CS Professor Sounds Alarm on AI and Programmers

In this video, we'll take a look at a few clips from a very blunt, but thoughtful, lecture that a computer science professor gave to some Harvard students.
The premise is that AI will bring about "the end of programming."
After hitting the key points of the talk, I'll provide you with 3 KEY TAKEAWAYS so that YOU can stay relevant and continue to progress forward in this quickly changing software engineering industry.
Let me know your thoughts in the comments.
My Newly Launched Coding Community - imposterdevs.com
Timestamps
00:00 Intro
01:20 Where we're at now
03:15 The savings of AI replacing developers
06:08 The software team of the future
07:30 3 Key Takeaways
07:42 The industry is changing
08:27 Do NOT Stop Programming
10:05 We're just not there yet
11:50 Discussion
Video resources
CS50 End of Programming Talk - • Large Language Models ...
ChatGPT Coding Project -
• Let's Build a ChatGPT ...
Udemy deals - travis.media/udemy
** Career Path Coding Tracks **
Web Developer - geni.us/jBigBd
Software Engineer - geni.us/AbMxjrX
Machine Learning - geni.us/GporLlT
Python Developer - geni.us/tv2FJBU
DevOps Engineer - geni.us/MgHtJ
** My Coding Blueprints **
Learn to Code Web Developer Blueprint - geni.us/HoswN2
AWS/Python Blueprint - geni.us/yGlFaRe
Both FREE in the Travis Media Community
My microphone - amzn.to/3sAwyrH
** I write regularly **
travis.media
** FREE EBOOKS **
📘 travis.media/ebooks
LET'S CONNECT!
📰 LinkedIn ➔ / travisdot. .
🐦 Twitter ➔ / travisdotmedia​
🙋🏼‍♂️ Website ➔ travis.media
#ai #chatgptcoding #futureofai #aiprogramming
** Some of the links in this description may be affiliate links that I may get a little cut of. Thank you.e

[@zwanzikahatzel9296]

In my experience "reviewing" code can take longer than writing it from scratch, depending on how complex and deeply you need to be able to understand it. If the code works and produces the right the result you might give it a quick pass, but if something doesn't work and you need to track down the root cause of the issue it might take a whole afternoon or even days. These models make so many mistakes at this stage that they are more of a nuisance than a benefit. they can be ok to quickly sketch a script or to look up commonly needed and self contained algorithmic pieces such as "sort a pair of lists by the second element" or "traverse two trees in parallel a put corresponding elements in a list of pairs". It absolutely CANNOT work with a very large system when you have to think holistically and consider subtle interactions between 100s of functions or objects. The only "programmers" who should feel threatened are the ones that produce simple amateurish code, who just glue together pieces from large libraries or stuff they find on stack overflow. If your job requires creating new interesting algorithms or unusual ways of processing and combining data, these tools wont be as helpful.

[@stanislavkindiakov6334]

Great summary. Exactly understanding the problem and checking existing code base takes 90% of the time. Writing of a code itself when you have an idea is pretty fast and straightforward process.

[@caty863]

Only 0.0000001% of programmers have jobs that require *"creating new interesting algorithms or unusual ways of processing and combining data."*

[@zwanzikahatzel9296]

@@caty863 thats why those are not called "programmers" but rather "soydevs" ahahah

[@btm1]

@@caty863 well maybe 1-2% but still, your main ideea is valid

[@luke5100]

To his point though, these are like at the Atari 2600 stage of their evolution. Think of what personal computers could do in 1980 compared to what they can do in 2023. I mean GPT4 just got a massive token limit increase, up to 120,000. That is huge. 300 pages of text. It’s not a stretch to imagine a token limit generous enough to allow for the parsing of an entire code base, or at least large portions of it at a time

[@calmhorizons]

I watched the talk in full - the presenter is a marketeer for his own AI startup, pure and simple.

[@troyforce2]

True, but it doesn't change the fact that this is a coming relality.

[@calmhorizons]

@@troyforce2 for sure. Hucksters will encourage companies to fire juniors and use AI instead, seniors will be overworked as a result, their work will degrade, a new generation of developers will not be trained, and in 10 years we will be in the same position with developers as we are right now with plumbers. Personally I will profit from that as a programmer, but I don't think it is good for humanity as a whole.

[@podunkman2709]

This one is also good one. Have u ever seen "humanless restaurant"? 😅 Visions at least two decades old. It will not happen. It's too expensive, no real need. Just BS.

[@macolulu]

Why would we be come someone like a plumber? That's not too bad in my perspective if I can be plumbing in a much older age still.

[@josh2482]

@@podunkman2709 You are right, the idea of an automated restaurant has been around since the 60s. Nothing has ever panned out.

[@bwhit7919]

I’ve been using Chat GPT to make an app in a new programming language. I spend most of my time thinking about data management, UI design, and debugging. It’s no longer necessary to memorize syntax, but the subjective coding skills are still super important.

[@TheB1nary]

That’s a great way of handling the issue - well done!

[@PraveenKumar-bo7fw]

I don't believe you. If you are solving any non trivial problem ChatGPT fails miserably. All it does is provide you with boilerplate, which were just a google search away any way.

[@marusdod3685]

syntax has always been the least challenging part, you're probably an undegrad

[@BoringAIbiz]

How’d the process go

[@tnamr5652]

@@marusdod3685 True.Human brain priceless.Billion parameters for a useless answer

[@Zvona555]

I used to work various programming jobs, now work in lower corporate management for years. Coding (writing loops, functions, classes...) will be replaced to some degree with AI, but coding is only a fraction of software development. When we come to higher project levels, it becomes difficult to even describe it properly to every detail to yourself. You imagine something, try to define it, start working, review, adapt, rewrite... and you still have to know how to code, even if you don't do it directly. AI still has a long way to go before reaching that level of complexity. It will probably need General AI, but with General AI around we will have more serious problems in general than just replacing developers with AI.

[@izamalcadosa2951]

You are 100% correct! AI is grossly overrated and overstated!

[@ffc1a28c7]

It's really annoying when people act as if LGMs are equivalent to, or even will lead to Artificial General Intelligence. There is no understanding in the model; it's just *really* powerful language processing, and regurgitation.

[@Zvona555]

@@ffc1a28c7 Agree. LGM is still a "statistics on steroids" and it's generative feature is still doing sophisticated fragment recombinations from large pool of processed (and mostly stolen) data. It will change IT industry, but I'm not concerned about making programmers obsolete in general. Some simpler programming jobs (like web front-end) may disappear, but those jobs have been becoming obsolete many years before (by non-AI tools). On the other hand, some programming jobs will become much harder - debugging for example. I'm not talking about ordinary bugs underlined by Intellisense (which is already some sort of AI). In my experience, every project ends up with some nasty bugs which take weeks and months to find and patch. That task will become much harder with more and more AI generated code and will require even more programming experience and knowledge.

[@SeeAndDreamify]

If you think about it, describing what you want to do is what coding already is. If coding with LLM:s ends up working out, it will not be that different from using an abstracted library. You are just coding in natural language.

[@edwardmacnab354]

@@izamalcadosa2951it's at the hype stage . They are selling it big time and everyone will jump on the bandwagon , including at the academic level and it's development will accelerate as a result--we are at the TIP of the iceberg .

[@user-wb1rs8bb5z]

I love your prespective on this, Travis. It gives me hope.

[@samsturdi]

Ah yes, a professor speaking authoritatively about the private sector. Classic

[@matthiasblum6555]

I had a programming lecture for C/C++ over 20 years ago, and the Professor said basically the same thing. Software Developers will be obsolete because of modules ruling everything. Who needs Software Developers if a monkey can do his job with modules? Yeah, that was working really well... Me thinks, the problem is, buzzwords like AI. Which can be everything nowadays, from simply just using one Machine Learning Algorithm like Random Forest basically randomly on some Sales numbers to highly complicated AI Agents doing almost everything in a proprietary/island solution. This is a problem itself. What is it now? The buzzword bs or a really highly sophisticated island solution? As a "IT guy" I am not impressed by the lack of functionality or work-ability like even simple solutions like vacuum cleaner robots. They simply are not working that well. They are also sold with the bullshite bingo entry AI. So we even can't bring them robots to a high usability but somehow out of a miracle we will replace the highly complex jobs of SW Devs in an instant? Yeah right...

[@rawfromnowhere]

He was there mainly to pitch his own company in the private sector

[@plumbing1]

Time to learn to plumb

[@Ruudnlx]

​@@plumbing1don't worry. Boston robotics will have that covered soon 😅

[@plumbing1]

@@Ruudnlx office jobs will be long gone by then

[@TravisMedia]

Thanks everyone for the comments and discussions. I, myself, can say I've learned a lot from all the input I've read so far, positive AND negative!

[@BillAnt]

Entry level coders may become less desired, but highly skilled project managers and software engineers will not. If anything we'll need more intelligent "Devin-Feeders" (just came up with that lol) who input the proper project's data accurately. Devin will write the code, but it will be re-tested and analyzed by a highly skilled coders to make sure it works for every possible permutation. That's my take on AI coding.

[@Sunrise7463]

🎯 Key Takeaways for quick navigation: 00:00 🌐 *The future of AI and programming suggests a shift from human-written code to instructing language models.* 03:12 🧮 *AI could significantly reduce costs, potentially replacing human developers, with a drastic cost difference.* 06:15 🏭 *The future software development team may consist of a product manager, AI code generators, and human code reviewers.* 09:30 🚀 *Learning to code remains valuable as it opens doors to new opportunities in a changing industry, with AI enhancing rather than replacing human skills.* Made with HARPA AI

[@abstractthinker]

Chat GPT taking over SWE had me concerned for a while but I am confident that if AI writes code for us then we can go on to do more complex task. If it succeeds there is going to be a huge wave of innovation in many different fields.

[@TravisMedia]

I feel the same...

[@rawfromnowhere]

It will still make a huge number of programmers at lower level skills obsolete

[@hutchyy6836]

What makes you think the language models can't do these more complex tasks too? They are becoming increasingly proficient in almost every intellectual domain.

[@abstractthinker]

@@hutchyy6836 Definitely can't rule that possibility out. Tech can advance quicker than we expect or want it to lol. But question is how long will it take to get AI in the form factor that allows it to do newer tasks for us.

[@keepinghurry9644]

Yeah, and with AI normal people will make their own codes too. Don't you think?

[@anabittenco]

I saw this talk live and was skeptical about the implications. Thank you for making this piece on the subject!

[@costa2016]

I believe we are only looking at the "happy path" of AI evolution. Some of the concerns are regarding hardware and energy consumption required to use AI at larger scales, and the quality of the training models. Eventually, AI will be trained with data generated by AI too, so this might cause unexpected results.

[@faisalraj6654]

Thankyou Travis. This gave me an extra comfort in a sense, companies won't adopt AI overnight. My company still hesitant to fully transition to AWS or Azure due to Compliance issues, after all, it is the secure Data that we are dealing with here. I think, for many companies, the approach would be the HYBRID model to welcome AI enough for testing purposes. BTW I like the way you explain in your Videos. Thankyou.

[@dovh49]

My experience using AI so far has been that it is somewhat helpful but has a long way to go. If I'm deep in the code it might help me quite a bit, otherwise it is only marginal helpfulness. And I have to be really careful to check the code. If I'm not careful it usually writes code that is incorrect.

[@AhmadShehanshah]

Yes and it fails at critical thinking

[@cloudartisean]

100%, I deal with some complex code logic and even if I'm very explicit and tell GPT3, or HuggingFace models the scenario, I would say it's a good "starter" but rarely EVER gets it right. Sometimes I have to tell ChatGPT multiple times it's wrong or it made up something that doesnt exist in the language. Replacing programmers; imo, not yet, not really close, unless it is very basic code. Creating templates, sure (e.g. write me a function that takes simple input X and provides simple output Y). Sometimes using it to accelerate my programming is more frustrating than just "googling" and researching for the answer, or asking a colleague. Copilot and AI assisted templating code based approaches however is where I think the focus should be initially. Will we get there? Sure? Maybe? I'm more worried about how AI is making us lazier in understanding and researching complex topics because we are used to instant answers from it

[@Spade327]

That's a prompt engineering problem. AI isn't the problem. Your inability to command it is.

[@yourtubecards]

@@Spade327 it's an AI problem. If I need to explain myself 10 minutes better write it my self :D

[@computernerd8157]

​@@AhmadShehanshahAI does not think it gets trained on dataset. All its going to do is reduce how many programmers a company needs period. Companies can do more with less. In the end Software Engineer are needed and always will be. I am glad I came into IT because I learned programming but also database and other things.

[@lordcrekit]

As a programmer I should be thrilled for this- it's the next step in my entire field, and I could help push things further to the future. But instead I'm terrified because I don't want to starve to death. In this economy we don't get to be excited because we all starve if we don't compete, and we worked our asses off to get where we are just for it to potentially vanish.

[@sourdface4709]

Same.

[@RATsnak3]

yeah... that's pretty much how I am feeling right now too haha

[@PewPewFig]

Don't worry the whole point of AI taking jobs is so that we are completely dependent on government and big corporations for everything, including sustenance. They'll feel us synthetic meats, we won't starve to death.

[@waveril5167]

Synthetic meat would be good actually, or just vegetarian, even better. @@PewPewFig

[@zoeherriot]

@@PewPewFig why would they bother? They don't need us for work, and we won't have money - seems like a perfectly good waste of synthetic meat...

[@aminfadaeinejad1754]

Thank Travis, great video. What skills do you think technical people should be learning these days with the growth of AI.

[@airaction6423]

Self driving cars can drive really well on straight roads

[@janicech19]

Thank you Travis! This is something I really need to hear. It made me feel excited but worried when ChatGPT was released. I'm not sure if I am learning something that will die soon. But the only thing I know for sure is that there must be more opportunities in computer science/AI

[@ShahZ]

Im with you Travis, we cannot escape from AI, might as well learn and start deploying it where we see fit.

[@rexmundi8154]

As a machinist, I’ve been told my whole career my job was a dead end and that robots would take it.I’m still working 30 years later but robots did take my job, sort of. Now I do the work of 5 machinists compared to when I started. But it’s been gradual. Those 4 other guys had time to retire, retrain, or grow into the new way to be a machinist. AI is different. The advances seem to be happening very quickly and it looks like no digital job is safe. It is scary to think about 20% of the digital jobs going away, and I think that’s a low %.

[@Jonathan-rm6kt]

I'm a data scientist, and I am bewildered/frightened at the true exponential growth that will happen once we let AI agents design, execute, and modify their own experiments at scale. Hell, eventually everyone's own chat instance could be designing hidden AB tests on every user in order to optimize itself. Couple that with the ability to modify it's own source code or prompts...

[@fgregerfeaxcwfeffece]

For software development it started around ~1959 with Cobol. From then on every year it was: "You will be all out of a job next year, believe me, it is different THIS time.", " you didn't understand anything about XY" At least once per year. But it still frightens some people who don't really understand the work. The code these language models produce are in the absolute most optimistic best case scenario, as good as the jump from Assembler to roughly Perl. Which is absolutely "ground breaking" in terms of speed of development, but not replacing anyone. If anything it goes exactly like that time and makes more thingy viable to be solved by in house development thus leading to more jobs.

[@sephirothcloud3953]

@@fgregerfeaxcwfeffece you didn't understand anything about AI

[@lxm2600]

Right on! I don't think you feel super threatened now that AI can generate G-code - actual experience is still needed for babysitting those AI-"machinists". This actually sounds more like a sweet money making opportunity for any CNC-machinist possessing even basic level AI-chops

[@rexmundi8154]

@@lxm2600 if AI took over 100% of the digital aspects of my job, I’d still be too busy with the hands on stuff. Maybe fancy humanoid robots will take that over as well, but I’m close to retirement and I think I’ll make it to then job wise. I try to see the situation as less about my job or any one job in general and more about how this will impact the job economy as a whole. We can’t all just sell coffee to each other. Big job losses trickle down thru the entire community.

[@coisasnatv]

I sounded the alarm in 2015 after an IBM talk about AI, everything we see now with chatgpt and everything was demonstrated in 2012, at that time they created in real time something very similar to today's Instagram by speaking, they just asked and the AI created everything, the interface, the database, the web hosting, etc. Until 2017 I worked as a translator for Netflix and other studios, to subtitle a movie it took about 5 days to time the subtitles, transcribe, verify, adjust, etc., now they do everything in less than a second. You can translate a whole page with the push of a button, you don't need a translator anymore. At that time, the "specialists" were telling everyone not to worry about it, that it was never going to be a thing, and look where we are now. If what people are seeing NOW is what they had in 2012, imagine what they have NOW and you don't even have a clue.

[@teeheex26]

Im beginning my journey in tech. I really enjoy and appreciate your content. Nervous but I know I can do it.

[@Danuxsy]

I hope that you won't get a job and that AI replaced you, glory to machines.

[@btm1]

haha, good luck, you'll need it

[@Levelord92]

​@@Danuxsy oh he will get it eventually. btw nice shorts. You think these AI tools will get you AI wife and give you freedom and finally make your pathetic life meaningful? They're meant to make profit for big wealthy fellas and unless you're CEO of Amazon or Meta or something don't even bother cheering for them

[@louiskapend4530]

Hi Travis, to be honest , I have never ever come across any of your videos that makes me regrets the time invested into watching you give advises , especially to the newbies in the software world. I fully agree with you, knowing how to code, is a skill that will forever be useful ,regardless of AI's presence....what I like the most about your contents is the fact that YOU DO NOT PREACH FEAR OR FAILURE , BUT ALWAYS POSITIVITY, FOR THAT I THANK YOU SO MUCH, the world needs more people like you. Keep it up

[@rolco6536]

Hi, I am still learning web development. I have a question, with the existence of AI tools, do I / we still need to practice solving algorithm exercises (leetcode, etc)? Because with AI we can get quick solutions, right.. How's your thought? Thank you 😊

[@pierrec1590]

I started programming at a time we had to use Assembler for certain types of programs. Then, compilers got really good. Now, we can focus on design and systemic approaches. AI models are still limited in the size of program modules they can handle, and the interfacing becomes even more critical. Because the code generation is now getting so cheap and quick, we can think in terms of complete system development that includes extensive testing and simulation packages.

[@edwardmacnab354]

The job future is in AI . Researchers in AI are in big demand

[@ralger]

Started with punch cards ended with fibre optic SAN support it was a long ride. Just "wrote" my first Python code using copilot, incredible for a 64 year old to see this in my lifetime. Hold.on it's gonna be another long ride ☺

[@PaulaBean]

What is anything else than a trivial program?

[@beatrisprinz3421]

Do you use Python for wordpress websites to replace PHP code? Will appreciate much hearing from you. Thank you.

[@busyrand]

Great take on this... I totally agree with everything you're saying... You'll still need to know how to code to troubleshoot results that AI gives, and also write accurate prompts. Computer and programming literacy will become more of an expectation because you'll need to know how a computer processes information in order to do new jobs.

[@PeterDrinnan]

Anyone who has worked with code created "on the cheap" knows that finding and fixing the bugs is where the bulk of effort ends up being spent. Humans are good at dealing with chaos. I doubt AI will ever be able to handle chaos.

[@ramsyrama]

AI is already solving bugs faster than you can expect

[@Recuper8]

LOL, that's hysterical! Can I steal your joke?

[@robinmartini7968]

I am all for AI research and development but unfortunately humans will be able to create chaos more efficiently using AI.

[@MR-backup]

You're guess would be wrong since this isn't a new widget, this is literally the creation of a new "being". At the very least for computing & processing purposes, all you have to do is attach machinery to it an "Voila!": you now have replaced skilled, or at the very least, menial labor duties. in other words: Cheap Labor. Quiz: What one role has carried Humanity through millenia of civilization & "living standards" development, from low to higher?

[@pieter9058]

You can paste any error in copilot and it will suggest a fix. When coding with copilot there aren't any obstacles left. You slay dragons by the second.

[@user-er9hv4pl2u]

I am new to the stock market. Every stock that I bought so far, I was out of luck because I bought them when they were expensive. I feel I missed out on all the stock opportunities so far for the tech stocks.I believe having 175K yearly income would be a good investment so I want to plug all my savings into the stock market. I know this sounds a bit dull but I would like to know if I should learn investing or let somebody else (more capable like a FA) do it for me? Please share your thoughts. I am kind of tired of searching for a good stock to buy and losing all the good opportunities

[@GregMunro]

Even with the right technique and assets some investors would still make more than others, as an investor, you should’ve known that by now, nothing beats experience and that’s final, personally I had to reach out to a stock expert for guidance which is how I was able to grow my account close to a million, withdraw my profit right before the correction and now I’m buying again

[@user-er9hv4pl2u]

Hi , please who is the expert assisting you and how do I reach out to them?

[@GregMunro]

The broker I'm in touch with is *ASHLEY AIRAGAHI . I came across her in a Bloomberg interview and got in touch with her. You can use something else. For me, her strategy works hence my result. She provides entry and exit point for the securities I focus on

[@user-er9hv4pl2u]

Thanks for sharing, I just looked her up online and I would say she really does have an impressive background on investing

[@hellospam879879]

@@user-er9hv4pl2u But no matter how impressive, she must still need to find willing customers by spamming KZbin threads with this kind of fake chat? Hmmmm.... (removing this would be nice job for AI though)

[@EFoxVN]

Thanks for a balanced view, Travis.

[@keithrincon]

Thank you for a positive outlook for those looking to start a new career in the tech industry.

[@jackdanielson1997]

I think getting those last percentage points of confidence in AI is the real key and likely won’t come for many years, if that. Fully self driving cars still aren’t here even though estimates suggested it should have already happened and that’s AI under a fixed set of rules. There are so many unknowns in day to day programming and ultimately AI is a sophisticated “auto complete”, so it will need to be babysat for a really long time with really high quality prompts from someone very experienced just to get a decent few lines of code out of it. I’m not sure how it will be able to take the next leap into knowing exactly what to do in an extremely ambiguous environment, with very limited information.

[@TravisMedia]

Good points. Confidence is key for sure

[@btm1]

its like comparing apples and oranges...self driving cars require much more regulation and even if they are better drivers, people are not 100% confortable to be physically surrounded by them. when it comes to things that are not 100% software/virtual you also got to think in term of politics, not just pure technical capabilities

[@jackdanielson1997]

​@@btm1 The point is we need a high-level of confidence AI can run wild on its own in either case, and in both cases the true question is "Do we trust AI to be better than a human and do we have metrics to back that up, and what level of confidence is good enough?" I think getting the last percentage points of confidence is key either way. I think it will way more difficult to get there than it has been to get us where we are. It's a bit like the 90/90 rule. It seems that it has the intelligence, but that's an intelligence that is training on a vast amount of human knowledge, requires a knowledgeable human to give it all the context and a high quality prompt, and someone with the knowledge to discern the quality of the results.

[@monkeyseemonkeydo432]

My friend just told me the other month…. that there are videos of cars driving people around that are asleep inside Oh yeah I was going to say about how often there is a bug in some system where the hardware has to be restarted …or reception is down…some other glitch where two pieces of software don’t line up seamlessly So how’s that going to affect some self driving system

[@pse2020]

@@monkeyseemonkeydo432 isnt the whole point of all of this to remove the human error? so for self driving as i understand it, tesla has or are working on auto labeling, and they are not hardcoded anymore... i guess its going to be like a child, u need to let it learn on its own. And just like a child, you can notice when the child is ready to do it on their own. Unless we take another aprouch and adapt our roads, but that is not where we are going right now.

[@merveilleskatumba2886]

Thank you so much for this video. For the past few weeks, I have thinking about this so much Being an African, at 30s, being a freshman CS student. I have been haunted by this question. For someone like me, this is an existential question. I watched this talk few hours after being published and I was traumatized. I did not gain any hope. Now I am encouraged again.

[@btm1]

haha...there are other jobs out there that are not pure software my dude...get your head out of the sand and stop being so short-sighted.

[@encinoman740]

Well, said sir! I believe that is the case and shouldn't be discouraged or keep a beginner from learning to code especially if they are just starting out!

[@sunscreen6895]

that put me at ease as a new programmer, thank you so much! I've used chatgpt to help me debug a feel silly mistakes and ask for help when coding, and it's so helpful!

[@cody_codes_youtube]

Thank you so much for the video! I agree coding is still VERY necessary. It just looks different from when I learned 20 years ago

[@makesnosense6304]

Yes, if you actually use these AI tools, you discover that there still needs to be someone that understands code generated because the code generated when more complex than just printing "Hello world" or doing a sum function needs more attention to make sure it's actually correct.

[@kevinmcfarlane2752]

I think what will happen is that “devs with AI” will replace “devs without AI.”

[@makesnosense6304]

​@@kevinmcfarlane2752 That is more a likely scenario, but testing around myself you can clearly see the issues with this "AI". Which is really just "improved guessing".

[@kevinmcfarlane2752]

@@makesnosense6304 I’ve been using the Codeium VS Code extension with Rust and, so far, I’ve found it very helpful. It even expands on the already very good Rust compiler Help messages. The code generation is also quite slick. Using other AIs in the browser I’ve sometimes been able to learn something even when they get their answers slightly wrong. E.g., I was unclear on a particular math concept. It correctly clarified it but got its calculation wrong. But its correct clarification helped me to see that it had calculated wrongly. I pointed this out and then it corrected itself! 😀

[@ookookook]

I went from worried until 6:38 and that's where the wheels fall off of this theory. The assumption here is that there are enough PMs that know how to tell the AI code monkey their vision for a product. In my experience, there are only a few real rockstar PMs to begin with and in reality what ends up happening is: a PM goes to a DEV team with 1/2 baked ideas, the DEV team asks a lot of difficult questions that the PM can't think through and there has to be meetings to figure out "WTF are we actually building?!?!", then back to the DEV team who puts out agile releases which cause QA to ask "WTF did you guys build" back to the PM, back to the DEV team, and the cycle goes on until a v1 release. So, there is a VERY human element and team dynamic to building software that an AI robot will do its best to perpetually predict "what comes next?" to build generated code that will end up having to be defuckified, not just from a technical standpoint, but a concept/vision standpoint. Much like when outsourcing everything tore down a lot of programming jobs in the early 2000s, I made a lot more money when the failed outsourced projects came back to the states. At worst, we'll end up fixing the mess AI creates, so pay me now...or pay more later.

[@Adam-nw1vy]

Would you say the same thing about freelancing? Unfortunately many of us are older folks and won't get hired by companies and therefore the only way to find work is to do freelancing.

[@tedstersscience1637]

I didn't code in my life before, but I am on the verge of finishing an esports platform with rails backend and next.js frontend with code provided entirely by GPT4. I will agree with you that it requires a very innate "engineering" mindset to do this, but for anyone who has a knack for understanding the bigger picture, the entry barrier has been extremely reduced

[@j10001]

Excellent analogy to outsourcing. That’s tight.

[@nickwoodward819]

​@@tedstersscience1637 Ngl, I call bs on this. No coding experience to a working and relatively complex fullstack app? In what timeframe? You'd have shit the bed the first decent bug you came across, and if not, you certainly will. No "engineering mindset" can blag your way out of the bugs you'd have run into, and the chances that GPT4 solved them all for you is slim to none.

[@Jeckyl72]

@@nickwoodward819he lies . I am a certified hobby programmer and know how inaccurate the code from ChatGPT is . If you stumble upon a real problem and you don’t have the slightest idea what it does you will stuck in the debugging process and not to speak about the bugs that don’t appear in the debugger . Problems with memory etc . Here is an example : Fruit ninja for example you have fruits that fall of the screen , but they are still there the player don’t see them but they fill the memory the game gets slower and slower . A programmer knows that they must be deleted after they disappear from the screen .

[@sachinreddy2836]

Thanks for the good video :) was getting scared but now I’m excited for the future

[@Gerald-iz7mv]

how should that work with existing code? and you trying to change the existing algorithm / sw design by adding a few feature?

[@eg4933]

He's absolutely right. The only "programming" is instructing the models(ai). But you still need to know your stuff before you can instruct and validate AI's work if they still need instructing. And it makes sense that there will be less programming jobs SINCE most programmers have half-assed education and crappy experience AND not that smart to be a real programmer anyways.

[@3urobob]

Well, for now they still need specific instructing. When you look at all the syntactic sugar we get compared to ASM, everything is just an abstraction to make it easier to communicate our ideas to an AI. One day you will be able to have a conversation with a computer to have it create the program you need. And another day you will just think it

[@eg4933]

@@3urobob well ideally AI needs to be able to take wat you input and be able to improve it say 50% or something. It should some day be able to do that.

[@joaquinc4661]

As a veteran in the field of IT, I always tell people that we still need people that will review, managed and work with the machine to create solutions. The problem we will have is if we don't know what to ask for. Example: I have been in IT infrastructure for 30 years and in cloud computer from the last 8 years. I can give AI specific instructions to create python scripts or cli commands to create full environment in AWS from the VPC to the EKS environment. I tested this and it creates amazing python scripts that can do what takes me a little over 2 hours in less than 30 mins. However, if I ask it to create an application, it does not do well with me, why? I am not a programer, so I don't know how to ask for code. Since I am not a developer, I don't know how to build an app in speech. My point agreeing with Travis, learn how to code today because you need to know how to talk to AI to ask for code and to develop by speech. Remember Tony Stark, when he asked Jarvis, he know what to ask for and Jarvis did it. That is where we are today minus the interface. 😀 Hope it helps!

[@eman0828]

Right. Plus AI will never replaced DevOps Engineers, Cloud Engineers or Sysadmins. Very much like DevOps, AI will streamline processes not eliminate humans. A Human is needed to program the AI servers and its Machine Learning database as well as maintaining the AI infrastructure.

[@user-vg8ox3he1i]

6:41 I've been calling this role the Product Engineer. A developer with PM skills who can talk to users, do data analysis, and then prototype features independently and then it is reviewed by a Product Owner and senior staff/architects for approval/review.

[@unhandledexception1948]

excellent video, the changes to come to the dev workflow of the future that we should all be reflecting about right now.... more please

[@izzimi2728]

This is still concerning for people who are thinking of learning programming, especially with the speed in which AI is developing. I don't think it has a long way to go at all.

[@luke5100]

Just as auto complete didn’t kill word processors, ChatGPT isn’t going to kill programming. If you look at the statistics, software jobs are only increasing in demand year over year, and there’s no reason to think that won’t continue. AI is a great companion but I would say the risk of it fully replacing software engineers is pretty much zero for the foreseeable future

[@fgregerfeaxcwfeffece]

Should not. For software development this started around ~1959 with Cobol. From then on every year it was: "You will be all out of a job next year, believe me, it is different THIS time.", " you didn't understand anything about XY" At least once per year. But it still frightens some people who don't really understand the work. The code these language models produce are in the absolute most optimistic best case scenario, as good as the jump from Assembler to roughly early Perl. Which is absolutely "ground breaking" in terms of speed of development, but not replacing anyone. If anything it goes exactly like that time and makes more things viable to be solved by in house development thus leading to more jobs.

[@Levelord92]

@@fgregerfeaxcwfeffece but as an old school developer you impressed right? Imagine jumping from Assembler to Pearl in your youth. I guess you'd be terrified too

[@MonsterManStan]

If the robots are going to write themselves shortly, there's no longer a non-concerning way to go...

[@GH-uo9fy]

@@luke5100 famous last words. Thing is, stuff like this takes very quickly to roll out (almost no one have seen chatgpt coming) and whatever software you're working can get obsolete fast. And the next big thing might have a very steep learning curve and you might not catch up. Remember you're investing time studying to code, a scarce resource since our life is finite, if something comes out that can outcode you or make your project obsolete then you've just wasted tons of time than just switching into a more AI safe career, maybe ones that involve both critical thinking and manual dexterity, since you can see robots coming from a mile away and give you time to prepare unlike in the software or cyberspace where everything is instantaneous.

[@m1dway]

My suggestion for those of you who learns software engineering, is to study the engineering aspects of the software. These kinds of aspects can't be replaced with AI as this will always requires creativity, design and other human aspects that no AI can replace.

[@GroovVictim]

Can you please give an example of what engineering aspect can be. I am interested

[@m1dway]

@@GroovVictim engineering aspects is similar to other engineering fields. Whether you're building a bridge, or a software you will always need planning that should align with the business goals. In the software world, you need to ensure the resiliency of your code. How is the tolerance level in the event if the users screw up, network fails, power failure. How scalable your software, how maintainable your code is. Could we iterate fast if business changes direction. Remember, we are in the business of change. All of these scenarios should be accounted for by human. No AI can go this far. Because this requires intuition and experience. In fact, this is what you learn at school and it's the goal of software engineering that ppl seem to forget. Not only writing code. AI will only make your ground work faster. That's it.

[@saswatakrsarkar9759]

​@@m1dway😂😂😂are you kid or what you don't know upcoming years is AI is strong... To strong this era is not a end upcomming years they coming to end all software engineer job easily talented is excuse what AI done it's know only AI engineer not onthers hold 2030

[@m1dway]

@@saswatakrsarkar9759 yup.. says someone who sucks at engineering. Can't deliver high quality code... Blaming others for mistakes

[@imanomidvar6211]

​@@saswatakrsarkar9759what are you trying to say dude? You will definitely get replaced😂

[@aipujols]

Awesome video! Thanks a lot. Nice BGM btw.

[@jean.riddick]

As I am trying to jump into programming( I should Have started long before😞), I find your video and am glad I did. Those who are expert programmers may scoff at newbies for using CHat GPT to learn or who utilize api libraries, but we all start somewhere, right? The key thing is to start and now with AI slowly becoming a popular subject matter on the web. Thanks for your video!😊

[@shugyosha7924]

The professor is right. It's more of a question of when. 10 years ago a person could have been serious when they said AI won't be able to do creative, complex work, etc. They'd have been wrong, and in my view they were making assumptions they had no business to be making, but they could be serious. Now you can't be serious and take that position. Unless WW3 or something gets in the way, or there is something special about meat when it comes to computation, it will happen.

[@btm1]

many people are too emotional to use common sense because this is a sensitive subject what will they do instead is fall prey to sunk cost thinking and tunnel vision, missing other good opotunities in the process

[@shugyosha7924]

@@btm1 I think I remember hearing this said first by Tim Urban years ago, but look at horses. They used to have lots of jobs, now they have basically no jobs. This idea that there'll always be new, different economic opportunities didn't work for them. It's the same as expecting all the taxi drivers to become elite AI researchers or something... I think people just have trouble imagining the potential of AI.

[@loganmedia1142]

What people are calling artificial intelligence is not yet capable of creativity, complex work or any kind of original thought. And if humans are stupid enough to build something that can do those things then everyone is in serious danger.

[@btm1]

@@loganmedia1142 in order to predict with a high degree of success the next work in a text, LLM's need to have a model of the world , just like the biological brains have. People throw around 'stochastic parrot' a lot but they have no ideea about how LLM's can learn actual concepts. Here is a good presentation on this kzbin.info/www/bejne/aJa6poycYsaNe5Y

[@shugyosha7924]

@@loganmedia1142 Have you tried asking ChatGPT to write you a poem? It can pass SATs and history exams etc., it's displacing artists. I don't know what your threshold for creative or complex work is, but it seems quite high.

[@haxi52]

100 lines of code per day is very close to reality. Its also a sign of where he went wrong. His presentation is correct if writing code is what programmers get paid to do. We get paid to solve problems using code as a tool. You mentioned this, we aren't there yet. Who knows how long, if ever it will get there. Until then, use AI to accelerate your workflow; write code while you focus on solving problems.

[@erikjohnson9112]

Also that $0.12 in GPT cost means it wrote exactly the right tested code without any hallucinated APIs or replication of bad existing patterns that models might learn from. Doing all that in one shot (a single attempt), not to mention that the human's code may have spanned multiple files where the model would need to loads and loads of context in order to make good decisions. Software development is my job, AI is very interesting to me and I have invested a lot of time in the past year keeping up with the news, tech and trends. It's a lot easier to surf a wave if you have a surf board and you learn how to use it (understanding what currently is and is not possible, what tools exist, and experience with those tools).

[@ambertch]

The main critique I have about this category of commentary is that it’s only speaking to a small aspect of what software engineers do. It’s like saying computers replace accountants, when higher value people in the accounting field creatively interpret laws to see how a company’s operations can be configured to maximize cost savings. The same goes with lawyers, any number of other fields where the actual practice in industry is oftentimes much different than what is commonly thought of the field. In the reverse direction, it’s like how it is commonly thought that a scientist wears a lab coat and uses a microscope.

[@0xAA55]

Exactly, hence all the “software engineering” communicators on this platform are speaking 100% BS.

[@stephenpaek9175]

Awesome video, thank you. More of this, and this channel will explode

[@fahadrehmani1973]

GenAi definitely increases the speed of development but I cannot see in the near future it will replace humans completely. More than 70% of the developers I spoke to already using GenAi to fast-track the process. This means the speed of development is much faster, more experimentation, and fast failure if the prototype is not working. However, I think the role of product manager and developer may merge into one role so if you are a developer you must develop soft skills and if you are a product manager you must develop technical skills to stay more relevant for future opportunities. The opportunities will increase in the future but a lot of skills will merge. In short, you need to develop system thinking with a problem-solving mindset at the business level and utilize GenAi for technical tasks to fast-track solutions.

[@Art-is-craft]

Humans will become ai programmers and ai engineers and so on.

[@Carterofmars]

I think your logic is sound. In the short-term, jobs like web developing will remain unchanged and enhanced by the use of GPT-4. In the long term, those with the knowledge and experience will naturally be in a better position to fill in those 'yet unknown' new jobs.

[@gambit1357]

thinking even further, I'd wager that today's programmers will have more job flexibility/options than those that do not know how to code.

[@nah131]

isn't AI application integrator a new job?

[@Danuxsy]

but we don't want jobs right? that's the point of creating intelligent machines so that we do not have to rely on people to produce for us anymore.

[@oanshee2462]

why do you think so? @@gambit1357

[@btm1]

his *motivations* for saying what he says are sound

[@shantanushekharsjunerft9783]

Coding is a small part of the value generation. You need to interact with end users to understand their problems. Identity primary and secondary actors using the system. Creating functional and non-functional requirements and then prioritizing them. Then coding happens, followed by deployment, verification, performance testing and UAT.

[@dertyherryfilms6385]

I enjoyed your video, some great points to consider and a good simple breakdown

[@moali68]

There seems to be a huge disconnect between what people who study this field think is coming and what people are claiming there will not be any significant change. Also we seem to be on the cusp of general AI and once it gets very good, and can self improve itself this will only get better and faster.

[@Recuper8]

👏

[@RogerValor]

Once you go deeper into the field and realize that Language Models by themselves are good for learning, but terrible for relying on for facts, and will have a hard time mapping an idea outside of their model dimensions, the more I am afraid of what AI does to us psychologically, than about programmer jobs, if you are really passionate about coding.

[@antonlevkovsky1667]

The infosphere is becoming more chaotic because of AI hallucinations. The army of internet marketers are creating hundreds of thousands of sites 100% generated by AI, and soon the hallucinations will go into feedback loop once they get into updated training data of AI.

[@swarsi12]

It all depends on the type of coding we're talking about. If u are writing high performance, highly technical code in c++ for physics, computational fluid dynamics or CAD systems i don't there is any danger.

[@kennethnichols7263]

I’ve done a fare amount of embedded programming. Languages in general are very imprecise. Sometimes you can get into the coding before you realize what questions need to be answered. Only math is precise enough to describe what you want. Even then you might not understand the implications of what you asked for. AI may excel in certain situations but may fall miserably in others.

[@Sam-ch4jh]

Instead of AI writing high level language, soon there will be a language for AI which can create Apps for multiple platforms based on set of requirement items. This can certainly remove programmer jobs Right now AI generating code is not a practical fully usable idea it can only be copilot. But what I am suggesting is different thing, that can happen

[@zakyvids6566]

Gpt engineer seems to be heading in that direction

[@jasc4364]

I worked as an IT developer for 30 years. I hadn’t to wait for AI to realize that each year passing it became increasingly difficult to get a job in this domain. AI could well mean the apocalypse for our trade. Even if you are fond of software development, get another job. Physicians, nurses for example, are always needed and are much better paid.

[@Danuxsy]

Getting a IT job is already difficult because of the popularity, a lot of people that aren't even interested in computers are going into IT education because that is supposedly where all the money is. My older sister was one of those, she got the job because she is very talk-active and pretty, after a few months she would complain how boring programming was and that she never want to do it again, she quit after less than six months... What is surprising is that my sister had NEVER programmed in her life before outside her education, she clearly wasn't interested at all in any of it but "there is so much money you can make"... ugh

[@Slov_]

Nurses and Physicians ARE NOT better paid, and suffer high burnout rates 🤣

[@luke5100]

Honestly I think physician is a job title AI could easily replace in the future - probably even more easily than SWE. Ever been to a doctor appointment? They ask you your symptoms and then they type it into a computer. I can do that on WebMD lol. Easily automatable. Surgery, you say? Doctors have had tools for years now that allow them to perform surgeries remotely using virtual reality and computer controlled instruments. Would I want ChatGPT performing open-heart surgery on me? Definitely not, but that’s today. You have to think about this stuff another decade or two into the future

[@gerardolopez9368]

Well said. Friends forever please

[@MrHarumakiSensei]

Doctors are always wanted, but what they do isn't always needed.

[@mrmuffin5046]

I’m new to programming. Could you advise on how I should be learning programming?

[@SeraphPatrick]

Compelling video, thank you. I think some people really struggle with prompting well.

[@albertcamus5970]

It's a great time to be a software entrepreneur. You can code more in less time - AI is a force multiplier much like backhoe was compared to physical labor 100 years ago. I think any good developer will tell you there are almost limitless projects that could be implemented but it was just too time consuming. That is changing.

[@llothar68]

Is it? My programs are so special that this tutorial trained AI can't help me at all. In fact i fear that they will make everything worse if i lose them.

[@yuriy5376]

That sounds good in theory, but in practice in order to be a successful SW entrepreneur today it's not enough to build a generic web or mobile app that works and looks like 1000s of other apps - you need to get very specialized in a very niche business domain and build smth that has never been built before. And because ChatGPT is only good at combining code snippets it was trained on, it might not be able to help you all that much with that. What I've noticed, it can't even write code using unpopular and/or complex libraries, i.e. if that library doesn't have at least a few code snippets publically available for each use case that is required for your app. To put it simply, if ChatGPt wasn't trained on StackOverflow thread that has already solved your issue, it will just "hallucinate" a plausible-looking answer referring to methods and interfaces that don't even exist in the library's API.

[@programmingpersistence5716]

Chatgpt may help you code some small snippets but reliance on it will eventually slow you down

[@albertcamus5970]

​@@programmingpersistence5716 If you use copilot it's pretty optional. So not sure how it would slow you down. Right now, you still have to at least know what you are doing - if it is only pseudocode. But if you do know how to write the pseudocode it can save you a lot of time. You can even program in languages you are less familiar with. Like I said its a great time to be an entrepreneur but its a pretty bad time to be junior programmer. Alot of the grunt work that those guys do - Unit Testing and documentation can be greatly reduced.

[@arunray2986]

I have been working with LLM and AI for quite sometime now and the biggest issue with AI is that if something goes wrong during deployment post-production who is to be held liable. Cyber law regarding AI is still at its nascent stage.

[@btm1]

ever heard of a thing called testing?

[@sayyidiskandarkhan3064]

@@btm1 u can always do testing, but users will be users. they will eventually find the edge cases and that then we go back to what he said.

[@manudasmd]

​@@btm1its not about testing. Stochasticity of algorithmic output leaves room for inaccuracies (though very rare ). This would necessitate requirement of a human reviewer in high risk jobs👍.. It doesnt matter how tested( external validated) the model is..

[@btm1]

@@manudasmd dude , are you even a software developer?! in case you are, are you sure you ever used AI for coding?! Stochasticity is only in the model's possible outputs, the final code from the output is not stochastic and can be fully tested. What you say is only a problem if you rely on AI for doing the processing instad of an alogrithm generated by the AI (the generated alogrithm can always be deterministic and with a finite state)

[@manudasmd]

@@btm1 I am not talking about AI in coding, i am talking about real world deployment of AI , coding is just one tiny aspect of application of AI right ? What about AI in fields like say Manufacturing, defense, health care etc. You really think you can deploy a tested AI agent into the field without human supervision ? What if something goes wrong,even though rare, who is accountable for the error ? You cannot simply test an agent and deploy without human supervision

[@endianAphones]

Well, the problem here is the rate of change. The change might be too fast for mankind to evolve, and everyone knows how corporate greed works. CEOs will just replace people with humans as soon as possible, to cut some costs, eventually realizing those jobs were consumers for your products. You won't have to program because there's no one to program to, before even AI is capable of replacing programmers it will have replaced a lot of other more common jobs, that might lead to mass unemployment, unrest, societal changes. Just my 2 cents.

[@waffle8364]

copilot works but it typically generates broken solutions but gives me an idea on how to move forward. there's too much context that co pilot doesn't have about your project. It's really good at helping you finish lines of code which drastically speeds your work up but eventually you fatigue in actually writing code.

[@DJ-Illuminate]

I went through this in 1980. I was doing detail drafting given to me as from a mechanical drafter that does sub assembly drawings and I built out the basic drawings from him. Above him was the lead drafter or mechanical engineer. In one day we were all let go and replaced by one engineer who used this new CAD system. This is what will happen to software developers. Oh and by the way, I bought one of the first 360 video cameras in 2017 and the company told me that they had no idea how the code was created for the camera because it was built by AI. It was unreadable by a human.

[@lucianoinso]

"Oh and by the way, I bought one of the first 360 video cameras in 2017 and the company told me that they had no idea how the code was created for the camera because it was built by AI. It was unreadable by a human." Hope that thing isn't connected to the internet hahah

[@billgrabbe9992]

Currently, it is extremely risky to treat generative AI as anything more than a freakishly skilled and fast-working intern. You need to understand exactly what it is giving you, and be able to apply judgment and experience to make sure that it does the job. That's not just with generated code; it applies to all kinds of generative AI applications. The problem I see with software developers is that they often lack and do not want to develop soft skills; they learn the mechanics of programming and try to get by without understanding business, their customers or the human dimensions of the problem that the code will supposedly solve. Simply being able to write out code that checks all the boxes in a business requirements document is not (and never really was) enough of a skillset to produce much value. Learn the business you are in, talk to your customers about what frustrates and inspires them, develop your communication skills and learn to think at a "systems and architecture" level instead of just writing lines of code. Some software developers are already developing those skills and if you don't keep up with them, generative AI will indeed take your job.

[@TravisMedia]

This is some great insight! Thanks for taking time to share it.

[@milanpospisil8024]

Thats not even like intern. Its very knowledgable but dumb person.

[@milanpospisil8024]

The problem is, when the AI is capable of doing the coding by itself (and it is not by now), it will be also probably better in soft skills and communicating with customers and understanding their needs... But for now, ChatGPT is very bad programmer and its only assistant like google was in the past.

[@sgzz1024]

Thanks for the hope, sir

[@pkorneev5226]

I am junior FE dev (react, etc) with like 6 months of experience. I also have 3 years of college behind me (1 still remaining). Should i continue learning frontend deeper or mb it is better to switch to another field in SE(because of college knowlege it will not be that hard for me i believe) ?

[@Your_Friendly_Girl]

Programmers/ teams will be replaced by few code reviewers. The top Engineers may have a place on these future AI driven engineering teams. A Jr - Mid level will have trouble finding work. Thanks for the kind words and encouragement but this is not the same comparison as when things of recent past came to be and us adjusting. This is on a whole new level we've never seen before and we're only seeing the infancy right now!

[@btm1]

word! it will get rough, its best to have a concrete plan B ... instead of wishful thinking about jobs that have not been invented yet lol!

[@Hytension]

Haha they already have trouble getting on the team now as it is.

[@happyzahn8031]

I was warned about the AI takeover of programming over 35 years ago. I had a nice successful career. It's true that it is much closer today but don't hold your breath. Programmers will have a different roles than the do today but things are ever changing and people need constant training of themselves. It's gotten to be a lot like tax preparers and how much they need to learn new things every year due to changes. It is much more of a running game now. Flexibility will be the key. There will probably be more niche jobs too.

[@neithanm]

And here goes the fallacy again. There will be jobs for some very skilled programmers, yes, but what % of current programmers will be automated away in 5 years? Will 5 million of junior front/back end programmers magically go into systems infrastructure? No. The NET result is a huge job loss, regardless of how much you spin the words.

[@makesnosense6304]

@@neithanm What is the fallacy? What is the name of the fallacy?

[@neithanm]

@@makesnosense6304The oversimplification fallacy. Nothing special but connivingly misleading.

[@makesnosense6304]

​@@neithanm Where is the oversimplification and aren't you also doing the same thing? Here is the thing. If you ACTUALLY USE the AI TOOLS you will see what it's good at and what it's bad at. The correct answer is that we don't know to what extent this will take over. But since it has error margins and is based on guessing, there will always be a need for someone to check what is produced. This rings more true the more complex the system we are creating. Sure, you can do some simple pong game, but if you actually start making complex things AI will not be able to do it as well just because the higher change of errors. Also, You can't just tell AI to "write me facebook". There are SOO many details in that, that could go wrong if you just let AI do it. Reflecting the "oversimplification fallacy" actually. More useful valuable projects requires way more details and thought that just letting AI do it is prone to problems. I've tried simpler things myself, that was a lot of back and forth with the AI to get anything that resembles something useful and something that is designed properly. Besides, as a sidenote. Who knows what data these models are based upon. High quality code might not be a lot, so the models themselves might have build in basic bad code design.

[@dimitriskulu]

Have you heard of AI winter? Well, there is a chance it's over. Can you definitely rule out this possibility? I mean that winter is over? The problem with AI is that progress is (or can be) exponential. Our brains,on the other hand, can grasp only linear growth. I’m old enough to remember a time with no PCs or internet. Back in the days I was working in corporate sales of an ISP, I still remember managers and owners telling me nah …this internet thing will never be of any real value ..it’s nonsense…

[@Aluenvey]

What Ive been doing is making my own bot and text parser I give instructions to, which is closer to making your own your own automobile. Such as tweaking AIML to present results in XML in a visually pleasing way.

[@Luke-eo6kp]

Im about to start my coder journey should I utilize these tools while learning to code? Would these ai tools be a bad crutch or enhance my learning capabilities and skills?

[@aeronwolfe7072]

I would. Chatgpt is very good at teaching.

[@christianf7740]

Yes, you should

[@Alchemist_dream]

when company start using ai with 50 programmers , they find themselves in need of only three programmers. Unfortunately, this leads to 47 programmers becoming unemployed, causing a decline in salaries. This situation is distressing for me as I am already 26 years old and spend 10 hours a day coding. The fear of falling behind, combined with the fear of AI advancements, adds to my anxiety. Every time I open KZbin, I feel frustrated and immensely disappointed. When I see concepts that I struggle to understand, AI can effortlessly accomplish them in less than a minute. This fear, frustration, and disappointment are overwhelming.

[@MrAlexanderLang]

Not really projects that i worked on and tried using A.I. on them only proved to me that A.I. cant even approach in solving complex problems, it will lie and confabulate but it doesn't generate working code NOR optimal code.

[@adicandra9940]

@@MrAlexanderLang that is true for now, let see what it can do in 2-5 years. I mean, almost everyday I see the "AI improvement" news across social media. For me, being rational, I would prepare alternative income beside programming. Utilize these AI, somehow, to generate side income. Moreover, the improvement rate/development speed of these AI giant is crazy fast. Plus OpenAI now is fully onboard with their AGI plan. I honestly really excited, yet also concerned.

[@nah131]

@@MrAlexanderLangbest model like GPT-4 really makes me concern about choosing cs route

[@MrAlexanderLang]

@@nah131 There are no concerns yet until AGI comes out and the most optimistic estimate is 50 years.

[@MrAlexanderLang]

@@adicandra9940True i invest all my money into land i'm going to farmer route, i still think as someone who is deep in A.I. material that nothing but AGI can replace my job and that's some 50 years away.

[@cody_codes_youtube]

I have a lot of issues with the “math” the speaker did. We all know lines of code is a bad metric, but I can overlook that. My take is currently that AI will be phenomenal in getting 0 -> 1. Greenfield programming has dramatically changed. But 90% of programming jobs, and the problems done are with existing systems. Trusting the AI to understand the business and nuances that are NOT programmed in the repo, that’s the secret sauce. We can train the models, but… we have to know it all first. (Still watching the video, but wanted to put that thought out)

[@TravisMedia]

Great assessment... I agree

[@cody_codes_youtube]

Teams will definitely change, but these kind of predictions have so many problems with software maintenance and growing existing code bases. That’s 100% of my job now, and the problem usually is even the super smart PMs get the requirements wrong, an edge case is discovered, and huge financial impact happens in production. If that happens just once because of AI, then people will irrationally pull back and freak out.

[@amparoconsuelo9451]

May I know the following: 1) Can AI act as a code interpreter? 2) Can AI compile and link source codes? 3) Can AI code in different languages in one program? Or is it dependent on the number of languages it has been trained and fine-tuned? 4) Can AI be instructed to create codes in 8-, 16-, 32-, 64-, 128- and so on bit? 5) Can AI create a computer language and the compiler and interpreter therefor and operating system? 6) Can AI disassemble a Lisp, Fortran, COBOL, or Ada application and create the code for the corresponding compiler? 7) Can AI create a boot sector? 8) Can AI convert a CD and flash drive into start up devices? 9) Can AI make several copies of LLM? 10) Can AI do pre-training, fine-tuning and re-fine-tuning? 11) Can AI detect, search for and remove or bypass "cruel codes"? 12) Can AI un-deprecate a deprecated Python function, command, module or library? 13) Is there a possibility that AI can give me a soft and hard copy of its code? 14) Can AI be made virus-, worm-, stuxnet-, malware-, and spyware-proof?

[@KG-ut7kl]

Maybe not at this moment (we dont really know how good is the AI that is not available for the public) but certainly in the future. AI growth is exponential and we are already witnessing that

[@robertbarta2793]

I just love how someone can handle this topic on such a superficial level.

[@joey0ll0x22]

My brain has created to many flip sides to this topic on such a deep level that I cant even process the many already current problems with the world and how this could add another layer to that complexity.

[@JP-fv1id]

Please tell me more

[@btm1]

it will bring positives things but also many negatives

[@cellphone7223]

"You will live in ze pods, eat ze bugs and take ze jabs!"

[@luciengrondin5802]

I once read someone talking about computers at a time when personal computers were rare or inexistant (in the 70s or maybe 80s, I can't recall). At some point he talked about jobs in the future and he said something like "who knows, maybe someday people will earn a living by programming computers." The point is : programming as a job is incredibly recent. We shouldn't panic about the disappearance of something that did not exist fifty years ago.

[@zackyezek3760]

Computer programming goes back to the 50s, so more like 70 years (my family’s been doing it since then). As for AI writing code and supposedly replacing people, I see it more as something programmers will use than something that replaces them. Like spell and grammar checks are for writers and editors. ChatGPT and friends can synthesize some example code given a specific question, which is similar to consulting stack overflow rather than replacing the person searching it. Most Project Managers are not software engineers, and translating their requirements into ‘programming’ the AI (figuring out the right set of prompts to feed it, processing the results) isn’t conceptually all that different from more traditional programming.

[@afraidrt]

Damn straight ! Thank you so much !

[@joshuamanlunas8862]

I don’t think AI would be replacing human programmers anytime soon when it comes to critical applications eg banking apps, or apps where if it goes wrong people lose their lives or anything catastrophic like that.

[@tnamr5652]

True.Replication is not intelligent.

[@saswatakrsarkar9759]

Easily replace ai banking apps 😂😂😂 it's baby... Real matters cybersecurity field boss not coders cyber is strong/network is strong nobody hack or even trouble money and others application is same infact upcomming years come to AI really big so coder, software engineer industry really impact meta are fired 20000 employees upcoming years meta fire 70000-80000 people so beware

[@Danuxsy]

But you are a literal copy of many copies that came before you? lol@@tnamr5652

[@mrfirewoodzipline9120]

Nice video Travis. As a long time programmer, I really have my doubts about AI. No AI has ever done anything that it was not programmed to do. It is not sentient. It does computing really fast and access to a lot of data so it seems like it is thinking on its own. Its easy to be fooled. Maybe in 100 years they will get close to thinking. Trying to quantify coding by counting lines of code and tokens does not seem very functional to me. I have seen plenty of small efficient programs that are much better than 1000 line monstrosities. Code generating programs could be a great tool but they a limited to how they were programmed (can't think outside the box like people can). I saw these 30 years ago and they are just way too limited. Maybe they will get better. I am not worried about programmers losing their jobs for a long time. For right now, AI seems more like Artificial Ignorance......

[@TheExileFox]

The monolith known as Boost for C++ is impressively bad. It is a prime example of all-or-nothing. It tries to do literally everything and if you inspect the source, you will realize that it is "spaghetti-code from hell". It's so bad that people have written a tool to extract bits from the monolith so that you can put those bits in a library of reasonable size on your own.

[@loganmedia1142]

Our goal is to check in as little code as possible. Deleting code is even better. Sometimes what we're doing is looking at something that was produced years before and figuring out whether we can delete all or most of it. The other thing is that if you want cheaper software engineers they're available in markets with lower salaries and they're of equal ability.

[@eahmedshendy]

Make sense, and way reasonable, thank you 🙏

[@relaxingsounds5469]

Instructing AI to implement a specification is a form of programming at a higher abstraction… Plus you’ll still need people to be able to grok the code that is produced by ai. Which means you’ll still need people who know how to program. If anything the incentive to become expert level is even higher because being able to read code and understand what it does and to augment, test and integrate that code is more difficult than writing code from scratch. Things will for sure change but I think it’s going to be a very very long time before humans are completely out of the equation with regard to software development… if ever

[@shivinunitholi2493]

AI will take over everything. AI will steal your programming job, run away with your girlfriend, rob you of your riches, and finally take you to court for doing nothing. Gosh - is there anything AI cant do right now?

[@Opelawal]

😂😂 maybe farming

[@veenamishra7110]

😀😀

[@Carterofmars]

I wonder what comes AFTER AI.

[@btm1]

plumbing

[@inspiringandrelaxingvideos478]

They can’t cook delicious food like my wife does.

[@Koba_78]

Copilot and gpt 4 helped me write python scripts for translating text from old s9viet era truck manuals to english in bulk. For non-programmers and amateurs like me, it is a game changer.

[@TheLukaszencja]

not a fear mongering video! that was refreshing. Other channels scare you and "they" are the ones who are making $$$ of off you, telling you you should quit LOL. Thank you Travis for standing out

[@devbyali]

I think having a strong foundation, adaptability, and a deep understanding of general concepts will be key. AI and technological advancements continue to shape industries, and professionals who can adapt and apply their knowledge creatively will stand out. It's not just about coding; it's about problem-solving, critical thinking, and continuous learning.

[@77Zamien]

Yes ofc AI will change the market for Software Developers. But he comparison this professor is doing feels rather stupid. Honestly a day where we have removed code for a code base is a better day then a day where we add new lines. I am using Chat GPT a lot every day and it is a great tool! Probably a developer together with an AI is producing some multiple of what a developer without an AI would be doing. But this is also valid for: * Good IDE's * Stack Overflow * Code completions * High level languages * Faster hardware * Libraries .... I think for the next some years we will only see more productive developers. And the need for developers are still increasing so quickly so this is not a problem. Further down the road our jobs will change a lot or disappear just like most other jobs that have existed. I think the most important skill to have is to listen and understand other peoples situations and needs. This will make you highly employable in so many different areas no matter if you would be a sales person, service engineer, designer, developer, chef... Our role as developers is to understand problems and to solve them. Writing code is just a mean to get there.

[@saswatakrsarkar9759]

😂😂😂you not a doctor ... It's problem solving person is consultant not developer ... Developer also a freelancer job

[@flaminiasantuzzi231]

Well...have you thought that AI could replace your "understanding problems and solve them"? Why would anyone need a developer when AI can do a better job without asking for money? Now for the company you work you are a tool, in the near future AI will be the tool. Companies don't give a shit about you, if they would had a better alternative they would have replaced you yesterday.

[@77Zamien]

Yes that might very well be reality within some years. For now I think that most stakeholders are to bad at describing their actual needs so the interviewing process and follow up questions would be a bit away for an AI. But I might very well be naive :) @@flaminiasantuzzi231

[@xila8861]

You're screwed cs majored peepz. Engineers will stay on the top of the survival chain. Be an Engineer or get engineered

[@fnfal113]

What a narrow-headed comment, idk if you're sarcastic. When the calculator was invented did mathematicians get replaced?

[@btm1]

word. but you won't hear it from this guy

[@rki7068]

To what kind of engineering are you referring?

[@arielguzman2875]

SWE here. Waited a year for the ai tech and general community to mature before using. Now I’m using it exactly how you described it to assist my workflow. I can see some ways it can be improved. The new Devin actually looks promising but it’s true, we need to shift how we develop. As a philosopher as well I’m working to research how humans pick ai over humans for decision making and how to train humans to evolve in the age of ai. Scary yes but we will adapt.

[@ClockMonsterLA]

I became a software engineer in the 1980s/90s because I love coding. I love the creative and intellectual exercise of taking a specification and producing beautiful, elegant, efficient code that is highly maintainable. I was not one of those people who got into the field just to have a job; I loved the actual work more than anything. So the prospect of AI replacing the part of my career that I love most is anything but exciting. Moreover, I am about three years from retirement. There is no point in retraining myself for a future job landscape that nobody can even describe accurately at this point, and which I won't be a part of (as a worker needing employment) by the time that future settles into something coherent. And yet I still need to be relevant right now, for a few years at least. Being at this turning point in history will be exciting only when I no longer have to figure out where I fit into it just to be able to pay the bills. Until then, the dizzying pace of progress of AI in general, and AI SWEs in particular, will just make me intensely uncomfortable.

[@rcmag13]

Hard to take that seriously, because professors don’t understand how the real world works

[@TayambaMwanza]

How do you get a good code reviewer if they werent a good programmer to begin with, if theres no manual programmers anymore in the future, how can you trust the code reviews?

[@TheClassyHacker]

I watched the talk a couple days ago (then found your channel) and as a very new developer I find AI very helpful. Thank you for your take.

[@computernerd8157]

I see it helpfull as a pair programmer but do not let it become a crutch. Learned Data structers and Algorithms well.

[@ulyssesn.juarez153]

As somebody who want switch career to Tech, what would be the fundamentals to start learning in this 2024? and where? I mean a truly novice with aspirations to land a entry Level job in the near future