not only a capable contributor but a leader in that project and so uh I came down to and we've talked about what the curriculum is and I keep changing it every time we talk but the latest version of the curriculum that I came up with is python C computer hardware architecture kinds of stuff and SQL is the four most basic things which is like 12 college credits and then the next three are HTML JavaScript and Java and I my theory is is that's like 21 credits basically a half a year if you master those not just take a crappy course on the syntax from a teacher who really doesn't even understand those language but you walk in as a master of those things there is nothing you can't learn there is just nothing in programming that you can't learn uh once you're there the other thing that's come to me and and part of the part of the if you go back and you look at our previous things I I don't like teaching JavaScript but JavaScript is moved into my core thing but angular and react are not right and so people teach JavaScript and the problem with JavaScript is it's just such a wretched language or historically has been such a wretched language that we got to build hundreds and hundreds of Frameworks to make it useful but those Frameworks are themselves shortlived and so yeah there's a lot of react programmers out there but react is kind of Legacy at this point and what's really cool is that the JavaScript Community is slowly but surely looking at things like jQuery react angular all the ones that get popular and say to themselves how can we pull those features into JavaScript itself right instead of like oh yeah angular's the answer or sevelt the answer and then fighting between angular felt no the the the actual browser developers the companies that make browsers are like that's a good feature for react how about we just like put it in the browser by default but clean it up a little bit and so I'm beginning to see a JavaScript class that has like a 10year shelf life because there's no such thing as a JavaScript or even a Java class that has a 10year shelf life and I don't like building courses that have two your shelf light I mean people can go on You Demi and put up a thing on react 12 and and angular 92 and whatever and they got you know they got like a six month shelf life because then a new version comes out and they'll go make a little money on you know angular N plus1 I don't do those kind of classes I'm trying to kind of create the classics I'm trying to say this yeah because I I build a course that has a 50-year shelf life right this is a 50y old programming language and and and so if it's 50 years old it's got a 50- year shelf life python is 25 years old so it has a 25e shelf life you know and I don't tend to do things things that just got invented last year because usually if something was invented last year it has like a one-year shelf life from a from a teaching so I don't I put so much love energy and devotion into every single class I create I don't want it to have a two-year shelf life if it was a crap class sure I'll make a little money it's crap and uh so I look for things that are truly permanent and are the basis of like a a deep truth in Computing and I'm beginning to think that in the next two years JavaScript is going to be so good that it's not going to change then for 5 to 10 more years and things like angular and react will just slowly fade away just like Jake where is slowly fading away I really want to thank you Dr Chuck for sharing your knowledge and your wisdom and also you like separating the hype because that's the big thing right people see all this hype sky falling and they worry but not only are you doing that but you're also giving people the opportunity who haven't got money to learn the stuff for free so I've put those links below Dr Chuck as always thank you so much thank [Music] you

Channel: David Bombal

Views: 199,894

Keywords: devin, devin ai, ai, artificial intelligence, programmer, coder, chatgpt, bard, chatgtp, chat gtp, Gemini, google gemini, amazon ai, chat gpt coding, chatgtp python, chatgpt c, chatgpt hack, terminator, ai jobs, ai robots, machine learning, cybersecurity, ai cybersecurity, ai sentient, python, dr chuck, dr chuck python, dr chuck python course, learn to code, software development, software developer, how to code, learn to code for free, learn to code python, coding bootcamp, nvidia

Id: iTjYuHDNooM

Length: 34min 3sec (2043 seconds)

Published: Tue Mar 19 2024