Literally anything AI does replaces a software engineer who would have written the code to do that had a computer not been trained to do it. You're not a software engineer or a quant.
People confuse Software Engineers with Programmers/Coders. A lot of software engineers also maintain systems, review PRs, fix bugs, do research, do cross-team communication and many others. And it's still pretty dumb overall (can't really create big reliable and secure systems). In many cases, especially when creating new frameworks or technologies, everything is super secretive and only internal engineers can develop certain products or systems. AI Software has no knowledge of what Apple for instance is working behnd the scenes. If we make an anology with construction: there are robots that 3-d print housese, but human constructors are still needed to watch those machines, to program them, to fill them, to stop and fix them in case something goes wrong, and to supervise the entire process. Seems like most people hear all these buzz words and hype and don't realize AI is more about automating certain tasks and making work easier and less about replacing humans. Airplanes fly on auto-pilot, but we still have human pilots. There are robots doing certain tasks in agriculture, but humans are still needed. And the list goes on. I don't think AI has any soft skills. If it had - it would have replaced teachers long time ago, and all other professions where people regurgitate the same material that's publicly available.
I doubt that anyone (who has some basic understanding of how software is created) would want to fly on a plane that had it's navigation software generated by LLM's or other ML mechanisms
@@bergweg These airplane softwares would be done by Electrical and Computer Engineers anyway - and these are niche type of jobs. Most software jobs are in apps and web development which is prone to AI takeover.
the moment AI can fully replace a software engineer, it won't take long until we can fully replace everyone remaining. Software itself is automation in the labour market!
@@martyr84 No, can you replace a policeman? a firefighter, a plumber, an electician, lawyers, nurses, at the current state of AI? Even traditional engineers are quite hard to replace with AI. Electrical, Mechanical, and Civil designs aren't as rigid as coding is and the practice of traditional engineering is protected by law through the P.E. license. Software Engineering on the other hand has no protection whatsoever. An H1-B indian or chinese guy can easily steal your job let alone AI.
@@martyr84except physical jobs such as trades (plumbers, electricians, etc) because it is widely more expensive to create one or more robots needed to replicate the many different tasks they perform for every single worker. That is, a physical product to replace every single worker, millions on millions of robots. With AI, once the program is created the software can simply be copied and you can replace 10,100,1000, etc workers for the same cost as only replacing one. Of course these jobs won’t last forever, but they will definitely fare better (until eventually tons of people are forced to get into trades and other physical jobs ramping up demand).
Yesterday i gave chatgpt with simple program of java which consisted of try catch final block, and each of them had print method to print some character. Even though there was no exception chatgpt still printed the character from catch block and did not printed the println method outside the try-catch blocks. And when i pointed it out then it printed it by apologizing, so i tried to trick it again, it apologized again and showed me the previous output. So my point is even for the small and easy problem like this it is giving wrong answers, so if we are to build some complex systems just by giving prompt even such a smallest mistake would bring whole system down and then we would have to debug for dont how long, since it was developed by some AI what assumptions it took and everything. And to assume that AI is just going to get better day by day and just will replace everyone within 5 years is little bit of stretch. Research takes time its not as easy as it seems that gather some huge data push into some complex NN and we get what we want. Also, the security aspect of it, if everyone can build anything by just typing in a prompt, it would be piece of cake for hackers to just duplicate the same app which they are trying to hack using the same prompt from which it was built and break it.
There's many angles you can take here. I'll take just one. Understanding existing code takes SIGNIFICANTLY longer than writing code. Even if AI can write the millions of lines of code required for a large-scale system in a single prompt, who's going to validate that the code does what it claims it does and confirm that it doesn't have critical bugs? Writing code is by far the least time consuming part of my job at FAANG. Our AI-based code completion already fails more than it succeeds. Even if we get to a world where AI can do most of the coding, it'll probably have to look like this: 1. SWEs work to understand requirements and translate that into prompts for AI 2. AI sends SWEs digestible pull requests that we thoroughly review, heavily tweak and approve. I don't see it going much further than that anytime soon. If John Doe wants to create a new solo project app, I think that'll become a lot easier since scope is small and stakes are low. Maybe we'll need a few less L3s at FAANG+, but I think most of us who work at scale are safe for at least the next 5-10 years.
As someone who is learning coding and wanting to do open source (It is what my ethics/morals guides me to do), I think AI has died down for me. I remember when ChatGPT first came out and I thought it was magic. Now, I know better that it is just a magic mushroom. It *feels* like magic, but it is a hallucination. That may discredit part of your experience if you just put in a prompt and it 'just works', but I also feel like my knowledge was stunted with these AI coding tools. The real artist behind the coder is what their end program looks like. Nothing more or less. AI will never replace that in the same way I think. Plus, only greed will want to replace the 'human element' of labor. I am sorry, but to me that is true. Anyway, I subscribed! Also, AI will never replace my dogs. I am convinced now they are smarter then Gemini or ChatGpt.
Great take! Good example of Say's law. I do think it will dumb down a lot of software engineers. Also create a lot of opportunities for 'security researchers' with all this code generated without much thought.
Exactly ! Security researchers would be able to replicate almost exact system they are trying to bring down and AI prompt most probably would almost give the identical codebase with some common assumptions which would be very easy for a hacker to break things. Because it is codebase that is what is most important and its underlying functionality, so if we can replicate even a few features of target it would be very easy to hack.
I can't wait for the day when chat gpt can perform at least the tasks I do at the same level, I'm not asking for it to be better, but unfortunately the level of neural retardation that chatgpt has right now after having tested it extensively on my projects makes it completely useless and it doesn't offer me any kind of confidence to the level of allowing it to perform the simplest task without having to check it. It goes without saying to ask him to give me the slightest innovative idea to create a trading strategy or to perform some quantitative analysis on time series. It is so useless that I will unfortunately have to do the work that I have been doing for more than 20 years myself.
Two words to y'all that kind summarizes the video's points: Jevon's Paradox Cars also didn't complete retired horses, we still use horses those to some extent :-)
The code you write is a manifestation of a creative process unique to you as an individual, it is an extention of your individuality. No machine can do that.
5:38 Mamba paper found a way to run the tokenization with unlimited context length and memory. I don't we'll all be dead. It will probably happen within a decade.
If an entity is as mentally and physically capable as the most brilliant humans is created it will be able to do anything. If it is also cheaper than humans at those tasks it will replace them entirely.
Ha ha...this is fun. I am software dev and after playing with chat gpt for some time I also come to conclusion : In terms of actually writing code i don't think I will use it...but writing regex could be useful. And may be some complex sql queries.
I'm an engineer as well, and I don't think our jobs are going away any time soon... But there's nothing fundamentally different between human minds and AI apart from the substrate on which computations happen (brain-meat vs silicon). LLMs already answer ethical questions better than most engineers (but perhaps not as well as ethics experts) if you judge it only on its output/behaviour.
I agree I cant see AI replacing senior SWE at least for now. Especially because of how buggy its code is, and its only evolving backwards at writing code, I have noticed its becoming worse in real time as months pass. Devin etc rely on ChatGPT Api calls so I dont think its gonna be anything better, once these LLMs get stuck on a bug they just run into a endless loop of trying to solve it. But it will 100% affect juniors, we need far far fewer interns/juniors than before, its essentially an enthusiastic smart junior dev for now
Coding Jesus, video suggestion. While I suspect you may traditionally serve in the Market Making space. I’d love to see from a Quant Devs point of view how you’d approach making a portfolio using C++, Python ect and some of the tools you picked up over the years. There’s a lot of BS on UA-cam about the topic, and believe you’d fill a interesting gap. Cheers!
This. I think you only really need to take one good computer science class to understand that software engineering isn't just knowing a computer language and coding. Its about abstract problem solving and is very creative in its own right. People just see a bunch of letters and numbers and think it must mean its a robot's job.
FINALLY. Someone who actually understands this stuff instead of believing that we magically have created AGI with Google. 1. Will AI produce better and more maintainable code or will it explode complexity within software projects? Who will translate requirements into the machine to produce the system? 2. The current LLMs and advancements represent the mapping of known data to generate statistical representations based on the context provided.
I agree with most of this except your implied timeframe of when iRobot will occur: we won't need to wait 100 years for AGI robots. It'll be much sooner, like this decade or next at the latest. But yet, until we (as in humanity and without too much interference from the government) reach AGI, there will be a need for a human software engineer.
I wouldn't worry until AI can actually drive a car. They've been saying that for decades and it hasn't happened yet. And self driving cars are much easier than software engineering.
Before deep learning even when you used a super computer, you couldn't bring the computer to answer a simple question. Now, everything is a question of scale and context length. People who work in the industry are really bad at predicting tech, otherwise every programmer would have created their own google or facebook. That only a tiny percentage did, means that the rest are really bad at predicting and understanding trends.
AI may not replace skilled software engineers who perform tasks greater than just writing code any time soon, but consider what a junior software engineer is, they're just a code monkey who is expected to learn on the job. The role of juniors will completely change as AI becomes more proficient. Look at Devin, for example, which is an AI that can solve >10% of GitHub issues. Eventually, these AIs will be able to reach much higher success rates and will be much cheaper than human developers. Software engineering may actually become a codeless career, unrecognizable from what it is today. It may literally become prompt engineering. And then, AI will learn to prompt for us too (i.e. a smarter version of AutoGPT), at which point our career will cease to exist. When will this happen? Who knows, but it will.
i ask chatgpt to write my phd thesis for me. i write in my basic bitch tone, chat turns it into smt that my boss wouldn't be too pissed at. chat also wrote my cover letters because wow, writer's block is real. chat even solved all my latex issues for me (after some prompting). i love chatgpt
I think the world's larger problems not necessarily software, will need software engineers to solve them. Software engineers understand systems, resources and incentives. Our current human systems like government are so full of holes and bugs they are being looted and their inevitable collapse will being on a lot of human misery. AI is very useful but for me it is basically a super search engine.
AI is a great tool to use, however, if you rely on AI tool too much, for me, I sometimes forget how to write code from scratch and learning from mistakes. It is good to utilize AI as tool for some repeatable tasks for now, not sure how we gonna do it in the future, it is a processing which takes time to figure out. If AI can replace software engineer, "business owner" who has no programming background will be replaced as well. everyone will be their own bosses, so there are bosses who don't know anything about software engineering and there are bosses who know software engineering. people won't just stay there do nothing and waiting for the worst outcomes., everyone learns and grows .
I keep offering those who think AI will replace all software engineers a challenge to build a trillionDollar company just with AI on their own to prove me wrong and they still haven't gotten back to me on that showing any money bag that would make Warren Buffett jealous.
ask an ai to setup the TWS ib_insync python wrapper to do algo trading and it will tell you everything but the correct way to do things. simply put when there are little to no examples of how to do something the "zero-shot" performance still sucks really really badly the benchmarks for all of that use toy examples that a highschool student could figure out in an hour
Well its kind of strange to leave your comment section enabled, post public videos, talk about AI & software, then say "people are leaving their unsolicited takes" in the comment section. Implying that's not what a comment section is for. Anyways heres my "unsolicited take": The main concern people have is about job prospects. Its not about "being replaced" or whatever that means. If we can have a single senior dev perform the work of 5 juniors, is this good for someone trying to enter the industry? Obviously software engineers aren't only coders, people DO understand this. But coding and implementing solutions is undeniably a decent chunk of the role. I think people already in the industry are relaxed and say things like "its always the ones who are the least experienced worrying about AI." Well again, obviously, because AI will impact entry level positions the most.
He doesn't believe in day trading. He's a quant who is a big believer in EMH and so am I after long time denial of it. There are outliers who beat the market, but in the long run only the market beats everyone.
Agricultural machines didn’t replace farmers, they made it so you need far less labor intensive work to get the same output. AI will do the same to software engineers. They’ll be able to do more with less effort.
Jesus, I am a student from a non-target who wants to break into quant. I have olympiad prep under my belt, so interviews are piss easy, But I can't secure interviews. Can you make an updated version of this ua-cam.com/video/FuNTNDEUwkw/v-deo.html. And this is your most popular video, so making one after 2 years would be good for the channel.
“AI won’t replace software engineers! Software engineering is so much more than coding. Software engineering also requires [generic soft skills]. You also need to problem solve, be creative, and think big picture.” AI won’t take your job. Mathematicians and engineers using AI will take your job.
😮 a new Challenger enters😮 have you heard of the brain organoid computer😮 we can make computers out of human brain cells by reprogramming skin cells and growing them and we've gotten even better at this with AI making them larger😮 and even more recently I just heard today connecting them to each other😮 like how the different parts of your brain are connected to each other😮 so what happens when they teach one of these things all the coding language😮 and everything that the best quant knows😮 it's your ass😮 it's like everyone just forgot biotechnology existed😮 The singularity means everything moves😮 because you made a breakthrough with AI that's helping you solve all of the problems which it is Gene problems😮 do you have any idea how many proteins were discovered in the last year like 2 million😮
LMAO you're the one who's coping. Look around you. When your ideology clashes with reality, you should update on reality, not double down on your ideology. 1. LLMs --> AI agents aren't problem solvers? How can you see techniques like Chain of thoughts and Devin and still say this with a straight face? Sure, it's crude. But you think that last 10% of whatever edge you have against Devin1 is just gonna...... stay after Devin2, Devin3, DevinTurbo..... DevinTurbo who optimizes DevinUltra ..... 2. AI can't consider ethics. Ethics matter because we still are a democracy. When labour is completely replacable, there wil be no need for ethics. Look at all the Claude/ LLaMa jailbreaks. Is it ethical to push out products that can be easily mis-used? No but it's not like it matters compared to the enormous economic gain that is to follow. 3. Have you been paying attention at how fast capabilities are scaling? "AI can't do x" gets refuted every few months, and yet you somehow imagine there's some unjustified final frontier that this global gold rush isn't going to solve. Is it against the law of physics that AI can't x? If no, you're just refusing to update on exponential capability ramp-ups. "chatGPT in 2024 is dumb, duh so my children's children will have jobs and AI as their assistants" 4. Do you think Geoffrey Hinton can't code? Do you think the thousands of software engineers (a good amount working daily on SOTA AI ) warning against further capability upgrades just.... can't code?
There is no such thing as infinite demand in any field. Consumers have a limited amount of money, time and attention. There are 24 hours in a day. AI will compete with programmers directly - it already does. AI will also compete with programmers indirectly. Some dude creating AI slop will compete with human made software for people's attention and money - AI generated software will have a natural advantage over human generated software because it's easier, quicker to make.
I was hoping for an interesting take as I personally also don't think that we need to worry that we will be replaced. At least not completely, or in the way that people think. But your reasoning was... naive is probably the right word? Like your time frame for when artificial general intelligence may be achieved just seems uninformed to me. My personal view is that it is actually very likely that some firms will hire one or two prompt engineers who are also software engineers who will prompt and oversee AGI based agents, in place of some teams. Unless you have some spiritual reason or something biasing your reasoning here I dont see how you can think humans are THAT special, so special that it will take generations before we even get close to AGI. Just 4 years ago most people in the industry laughed at the idea of having an AI based assistant. We absolutely did not think we'd all ask a chat bot to write code for us from time to time, by asking it in pure english, no less. Anyway, my personal view on why traditional software engineers aren't going away any time soon is that we as humans like when other humans do things. It just feels good. It is built into us to like what other people produce. And the keyword there is *people*. That will not go away. Just like artists will never go away, or musicians, etc. We will always crave that as human beeings. That coupled with the fact that most companies in the world move at a slow pace. Most aren't cutting edge, even if they want to be. Mostly because they are risk averse. They want to see something working for a long time before they switch to that. I'm glad you're back here on UA-cam, but I do have to say, this video was a tiny bit disappointing. But still, interesting to hear what you think, since we agree on the what, even if we don't agree on the why.
Literally anything AI does replaces a software engineer who would have written the code to do that had a computer not been trained to do it. You're not a software engineer or a quant.
The cope is unreal.
People confuse Software Engineers with Programmers/Coders. A lot of software engineers also maintain systems, review PRs, fix bugs, do research, do cross-team communication and many others. And it's still pretty dumb overall (can't really create big reliable and secure systems). In many cases, especially when creating new frameworks or technologies, everything is super secretive and only internal engineers can develop certain products or systems. AI Software has no knowledge of what Apple for instance is working behnd the scenes. If we make an anology with construction: there are robots that 3-d print housese, but human constructors are still needed to watch those machines, to program them, to fill them, to stop and fix them in case something goes wrong, and to supervise the entire process. Seems like most people hear all these buzz words and hype and don't realize AI is more about automating certain tasks and making work easier and less about replacing humans. Airplanes fly on auto-pilot, but we still have human pilots. There are robots doing certain tasks in agriculture, but humans are still needed. And the list goes on. I don't think AI has any soft skills. If it had - it would have replaced teachers long time ago, and all other professions where people regurgitate the same material that's publicly available.
I doubt that anyone (who has some basic understanding of how software is created) would want to fly on a plane that had it's navigation software generated by LLM's or other ML mechanisms
@@bergwegthis.
@@bergweg These airplane softwares would be done by Electrical and Computer Engineers anyway - and these are niche type of jobs. Most software jobs are in apps and web development which is prone to AI takeover.
the moment AI can fully replace a software engineer, it won't take long until we can fully replace everyone remaining. Software itself is automation in the labour market!
You are mistaken my friend, not all jobs are software related.
If AI does fully replace software engineers (fully autonomous agents) then you can bet that every job in the world would be disrupted.
@@martyr84 No, can you replace a policeman? a firefighter, a plumber, an electician, lawyers, nurses, at the current state of AI? Even traditional engineers are quite hard to replace with AI. Electrical, Mechanical, and Civil designs aren't as rigid as coding is and the practice of traditional engineering is protected by law through the P.E. license.
Software Engineering on the other hand has no protection whatsoever. An H1-B indian or chinese guy can easily steal your job let alone AI.
@@martyr84except physical jobs such as trades (plumbers, electricians, etc) because it is widely more expensive to create one or more robots needed to replicate the many different tasks they perform for every single worker. That is, a physical product to replace every single worker, millions on millions of robots. With AI, once the program is created the software can simply be copied and you can replace 10,100,1000, etc workers for the same cost as only replacing one. Of course these jobs won’t last forever, but they will definitely fare better (until eventually tons of people are forced to get into trades and other physical jobs ramping up demand).
Thanks for coming back. Going into my masters in Econ.
Yesterday i gave chatgpt with simple program of java which consisted of try catch final block, and each of them had print method to print some character. Even though there was no exception chatgpt still printed the character from catch block and did not printed the println method outside the try-catch blocks. And when i pointed it out then it printed it by apologizing, so i tried to trick it again, it apologized again and showed me the previous output.
So my point is even for the small and easy problem like this it is giving wrong answers, so if we are to build some complex systems just by giving prompt even such a smallest mistake would bring whole system down and then we would have to debug for dont how long, since it was developed by some AI what assumptions it took and everything.
And to assume that AI is just going to get better day by day and just will replace everyone within 5 years is little bit of stretch. Research takes time its not as easy as it seems that gather some huge data push into some complex NN and we get what we want.
Also, the security aspect of it, if everyone can build anything by just typing in a prompt, it would be piece of cake for hackers to just duplicate the same app which they are trying to hack using the same prompt from which it was built and break it.
If youre not using the paid version your getting chatgpt with half its brain tied behind its back.
There's many angles you can take here. I'll take just one. Understanding existing code takes SIGNIFICANTLY longer than writing code. Even if AI can write the millions of lines of code required for a large-scale system in a single prompt, who's going to validate that the code does what it claims it does and confirm that it doesn't have critical bugs? Writing code is by far the least time consuming part of my job at FAANG.
Our AI-based code completion already fails more than it succeeds. Even if we get to a world where AI can do most of the coding, it'll probably have to look like this:
1. SWEs work to understand requirements and translate that into prompts for AI
2. AI sends SWEs digestible pull requests that we thoroughly review, heavily tweak and approve.
I don't see it going much further than that anytime soon. If John Doe wants to create a new solo project app, I think that'll become a lot easier since scope is small and stakes are low. Maybe we'll need a few less L3s at FAANG+, but I think most of us who work at scale are safe for at least the next 5-10 years.
As someone who is learning coding and wanting to do open source (It is what my ethics/morals guides me to do), I think AI has died down for me.
I remember when ChatGPT first came out and I thought it was magic. Now, I know better that it is just a magic mushroom. It *feels* like magic, but it is a hallucination. That may discredit part of your experience if you just put in a prompt and it 'just works', but I also feel like my knowledge was stunted with these AI coding tools.
The real artist behind the coder is what their end program looks like. Nothing more or less. AI will never replace that in the same way I think. Plus, only greed will want to replace the 'human element' of labor. I am sorry, but to me that is true.
Anyway, I subscribed! Also, AI will never replace my dogs. I am convinced now they are smarter then Gemini or ChatGpt.
"Only greed will want to replace the 'human element of labor". Yes and unfortunately most people in power are greedy
Great take! Good example of Say's law. I do think it will dumb down a lot of software engineers. Also create a lot of opportunities for 'security researchers' with all this code generated without much thought.
Exactly ! Security researchers would be able to replicate almost exact system they are trying to bring down and AI prompt most probably would almost give the identical codebase with some common assumptions which would be very easy for a hacker to break things. Because it is codebase that is what is most important and its underlying functionality, so if we can replicate even a few features of target it would be very easy to hack.
Thank you mewing jesus for another sharp point of view
I can't wait for the day when chat gpt can perform at least the tasks I do at the same level, I'm not asking for it to be better, but unfortunately the level of neural retardation that chatgpt has right now after having tested it extensively on my projects makes it completely useless and it doesn't offer me any kind of confidence to the level of allowing it to perform the simplest task without having to check it. It goes without saying to ask him to give me the slightest innovative idea to create a trading strategy or to perform some quantitative analysis on time series. It is so useless that I will unfortunately have to do the work that I have been doing for more than 20 years myself.
Two words to y'all that kind summarizes the video's points: Jevon's Paradox
Cars also didn't complete retired horses, we still use horses those to some extent :-)
The code you write is a manifestation of a creative process unique to you as an individual, it is an extention of your individuality. No machine can do that.
**Greg Rutkowski enters chat**
Hey, I'm super glad you're back making vids.
5:38 Mamba paper found a way to run the tokenization with unlimited context length and memory. I don't we'll all be dead. It will probably happen within a decade.
If an entity is as mentally and physically capable as the most brilliant humans is created it will be able to do anything. If it is also cheaper than humans at those tasks it will replace them entirely.
Agreed, I use gpt often enough as an assistant. As for when AI becomes AGI, who knows, but it could come faster than anticipated...
Where's the Baristas?
Coping in the comment section.
Ha ha...this is fun. I am software dev and after playing with chat gpt for some time I also come to conclusion : In terms of actually writing code i don't think I will use it...but writing regex could be useful. And may be some complex sql queries.
Chat GPT will evolve, this is what you didn't mention!
I'm an engineer as well, and I don't think our jobs are going away any time soon... But there's nothing fundamentally different between human minds and AI apart from the substrate on which computations happen (brain-meat vs silicon). LLMs already answer ethical questions better than most engineers (but perhaps not as well as ethics experts) if you judge it only on its output/behaviour.
I agree I cant see AI replacing senior SWE at least for now. Especially because of how buggy its code is, and its only evolving backwards at writing code, I have noticed its becoming worse in real time as months pass. Devin etc rely on ChatGPT Api calls so I dont think its gonna be anything better, once these LLMs get stuck on a bug they just run into a endless loop of trying to solve it. But it will 100% affect juniors, we need far far fewer interns/juniors than before, its essentially an enthusiastic smart junior dev for now
Coding Jesus, video suggestion. While I suspect you may traditionally serve in the Market Making space. I’d love to see from a Quant Devs point of view how you’d approach making a portfolio using C++, Python ect and some of the tools you picked up over the years. There’s a lot of BS on UA-cam about the topic, and believe you’d fill a interesting gap. Cheers!
Upcoming CS freshman and aspiring quant dev here, second this!
@@adipandey06 third this!
This. I think you only really need to take one good computer science class to understand that software engineering isn't just knowing a computer language and coding. Its about abstract problem solving and is very creative in its own right. People just see a bunch of letters and numbers and think it must mean its a robot's job.
FINALLY. Someone who actually understands this stuff instead of believing that we magically have created AGI with Google.
1. Will AI produce better and more maintainable code or will it explode complexity within software projects? Who will translate requirements into the machine to produce the system?
2. The current LLMs and advancements represent the mapping of known data to generate statistical representations based on the context provided.
Clarity. Thank you
I agree with most of this except your implied timeframe of when iRobot will occur: we won't need to wait 100 years for AGI robots. It'll be much sooner, like this decade or next at the latest. But yet, until we (as in humanity and without too much interference from the government) reach AGI, there will be a need for a human software engineer.
Thank you Jesus - I feel much better now!
I wouldn't say 'never' but I feel pretty confident saying 'not in my lifetime'!
I wouldn't worry until AI can actually drive a car. They've been saying that for decades and it hasn't happened yet. And self driving cars are much easier than software engineering.
Actually AI can drive a car pretty well now.FSD 12!
ua-cam.com/video/r3n5Rh_fyd0/v-deo.html
Prospects for Self Driving Cars: Discussion with Gary Marcus, NYU
Yeah my car drives me to and from school everyday
When is the 3rd live session of the Orderbook going to be ?
It’ll be a video not a livestream. Livestreams seem to perform poorly. Next week.
@@CodingJesus It is fun to participate live but not as fun to watch the vod imo. Keep doing the content tho, its super interesting!
Before deep learning even when you used a super computer, you couldn't bring the computer to answer a simple question. Now, everything is a question of scale and context length. People who work in the industry are really bad at predicting tech, otherwise every programmer would have created their own google or facebook. That only a tiny percentage did, means that the rest are really bad at predicting and understanding trends.
Worried less about artificial assistants and more about human code reviewers…yep, that will probably compile and run…
AI may not replace skilled software engineers who perform tasks greater than just writing code any time soon, but consider what a junior software engineer is, they're just a code monkey who is expected to learn on the job. The role of juniors will completely change as AI becomes more proficient. Look at Devin, for example, which is an AI that can solve >10% of GitHub issues. Eventually, these AIs will be able to reach much higher success rates and will be much cheaper than human developers. Software engineering may actually become a codeless career, unrecognizable from what it is today. It may literally become prompt engineering. And then, AI will learn to prompt for us too (i.e. a smarter version of AutoGPT), at which point our career will cease to exist. When will this happen? Who knows, but it will.
i ask chatgpt to write my phd thesis for me. i write in my basic bitch tone, chat turns it into smt that my boss wouldn't be too pissed at. chat also wrote my cover letters because wow, writer's block is real. chat even solved all my latex issues for me (after some prompting). i love chatgpt
I think the world's larger problems not necessarily software, will need software engineers to solve them. Software engineers understand systems, resources and incentives. Our current human systems like government are so full of holes and bugs they are being looted and their inevitable collapse will being on a lot of human misery. AI is very useful but for me it is basically a super search engine.
AI just replaced Stackoverflow only
AI is a great tool to use, however, if you rely on AI tool too much, for me, I sometimes forget how to write code from scratch and learning from mistakes. It is good to utilize AI as tool for some repeatable tasks for now, not sure how we gonna do it in the future, it is a processing which takes time to figure out. If AI can replace software engineer, "business owner" who has no programming background will be replaced as well. everyone will be their own bosses, so there are bosses who don't know anything about software engineering and there are bosses who know software engineering. people won't just stay there do nothing and waiting for the worst outcomes., everyone learns and grows .
I keep offering those who think AI will replace all software engineers a challenge to build a trillionDollar company just with AI on their own to prove me wrong and they still haven't gotten back to me on that showing any money bag that would make Warren Buffett jealous.
irrefutable logic indeed.
I mostly agree with you, but ... John Carmack.
What do you think of robot's trading
AI is already replacing software engineers, but its not DISplacing them
Exact tautology can be used for algorithms. Developers who rely on bad algorithms and don't correct - will lose their jobs.
making swers more efficient = replace
replacing & displacing are for all intents and purposes- synonymous
ask an ai to setup the TWS ib_insync python wrapper to do algo trading and it will tell you everything but the correct way to do things. simply put when there are little to no examples of how to do something the "zero-shot" performance still sucks really really badly the benchmarks for all of that use toy examples that a highschool student could figure out in an hour
Coding Jesus to the moon.
Well its kind of strange to leave your comment section enabled, post public videos, talk about AI & software, then say "people are leaving their unsolicited takes" in the comment section. Implying that's not what a comment section is for.
Anyways heres my "unsolicited take":
The main concern people have is about job prospects. Its not about "being replaced" or whatever that means. If we can have a single senior dev perform the work of 5 juniors, is this good for someone trying to enter the industry?
Obviously software engineers aren't only coders, people DO understand this. But coding and implementing solutions is undeniably a decent chunk of the role.
I think people already in the industry are relaxed and say things like "its always the ones who are the least experienced worrying about AI." Well again, obviously, because AI will impact entry level positions the most.
Is copy trading a good strategy.
He doesn't believe in day trading. He's a quant who is a big believer in EMH and so am I after long time denial of it. There are outliers who beat the market, but in the long run only the market beats everyone.
What about Devin?
Whole demo was faked. Its really quite a useless piece of software
Mogging thumbnail 😂😂
Anti-software engineer AI enthusiast copers got mogged by the thumbnail. 😂
@@CodingJesus based
Amen Jesus! 👏
In copy trading you copy traded from other traders doing well
Agricultural machines didn’t replace farmers, they made it so you need far less labor intensive work to get the same output.
AI will do the same to software engineers. They’ll be able to do more with less effort.
mogging jejus
Jesus, I am a student from a non-target who wants to break into quant. I have olympiad prep under my belt, so interviews are piss easy, But I can't secure interviews. Can you make an updated version of this ua-cam.com/video/FuNTNDEUwkw/v-deo.html. And this is your most popular video, so making one after 2 years would be good for the channel.
Will do
Just to be clear for qt role internships
People who say AI will replace x job are either just insecure or they know nothing about the profession they are talking about
Amen🙏
its like mathematicians are done because we have calculators
I like
This
god has spoken
“AI won’t replace software engineers! Software engineering is so much more than coding. Software engineering also requires [generic soft skills]. You also need to problem solve, be creative, and think big picture.”
AI won’t take your job. Mathematicians and engineers using AI will take your job.
Yea lotta those out there
If you listen to him talk but look away from the screen, you would swear Tim Pool was talking.
😂
@@CodingJesus
You sound just like him. Not anything you said, but just your voice sounds exactly the same as his.
Jesus is here again. Thanks coding Jesus!
Devin entered the chat, CEO of NVIDIA entered the chat
You kept it real! 🤣🤣🤣🤣
as time goes on your start losing hair as a quant
AI will replace devs and quants 100%.
no, people < LLMs.
😮 a new Challenger enters😮 have you heard of the brain organoid computer😮 we can make computers out of human brain cells by reprogramming skin cells and growing them and we've gotten even better at this with AI making them larger😮 and even more recently I just heard today connecting them to each other😮 like how the different parts of your brain are connected to each other😮 so what happens when they teach one of these things all the coding language😮 and everything that the best quant knows😮 it's your ass😮 it's like everyone just forgot biotechnology existed😮 The singularity means everything moves😮 because you made a breakthrough with AI that's helping you solve all of the problems which it is Gene problems😮 do you have any idea how many proteins were discovered in the last year like 2 million😮
All interns are in danger 😊
nah it's getting better fast. ur coping
LMAO you're the one who's coping. Look around you. When your ideology clashes with reality, you should update on reality, not double down on your ideology.
1. LLMs --> AI agents aren't problem solvers? How can you see techniques like Chain of thoughts and Devin and still say this with a straight face? Sure, it's crude. But you think that last 10% of whatever edge you have against Devin1 is just gonna...... stay after Devin2, Devin3, DevinTurbo..... DevinTurbo who optimizes DevinUltra .....
2. AI can't consider ethics. Ethics matter because we still are a democracy. When labour is completely replacable, there wil be no need for ethics. Look at all the Claude/ LLaMa jailbreaks. Is it ethical to push out products that can be easily mis-used? No but it's not like it matters compared to the enormous economic gain that is to follow.
3. Have you been paying attention at how fast capabilities are scaling? "AI can't do x" gets refuted every few months, and yet you somehow imagine there's some unjustified final frontier that this global gold rush isn't going to solve. Is it against the law of physics that AI can't x? If no, you're just refusing to update on exponential capability ramp-ups. "chatGPT in 2024 is dumb, duh so my children's children will have jobs and AI as their assistants"
4. Do you think Geoffrey Hinton can't code? Do you think the thousands of software engineers (a good amount working daily on SOTA AI ) warning against further capability upgrades just.... can't code?
There is no such thing as infinite demand in any field. Consumers have a limited amount of money, time and attention. There are 24 hours in a day.
AI will compete with programmers directly - it already does. AI will also compete with programmers indirectly. Some dude creating AI slop will compete with human made software for people's attention and money - AI generated software will have a natural advantage over human generated software because it's easier, quicker to make.
I was hoping for an interesting take as I personally also don't think that we need to worry that we will be replaced. At least not completely, or in the way that people think. But your reasoning was... naive is probably the right word? Like your time frame for when artificial general intelligence may be achieved just seems uninformed to me. My personal view is that it is actually very likely that some firms will hire one or two prompt engineers who are also software engineers who will prompt and oversee AGI based agents, in place of some teams. Unless you have some spiritual reason or something biasing your reasoning here I dont see how you can think humans are THAT special, so special that it will take generations before we even get close to AGI. Just 4 years ago most people in the industry laughed at the idea of having an AI based assistant. We absolutely did not think we'd all ask a chat bot to write code for us from time to time, by asking it in pure english, no less.
Anyway, my personal view on why traditional software engineers aren't going away any time soon is that we as humans like when other humans do things. It just feels good. It is built into us to like what other people produce. And the keyword there is *people*. That will not go away. Just like artists will never go away, or musicians, etc. We will always crave that as human beeings. That coupled with the fact that most companies in the world move at a slow pace. Most aren't cutting edge, even if they want to be. Mostly because they are risk averse. They want to see something working for a long time before they switch to that.
I'm glad you're back here on UA-cam, but I do have to say, this video was a tiny bit disappointing. But still, interesting to hear what you think, since we agree on the what, even if we don't agree on the why.
You remind me of the "as a software engineer" guy who went to prison for fraud and who was never at any point a software engineer.
Cope harder
Coding Jesus, have you resurrected from AI coding world ? 😅😮🎉❤🫡😺🥳