CS Professor Sounds Alarm on AI and Programmers

I mean what if all Industries got replaced by AI like everybody's field got replaced by Ai and no humans had jobs would everyone really be out of a job no people will do what people do and have always done they create new opportunities opportunities that are not even known about as of this moment and it will require technical people but we just don't know what those opportunities are that's exciting what's up everybody Travis here from travis. media today we're going to look briefly at a very blunt but thoughtful video on the future of AI and programmers yes we're going to go there so a professor Dr Matt Welsh gave a talk to a group of Harvard computer science students and the talk was titled large language models in the end of programming kind of a bold subject to present to a group of students learning programming anyway the talk was a bit scary but I think we need to discuss it and I think we need to take the caution seriously so that we can figure out where to go next and when to go next and I want to leave you with three key takeaways later in the video to help you stay progressing in this quickly changing industry now I'm a big advocate of learning to code right now as you already know my stance hasn't changed and I'll discuss why again later in the video but it doesn't help to brush the concern under the rug and act like we aren't in a pivotal point in history so today we'll look at three clips from that talk and be sure to comment your opinions below because I think it would be a great discussion so we'll start with clip one which is the premise of the talk so I think what I'm saying here is we are now in an era where we have machines that can take natural language in and produce results algorithmic results computational results but for which no human has written a program in anything resembling a conventional programming language and I claim that these models are going to get so good at doing this that our whole concept of programming computers is going to get replaced over time with instructing language models to do things for us okay that's the premise of the talk this is where we're currently at so we're currently at this point in history we now have machines that can take natural language in or prompts and produce algorithmic and computational results and our model eventually is going to change from us doing the programming to primarily instructing language models to do things for us that's the premise That's the basis of the talk now let me say quickly here we are in this point in history you 100% should be using GitHub co-pilot chbt a vo by verel something like this to Aid in your programming if you're not you're still riding the horse while everyone else is buying an automobile like if you need to use python in the requests package to pull down data from an API and save it to a spreadsheet you can use these tools to produce a starting blueprint instead of having to start from scratch these tools are here to help you and you either start using them and finding the benefit in your own way or you'll end up working slower than everyone else everyone should be utilizing these tools so this next clip this is the scary part in the talk and it's an estimate of how AI can replace the everyday developer now I think this is a bit of a bold claim given the ethical battles that are currently being fault in the inaccuracies in the trustworthiness of the models to be able to work on their own but it's an interesting conclusion one that may come to pass further down the road and again I'll talk about that here in a few but check this clip out clip two how much does it cost to replace one human developer with AI well I did the math so let's say that a typical software engineer salary in Silicon Valley or Seattle around 220,000 a year that's just the base salary doesn't include benefits doesn't include Equity packages doesn't include your free lunch and your bowling alley and all that kind of stuff so let's just assume that that stuff cost you know 92k a year this is again a little conservative so the total cost to your employer is roughly 300 312 k for One S how many working days are there in a year about 260 and so it cost $1,200 a day to employ you as a s at one of these companies software engineer s fair enough okay so let's do the math how many lines of code do you think an average developer checks into the code base every day I mean finalized tested reviewed and approved lines of code most of us whove worked in Industry know that the uh the median value is zero because there's so many days that you go by where you're waiting on somebody else or you're in meetings all day you didn't get anything done you didn't check it in but let's just be generous here and say it's about a hundred I know 100 doesn't sound like a lot people are like but I was programming all day yes but 90% of your code you ended up throwing out or somebody reviewed it and said it was no good you have to rewrite it you were trying to figure out what to do you were revamping it so like the final result of your output is something like a 100 lines of code a day that's the final result how many gpt3 model tokens is that it's about 10 uh tokens per line more or less so and the cost for gpt3 current actually this is probably a little out of date but at the time I made this slide it was 2 cents for 1,000 tokens okay so if you do the math then the total cost for the output of one human software developer on gpt3 is 12 cents this is a factor of 10,000 this should scare us all that's a scary statistic what he's saying is that for 100 lines of code per day instead of paying a software engineer 1,200 bucks a day to write that AI will do it for 12 cent or cents I mean this is what we're all afraid of this is their biggest concern will AI take my job well if it's 10 10,000x cost cutting then yes it will but if that happened what would it look like what would a workflow like that look like well in this last clip he gives an example of what the future may look like with this model what will software Engineers be doing check this clip out but if you imagine what the software team of the future might look like I think this is one very plausible uh approach which is you have a product manager this is probably still a human taking the business and the product requirements the user requirements and translating them into some form probably English maybe a little bit technical English that you then can provide to the AI the army of AI code generators the AI code generators give you a whole bunch of code and probably for a while still we still have humans reading in reviewing the code to make sure that it does what it was supposed to do so the software team of the future may be made up of project managers AI writing code and then code reviewers and unless you or I got into product management we'd probably be the code reviewers in that model making sure that the code actually works that it does what it intends to do and overall Works according to how the product manager is positioning the product or project that's an option so what do we do with all this information what do we do if this is the case for years here on YouTube I've been helping people learn to code and exhorting them to get into the industry do I advise differently now no let me explain I'm going to give you three takeaways from this video and I encourage you to go watch the whole thing it's rather fascinating I left a lot of it out but it's on the cs50 page the cs50 YouTube page and it's well worth your time but here are my three takeaways from the video number one the industry is changing this isn't a bad thing change is good progress is good but it's not for those who ignore it and fall behind talks like this are helpful not only to begin processing where things are heading but where you fall into that picture if you're not using GPT or copile or some similar AI helper then I would suggest that you give it a shot and see if it helps you out and helps you become faster and more efficient at your job it isn't going to go anywhere I would also suggest that you build some sort of app with the open AI API to get familiar with it now I have a video where I walk you through how to build a chat GPT interview bot with fast API and react I'll put a link to that above if you don't have any ideas you can just follow that tutorial but I would suggest you use these tools and get familiar with them number two is that you should not stop pursuing a career in programming think about the jobs out there that'll easily be replaced like cashiers customer service Representatives bank tellers par illegals I mean do I really need someone a person to take my order at McDonald's sure it's nice to interact with people but a computer will probably do it faster actually I just saw today that the creator of Chipotle is about to launch a new meatless restaurant in Manhattan that requires only three humans the rest will be robots this will be the future but when you think about Ai and Rob robots and machine learning Etc what do you think about well you think about programming you think code you think technical people you don't think cashiers and data entry it's a technical field and here's what I think I think anyone with an understanding of computers and computer programming will be of value perhaps it'll be a baseline skill for future jobs that haven't been created yet of which you'll already have that skill I mean what if all Industries got replaced by AI like everybody's field got replaced by Ai and no humans had jobs what everyone really be out of a job no people will do what people do and have always done they create new opportunities opportunities that are not even known about as of this moment and it will require technical people but we just don't know what those opportunities are that's exciting this skill of programming can be seen like a package that you have that validates you into a new era of opportunities it's a baseline skill set so don't think in terms of limiting jobs like my job will be taken by AI think of it more as a huge expansion of new opportunities that will require technical people to bring to life and the third key takeaway is that we just aren't there yet while the open AI Dev day introduced a huge leap in the advancement of AI it's still just a atori right now it's so early we just don't trust AI yet in the big picture there are legal battles there's lots and lots of Errors happening in fact just the other day on LinkedIn I saw this post about gp4 and this study was done this paper produced but basically here were the results gbd4 multi-turn accuracy which means a conversation turns more than once so you ask it a question then you shift to something totally different GPT 4 multi- turn accuracy for editing a PPT was only 6% and then even GPT 4 single turn accuracy was low at 72% creating and 28% editing there's a whole paper documenting all this stuff but AI has a long way to go we're not ready to unleash it fully on anything we're far off from allowing AI to do the big picture programming for Air Products some organizations won't even allow AI in the door you even if you're in a stage where you're just starting to learn to code I would say that you're learning a valuable skill that will be utilized in the future currently it'll be utilized as it is with the goals that you're aiming for but even in the future it'll be a valuable skill how I don't know yet I mean maybe you'll still be a web developer maintaining a large website or a cloud architect or devops engineer automating and specking out systems and while AI can write code there's so many more facets to being a software developer that doesn't involve writing code and that human aspect of it is so vital to so many companies so again I think learning to code is still such a vital skill in this wide- ranging industry but one that will allow a baseline skill set for all of the jobs to come as AI gets more and more adoption and more and more accurate and trustworthy so what do you think what are your opinions where do you think the industri is going have you seen the talk I'd love to hear your comments below and love to have a discussion with you if you found this video helpful give it a thumbs up if you haven't subscribed yet consider doing so and I'll see you in the next [Music] video

Channel: Travis Media

Views: 283,747

Keywords: ai and programming jobs, future of ai and programming, ai and computer programming, chatgpt 4, chatgpt programming, github copilot, ai, machine learning, will ai replace programmers, will ai replace software engineers, the future of ai technology, will ai replace software engineers in future, what coding will look like in the future, travis media, ai coding

Id: uuvjOvAdnOk

Length: 12min 20sec (740 seconds)

Published: Wed Nov 15 2023