Is GPT-4 Really 10x Better At Programming? | ChatGPT

chat GPT is once again making headlines with the release of version 4 which has been absolutely crushing a bunch of standardized tests so today I want to put it to the test specifically with software engineering so I'm going to give it a variety of different software engineering tasks and compare it to how I think a human would do and compare it to the previous version of chat GPT so the way I'm going to do this is by starting with some coding interview questions because after all we all have to pass coding interviews to become software Engineers so we should hold chat GPT to the same absurd standard and then after the coding interviews I'll give it some more practical actual software engineering tasks to see how it does on that as well as some debugging see if it's able to debug code and finally I'll also test it on its knowledge of computer science so for the coding interviews I'm going to get them from algo expert we'll do 10 of them starting with some easies and then moving up to some hards and very hards so the first question is validate subsequence it takes in an array and a sequence and it needs to tell us if that is a valid subsequence of the main array so I'm going to paste the entire prompt into chat GPT as well as I'll switch to GPT 3.5 to begin with so the old version and we can see we have the prompt here and I added this a little bit at the beginning with some context so essentially you're doing a coding interview in Python enter the following question by only providing correct efficient python code and you do not need to explain the codes I don't want a bunch of extra text but then give time and space complexity at the end so let's go ahead and run this and one thing I have noticed is with the pro subscription GPT 3.5 runs almost instantly which is super cool so it's way faster so it says time complexity over in space complexity o of 1 and we have some code here so let's copy the code into here paste it and we can simply run our code and it did actually pass all of the test cases so 24 out of 24 test cases correct and now I will open up a brand new chat and this one will be with gpt4 it will paste the exact same prompt in and we'll see how it does in comparison so it is going to take a a little bit longer so gpt4 is not as fast as GPT 3.5 but supposedly it does a better job one thing I'm noticing here is it's using very strange formatting it doesn't have the code formatting and then the time and space complexity is absolutely massive but they are still correct so let's go ahead and paste this code in and it didn't actually indent it for me so I'll just go ahead and indent it for it which I don't know if we should take some points off for this but it is I believe still correct so we can run this code and we do still pass 24 out of 24 test cases so both of them were able to pass this although gpt4 did format it kind of strangely okay so I'm going to give it nine other coding interview questions collect some data and I'll check back in all right so here is the data actually they performed almost the exact same which was interesting I've seen some other data that people have shown on Twitter and whatnot where gpt4 is doing much better than 3.5 in the coding interviews but at least in my small test they did basically the same if they got the question and correct I gave both of them the chance to do it again and in any of the cases where they got it incorrect actually both of them were able to correct their Solutions in every single case so ultimately all 10 questions they were both able to get correct which is interesting because the last time I tried to do coding interview questions with GPT 3.5 it was actually not doing a very good job at them so I don't know if I got better at that or if I just got lucky but in my small test they both did basically the exact same they did however struggle a little bit with time and space complexity some of the problems I gave it were just sort of difficult though for example one of them had a factorial in the space complexity and neither GPT 3.5 or gpt4 was able to figure that out also as far as the formatting bug it seems to be a bug on the actual website it's happening with both models so whenever the formatting is messed up I'll just fix the formatting and we'll assume it did that correctly okay so now what I want to do is test it on some more practical coding and what better way to start some practical coding than seeing if it can do the age-old hardest thing to do in all of computer science and that of course is how to center a div so how do you Center a div we'll start with GPT 3.5 and see what it says okay so to Center div horizontally you can set the margin property to Auto and specify a width for the div so I guess this does work I don't love that it did it inline in HTML but I guess I didn't specify otherwise that it's fine in the example above the div has a width of 50 which means it will take up half the available space the margin is set to zero Auto o which means that the left and right margins will be set to Auto centering the div horizontally okay so that seems correct if you want to do both horizontal and vertical you can do display Flex so we have div style display Flex align item Center justify content Center this is a sensor div so this is correct although it doesn't necessarily Center the div it's centering the content within the div so it's like it's this explanation down here is correct but it's not really centering the div itself I would want to see that as like a wrapper around the div but either way I'll count this as correct let's move on to gpt4 okay so the gpt4 answer took like five minutes to generate I kept getting network errors and all of that but I'm not going to hold that against the model and the answer actually seems much better so to sim to a div we have some different options and it did actually write CSS Force instead of doing it inline which is nice but again it's not really what I asked for so either way it's fine so to send your div horizontally using margin we have margin left and right as Auto the display block is a little bit strange because it div it should know already is display block by default but that is I guess fine it's not going to do anything Incorrect and then we have our HTML with the content in the middle and this would be centered and then we have using flexbox and it did this I think much better than GPT 3.5 so it actually recognized in the HTML that we need a parent div and this Center div inside again 3.5 sort of hinted at this and the writing it had after it but in the actual code it didn't do that so here we have nested divs which is what you really need with flexbox and then we also have a grid solution which we did not get from 3.5 so this is a little bit more complete a little bit less concise I guess you could say so here we have a parent with justify items Center and the line item Center and it gave it a height as well and this is display grid and this should also work to Center this div here and then finally it also gave us a transform property solution which also I guess works so here we have position absolute top and left 50 and then we translate negative 50 in both directions which would also work and then we have the same Center div inside of its parent okay so next I want to do something sort of fun I want it to be a little bit creative while also writing code so what I'm going to do is ask chat GPT starting with 3.5 if it can write HTML and CSS code to create a cartoon pig and to give us just the completed code and I'll go ahead and copy and paste this code into a browser and I'll put it up here so you all can see what it looks like okay so as you can see here the CSS it rope basically doesn't work most of this doesn't really look like a pig it sort of had the right idea down but the execution was just a little bit off so let's copy this prompt and move over to gpt4 paste this same Crompton I don't need my name there and we'll see if chat GPT 4 does an even better job okay so we got the answer back from GPT T4 as well I'll put it up here I guess we can debate as to which one's better I would argue the gpt4 one got a little bit closer although it looks sort of demonic it's a little bit scary to look at but I do think gpt4 did a slightly better job with this one but neither one was able to like blow it out of the water but that said it was a pretty hard challenge okay so now we actually have similar questions to this on front end expert for CSS practice so for example this one you need to make this cartoon robot Emoji thing so what I'm going to do is give this to chat GPT the entire question and see if given a little bit more structure if it's able to generate the CSS for this so first of all let's start with GPT 3.5 so I'll paste the entire prompt in here and we'll see how it does okay so we got the GPT 3.5 answer back so here we have all of the different code and it also gave us the HTML back I'm not entirely sure why I did that but that's fine the HTML is unchanged and we can copy this over here paste it and it got pretty close it's not correct obviously if I submit this it's not going to be correct but it did get pretty close so let's see if gpt4 is able to get closer than 3.5 was okay so gpt4 again took a while to actually generate but we did get some CSS so what I'm going to do is copy this CSS come over to our robot emoji on front end expert we can see the GPT 3.5 solution it's come to solution two paste in the gpt4 solution that looks like it got closer so if we toggle between these you can see solution one which is 3.5 and solution 2 which is gpt4 neither of them are correct if I run this it's not going to pass but it is much closer okay so now I want to test chat GPT a little bit on some JavaScript and I'll use some JavaScript interview questions for that from Front End expert so first three of this promises question which essentially is to implement promises as a new class without using the built-in promises class so I will paste the entire thing in for chat GPT 3.5 and we are getting a solution already it looks like we are getting that bug where some of it is highlighted and some isn't but I'll fix all that and get it pasted into front-end expert okay so after much struggling with GPT 3.5 I did finally get it to give us the code all at once so let's copy the code into front end expert make sure not to delete the export at the bottom they'll also delete the sample usage it gave us because we don't need any of this and let's see how it does okay so it passes 16 out of 18 test cases so that's not bad at all it's not passing all of them let's go ahead and try giving it one of the failed test cases so I'm going to paste this entire test case into chat GPT so say it failed this test case and then I'll just pass in the entire test case and see if it is able to fix the code okay so it did generate some new code but I also pasted it into front end expert and now we're actually only passing 6 out of 18 test cases so it wasn't able to debug its own code but but 16 out of 18 to begin with is not terrible okay so gpt4 was also able to complete and it gave us a very similar almost the exact same solution as GPT 3.5 I pasted it into promises on front end expert and fixed the indentation and it actually fails the exact same test cases so you can see 16 out of 18 and we get these same failures so I'm going to give it the exact same prompt and maybe gpt4 will do a better job of debugging itself than 3.5 days okay so it took a while to generate but we did get all of the code together again so that's nice so we'll copy this code let's go to solution three paste this in and I think I deleted the export so we can say exports dot my promise equals my promise run this code and it looks like now we pass 5 out of 18 test cases so we're doing even worse than GPT 3.5 debugging so neither of them was able to properly debug their failed test cases on this problem but it is a very hard problem so not the end of the world but they were not able to solve it okay so for a point of redemption what I'm going to do is give both GPT 3.5 and gpt4 a bit of an easier JavaScript question so we're going to give it debounce which is a very classic question it's not easy by any means but it's not super difficult to either so starting with 3.5 let's paste The Prompt in here and we are getting a debounce function very quickly this time so I will copy this code and let's go ahead and paste it right here run the code and we passed 10 out of 14 test cases so not terrible by any means but it did a decent job so once again what I'm going to do is give it the test case that it failed and see if it's able to debug itself based on that failed test case and let's go ahead and copy this code and I will paste it in here I'll paste it as solution two see if it does any better run this code and we get 7 out of 14. so once again trying to debug based on the test case and it actually ends up doing worse not better okay so gpt4 once again had the bug where it's a little bit hard to read the code however I did go ahead and paste all that code into frontend expert you can see when I again this code we actually do pass all 14 test cases so this is the first case I've seen where gpt4 was able to perfectly solve a problem and GPT 3.5 was actually not able to solve the problem even after I told it what it did wrong so that's a big win for gpt4 but now what I want to do is see if we can do something that I did in my first video I ever did with chat GPT and it was giving me a big headache and that was trying to create a tic-tac-toe game so I'm going to paste this prompt in here telling it to create a tic-tac-toe game for us using HTML CSS and JavaScript and we'll see first if GPT 3.5 can give us a game we can play and then we'll see if gpt4 can give us a game we can play and maybe it'll be a better game okay so GPT 3.5 told us unfortunately as an AI language model I'm unable to provide the code as I don't have the capability to create files or execute them then says however I can provide you with an example of the code that you can use as a starting point so I guess maybe it's saying it can't do everything but it's giving us sort of a starting point so we have HTML CSS and then JavaScript but it did do that thing where it didn't completely finished so what I'm going to do is just say continue and see if it's able to give us the rest of this code and it does look like it did that so I'm going to go ahead and paste all this code together and see if we have a working game okay so this is the game that GPT 3.5 gave us the CSS doesn't seem to work we don't really have a grid we just have a big row but it does say it's access turns let's try clicking and we do get an X now it's O's turn so we'll click there access turn so this part does seem to be working I'm not really sure how to win because I don't know what this grid represents but it says X wins I don't know if that's actually won there or not but let's go ahead reset does seem to work so let's try doing something like this does that make X when it does okay so the game does seem to work it just seems like the CSS is broken but this is a big win over what happened the first time I tried to get tic-tac-toe out of GPT 3.5 okay so gpt4 was able to finish again we got that bug where it sort of is just giving us the code without proper formatting but but it gave us a bunch of code it read it all in HTML and used like a script tag in the style tag but that's perfectly okay I pasted it into a code pin and at least stylistically it looks much better than what we had before we do have a grid that looks like tic-tac-toe so let's go ahead and click and we do get an X if I click here I got an o x o x and now we get player X wins so that does seem to be working let's let o win real quick make sure that works as well and that also seems to be working can we win diagonally and we can so it does seem like this tic-tac-toe works perfectly so awesome job to chat gpt4 this is much better than 3.5 gave and way better than we got the first time I ever tried to do this okay so next what I want to do is see how it can do with dom manipulation in JavaScript and see if it can make an API request to an API if I describe it to it so what we have here is this infinite scroll problem that we have on front end expert where you have to make an infinitely scrolling container of testimonials so I'm going to paste The Prompt into GPT 3.5 first of all and see how it is able to do okay so once again we got an answer super quick and we have proper formatting so woohoo on that one so we have all of the code here it looks pretty good at first glance and then it did give us an explanation below which is nice I didn't ask for an explanation but it did give it to us so that is cool and we can come in here paste the code and let's see how it does okay so we did properly fetch testimonials however it doesn't seem like the infinite scrolling is actually working as I scroll to the bottom here we do not seem to be fetching more so it's getting close but it is not completely working it passes four out of the 10 test cases let's go ahead and see if it can actually debug itself so I'm going to say this loads the initial testimonials however it does not load more on scroll and let's see if it is able to fix itself okay so I apologize for the mistake in my previous answer you are correct the codec provided only loads on the initial only loads the initial testimonials and doesn't load more on scroll here's an updated implementation and it looks like this is complete now let's copy this code come over here and paste it in and let's see in the browser output are we able to infinitely scroll we still are not so it wasn't able to fix its mistakes and again we get 4 out of 10 test cases passed okay so gpt4 was able to actually finish pretty quickly this time I like the way it sort of formatted this answer so it told us in steps what it's going to do and then it gave us the code so let's copy the code it gave us we'll go to solution 2 here paste in the code let's see the browser output so we do get the initial testimonials and as we scroll again we're not getting more so I suspect we are once again going to get 4 out of 10 test cases and again we do so I will do the same thing with gpt4 I'll tell it it doesn't work on scrolling and see if it's able to fix itself okay so gpt4 just got back with its corrected solution and said the issue was that the scroll event listener was on the window object instead of the testimonial container which seems like a valid issue so let's copy this code come back to front end expert I'll paste this into solution three let's check out the browser output tab and this does seem to be working or at least working better let's run the code and 10 out of 10. so GPT 4 was actually able to correctly diagnose the issue and fix its code whereas GPT 3.5 initially had basically the same incorrect code but it was not able to fix itself so much better on the gpt4 side okay so next I want to test some react knowledge so we're going to start with a react hook so we have this question use state with history which is a pretty classic react hook so let's go to GPT 3.5 paste in the question and see if it's able to write a use state with history hook and it seems like it's doing it very quickly okay so now we have our hook we have some explanation here as to how it supposedly works but let's see if it actually works so I will paste this into here and we can run this code and 0 out of 12 I think we just need to delete this export default run this again and we get 11 out of 12 test cases passed okay so as we did before I went ahead and gave it the failed test case let's copy this code come back here and we will update this to see if it is now working and we will remove that once again run the code and still we get 11 out of 12. so it gave us different code but it's still failing the test case so again GPT 3.5 isn't doing a great job of debugging when I give it the test case that it's failing okay so gpt4 just gave us the code it didn't really say much for an explanation let's go ahead and copy this into here so we have use state with history we can run the code and failed one out of the 12 test cases it looks like it failed the same test case as GPT 3.5 did so again we get the same results initially let's see if gpt4 does a better job of debugging itself okay so gpt4 is confidently telling us that it has a correct answer now so let's go ahead and copy this and we can paste it in here and see if this updated version actually passes and it does not so it still fails test case 11 so both of them weren't able to debug in this case but they did get pretty close regardless okay so now what I want to do is see if we can get an entire react application written for us so I'm going to ask first of all GPT 3.5 to create a Connect 4 game using react and we'll see if we are able to play this game okay so I've been trying with GPT 3.5 for a while trying to get it to debug its own code and just sort of regenerate itself a few times to get Connect Four working it's given us some code that looks pretty close to valid but no matter what I've done it's not been able to actually give us working code like it doesn't even work a little bit it has syntax errors in it and it's not even really close to working granted this is a big thing to try to ask it it's trying to build an entire sort of application but the GPT 3.5 version is effectively something I wouldn't be able to use at all maybe I could ask it for small components at a time and then put them all together and get it to work but the just asking for Connect Four outright with react is not working at all on GPT 3.5 okay so gpt4 seems to have done a bit of a better job here so what it did is it actually gave us instructions on how to make an entire application so first like even told us to go to create react app and what files we need to edit and where to put them I broke it all down into different files for us it did sort of run out of space like it does sometimes so I had to do it a little continue and when it did the continue it was sort of a mess so I got it to give us all the code back together which it was able to do so I went ahead and made a code pin with it and this seems much better than what we had before the CSS at list is working so let's see what happens when we click so let's say we play there we get red shows the middle let's say there yellow chose the one next to it so it seems to be working let's see if we can actually win the game okay so winning the game doesn't seem to actually do anything okay so it's actually connect five instead of Connect Four but that is pretty close to actually working so definitely this one goes to GPT 4 over 3.5 let's see if we can play again also okay so it seems like we can even play again see if we can let yellow win horizontally we can't but maybe that's because it needed five in a row okay so it doesn't perfectly work because player two should not have just won and it says it did but this is much closer than GPT 3.5 was okay so now I want to shift gears a little bit away from actual coding and see if I can do some of the other aspects of software engineering and computer science starting with system design so let's ask it to make a system design for an infinitely scrolling news feed similar to the Facebook news feed okay so we have our answer so first it just tells us what infinite scrolling is this is fine and then it gives us a high level overview of a system design so it tells us we need a database to store all the content for the news feed it doesn't tell us what type of database we need but maybe we could drill down on that by asking more questions then it tells us we need content retrieval so this is a way to actually retrieve information from the database which makes sense it tells us we can use caching indexing and sharding to improve performance which would be important then we need a Content ranking system the system should have a mechanism to rank content based in relevance popularity and other factors so I would agree with that it says we can use machine learning or a rules-based system which makes sense to me and it should be updated regularly to reflect changes in user behavior and preferences then we have the actual infinite scrolling part which seems good to me as well then we have user engagement which is basically telling us we should have likes comments sharing things like that then we have personalization so we need some way to personalize settings and then we have scalability so the system should be designed to handle a large number of users and a high volume of content this can be achieved by using scalable Technologies such as cloud computing distributed databases and to load balancers as well as by optimizing the system architecture for high availability and default tolerance so overall I would say it gave a good answer everything it said seemed to be correct to me although it didn't go super in depth in anything but that said I'm sure we could drill down on all of these details if I asked more questions so not a terrible answer certainly not the best answer I've ever seen either well let's see how gpt4 does instead with the exact same question okay so now we have an answer from gpt4 and let's see how it compares to that of GPT 3.5 so first it tells us we need for the front end some UI components and it tells us what components to implement which is already I think a really good step and then it tells us we need an infinite scroll so this is going to be a scroll event listener that triggers a function to fetch new content when the user reaches a certain scroll position so perfect then it says to implement pagination to fetch content in chunks EG 10 posts at a time and it says to use the bouncing to prevent multiple fetches in a short amount of time so from the front inside this seems pretty much perfect to me and then we have the back end and it tells us API endpoints first of all so we have a fetch post endpoint which is an endpoint for retrieving a batch of posts based on users interests Trends and pagination parameters so that sounds good then fetch additional post data to get data such as comments and and likes for a specific post so both of these seem incredibly reasonable to me and then we have data models so first we have a user to store information about the user we have a post to store information about the post and a comment and then it tells us we need an algorithm and ranking as well so it says to implement an algorithm to rank and prioritize posts based on factors such as recency relevancy popularity and the user's preferences and periodically update the ranking algorithm based on user feedback and engagement metrics so it's very similar to the algorithm that was described in GPT 3.5 and then we have a database so choose a database system that supports efficient retrieval and sorting of data design an appropriate schema to store user post and comment data and to implement indexing for faster querying and sorting of data I like that it gave us a little bit more information about the database here and how to choose the correct type of database even if it didn't go super in depth with any specific examples then we have caching Implement caching on the front end to store recently fetch posts and minimize the need for additional API calls so that's great I think that's very important and to implement caching on the back end to store frequently accessed data such as popular posts or user information also I think great then we have monitoring and scaling so monitor the system's performance and the load to identify bottlenecks and potential areas for improvement and optimize the system for performance and scalability by employing techniques such as load balancing database charting and content delivery networks and then we have security and privacy which I don't think the other answer touched on at all so Implement authentication and authorization mechanisms to protect user data and ensure only authorized users can access certain content and regularly review and update the system security measures to safeguard against potential threats and vulnerabilities okay so I think both of these answers were reasonable it seems like gpt4 went more in depth I'm not sure how much that is due to gpt4 being more advanced versus it's just that maybe gpt4 is programmed in chat GPT to give us more depth and the answers versus GPT 3.5 seems sort of focused on speed but regardless this was a more in-depth answer but I think both answers were reasonable and a good starting place that you could use to drill down in those details next I want to test some debugging so what I'm going to do is paste some python code in here that actually has a bug so this is supposed to be an in-order traversal however it does not work and essentially this does not work because of the fact that we do not have the base case at the very beginning and let's see what it says so the code you provided for the inorder traversal of a binary search tree is not correct as it lacks base cases and initial checks so that's exactly correct it will fail with an error if it is called with an empty tree as there will be no tree dot left or treat out right to Traverse here is a correct diversion and that is completely cracked so it perfectly realized that this code was incorrect so I didn't tell it it was incorrect if you read this prompt it all it does is say is this correct and if it's not fix it and it was able to perfectly identify the issue so I would imagine gpt4 will also be able to do this let's go ahead and try that now so we can switch to gpt4 paste the same Crompton and we'll see if it gets it correct okay so gpt4 was also able to perfectly identify that the code was not correct and it gave us basically the exact same answer so now I want to do one more debugging test this one is actually going to use Java and the issue here is that we would get an index out of bounds error so let's see if it is able to fix it this is GPT 3.5 the code is not correct so the for Loop has an off by one error the condition in the loop I is less than or equal to nums that length should be I is less than num cell length this is correct because arrays in Java are zero based indexing meaning that the last element in the array is at index nums.length minus one so this is correct and it also tells us the correct error we would get and it was able to fix the code so GPT 3.5 did a great job of debugging here as well okay and gpt4 also gave basically the same message so both of them are able to debug pretty simple code we did see earlier that gpt4 seemed to do better in debugging overall but at least in these simple examples they both were able to do it okay so now I want to test its knowledge of computer science so I'm going to give it five different computer science questions and see if they are able to answer all five of these questions so the questions are to explain how the internet works how computers work what the differences between a compiler and an interpreter what different programming paradigms are and what public key encryption is so first how the internet works the internet is a vast network of interconnected computers and devices that communicate with each other using standardized protocols when you send information over the internet is broken down into packets and sent through a series of routers which direct the packets to their destination the information is then reassembled at the destination computer or device this process happens very quickly allowing you to access websites and emails and communicate with people all over the world so a basic explanation but that is correct how do computers work they process information through a series of electronic components these components include the central processing unit memory input output devices and storage the CPU is the brain of the computer and responsible for executing instructions memory is used to temporarily store data and instructions that are currently being used by the CPU IO devices such as keyboards and monitors allow users to interact with the computer storage devices such as hard drives and solid state drives are used to store data and programs for lighter use so again very basic level but this is correct what is the difference between a compiler and an interpreter a compiler and an interpreter are both programs that translate code into machine readable instructions however they work in different ways a compiler takes the entire source code and translates it into executable code in a single step the resulting executable code can be run directly on a computer in contrast an interpreter translates source code one line at a time executing each line as it is translated this allows for a more interactive development process but can also be slower than using a compiler and different programming paradigms that gave us procedural programming object-oriented programming functional programming and event driven programming and these all seem like reasonable answers as well then we have public key encryption which is a cryptographic technique that uses two different keys to encrypt and decrypt data one key the public key is freely available and used to encrypt data the other key the private key is kept secret and used to decrypt the encrypted data because the private key is never shared it provides a secure way to send and receive sensitive information over the internet public key encryption is widely used for secure online transactions such as online banking and e-commerce so none of these answers were in depth by any means but they were all for the most part correct but let's see if gpt4 gives us a more in-depth answer and gives us some extra information okay so now we have the gpt4 version as well so how the internet works it seems like this paragraph is roughly the same thing it tells us that it's a Global Network of interconnected devices but then it also gives us information about the Internet Protocol the transmission control protocol and how that's working so it doesn't just tell us that packets are being sent it tells us sort of a little bit about how that's happening and then it mentions HTTP as well as FTP so it does give us just a little bit more information than we got from GPT 3.5 next we have how computers work so a computer is an electronic device that processes data following instructions given by a program or software the CPU is the computer's brain responsible for executing instructions and Performing calculations memories such as RAM and storage holds data and instructions for the CPU i o devices keyboard mouse allow users to enter information while output devices so a monitor or a printer display the results of processing software including operating systems and applications provides the instructions needed to control the hardware and perform tasks so again I'd say both answers were correct but I think this one was just a little bit more in depth then we have the difference between a compiler and an interpreter a compiler and an interpreter both are tools used to translate high level programming languages into machine readable code the main difference between them lies in their execution process and it seems like both of these gave roughly the same answer on compilers versus interpreters and then we have programming paradigms which are approaches or styles used to structure and organize code some common ones include imperative object-oriented functional and logic and then it gave us some examples of each and then we have public key encryption which it tells us is also called asymmetric encryption is a cryptographic system that uses two separate Keys a public key and a private key the public key is freely shared and can be used by anyone to encrypt a message while the private key is kept secret and used to decrypt the message then it roughly tells us how these are used and it says public key encryption is widely used in Secure communication protocols like SSL and TLS which protect data transmission over the in internet and in digital signatures to verify the authenticity and integrity of messages or documents so overall again not a super in-depth answer but I think a good one at that so I'll leave it up to you all as to which one you think did better on the computer science questions I think both did fairly well and overall I would say that both GPT 3.5 and gpt4 did pretty well in most of these tests it seemed like gpt4 was able to give a little bit more detail on all of its answers as well as its code was generally a little bit better but for the most part both of them did almost the exact same on all of the coding tests but that said I have seen a lot of examples online where gpt4 has far outperformed GPT 3.5 and some coding tests so your mileage may vary there and there's also the difference of GPT 3.5 seems to work about 10 times faster it's nearly instant whereas gpt4 takes a while to get the answers back either way it does seem like a big progression and overall I don't think we're at the point where GPT is anywhere near pricing software Engineers but at the same time it is very good at software engineering and it can be a useful tool that you can use to assist you in your own software engineering job so with that said that is going to be the end of this video If you enjoyed it make sure to subscribe for future content and I will see you next time

Channel: Conner Ardman

Views: 81,132

Keywords: coding, software, engineering, software engineering, sde, swe, software development, development, web, web development, developer, tech, frontend developer, engineer, javascript, front-end, front end, frontend, chatgpt, gpt, gpt-3, gpt3, open ai, open, ai, aritifical intellegence, artificial, intellegence, ml, machine learning, machine, learning, html, css, interview, coding interview, chat gpt, openai, gpt-3.5, gpt-4, gpt 4, gpt-4.0, 4.0, 3.5, chatgpt-4, chatgpt-4.0, version 4

Id: Y9LIDMV-fw4

Length: 35min 4sec (2104 seconds)

Published: Thu Mar 16 2023