As a guy who runs an AI consultancy firm, this is an illusion. Coding is like 10% of the job. The main problem is the training data (to acquire, label, prepare, etc), and I don't see much improvement in the last 3 years.
@@JimMilton1 LOL! i'm so excited to see the outcome of this! A super powerful bullshit generator that eats it's own excrement ...what could possibly go wrong?
A lot of non-tech companies, including large ones, don't even have good data governance. They don't have policies regarding the collection, storage, maintenance, and disposal of data. It's sort of a hodge podge of different departments doing their own thing with data. So, even before AI can make a positive impact, these companies need to get their data governance policies in place.
@@tongobong1 Far easier said then done. There will be a ton of office politics that a good consultant would have to manage in order to leverage the full capability of AI. Department heads won't give up their data (power) easily .
@@tongobong1 Most definitely. That's been my experience as a Team Lead. AI is a different animal entirely though. Adaption will lead to people losing their jobs/careers.
Although Im totally with you on data and its importance, I still call to listen to client more. Right now I have a client with absolutely garbage data infrastructure, they have several offers from IT companies that give a perfect plan (step by step, all is really good etc). The only problem is that the client needs quick wins to justify AI/ML investments, and they are not ready to launch a project when they will get feasible results only in like 9 months or more. So my advise to them now - “ok, lets try some basic cheap AutoML stuff, if you’ll value in, then we do everything right”. And yes, i told them all pros and downsides of such approach
From a Senior Software Engineer at a huge bank: AI is the future... but the future is pure garbage. We are currently having a problem with people even understanding the basics of JSON and YAML and when you throw in an LLM engine that just regurgitates stupidity... well... we are all doomed. Good luck, everyone.
Having worked in tech I agree with the original comment. I don't see how someone with a background in AI learning rudimentary full-stack development helps with the issue @@MARTIN-101
That's one of the many issues: AI cannot be double checked for consistency and accuracy in it's answers/outcomes. So, unless you are a very skilled professional that knows what is BS and what is not, many entry level professionals could potentially get stuck in something that doesn''t work and that they are unable to make it work.
I've been a data scientist for 20 years and I've never seen Andrew so far off the mark. I have delivered dozens of these solutions in the tail. The model is typically you get a data scientist like myself to find what the thing is. Build out a prototype using open source libraries. Plenty of code there and then you get some data engineers and machine learning engineers in to harden up the code base and scale it
@@john.8805these would be your standard regression and classification models which have to be bespoke to the use case of the specific client. Lead generation. Price prediction. Churn likelihood. Probability to fail. This is the whole job of a data scientist. I don't understand how this is news to anybody
I’ve been thinking about this lately. I told my friend that I was getting into tech and I wanted to start by learning software development. I was surprised by what he told me. He said that’s a basic skill rn and what I need is to either get into cybersecurity or cloud. All these areas use Python as their main language. I started learning python and I’m impressed by the versatility of the language. I mean you can use this language to do almost anything. From data analysis (Python is now in excel), data visualization, network automation you name it!
Yes but you shouldnt run applications off it. You dont learn programming languages. You just learn "programming" and most languages look the exact same.
python is a jack of all trades, it can do absolute everything but is a master of none. The trade-off for a company is fast deployment and low cost (low cost due to fast deployment, need to pay less hours of working to employees) vs generall "poor" performance. By "poor" it means for very specific situations where speed and stability is critical, like a mobile or web app with millions access, , trillions rows of data, then an AI in Rust for example would perform better at the cost it takes a bit more time to fine tuning it. Or just get Azure, GCP, AWS, and their top tier engineers already did all of this hard work 😄. For small applications, self projects, prototypes, python all the way. But! Python is getting closer and closer in terms of performance. Getting rid of GIL is a feature is something will be a game changer for python to support the speed of true parallelism. Is already on the table but may take few years to me implemented. A glimpse of this would be Mojo language.
Python is a great programming language. It's straightforward, consistent and very intuitive. But occasionally it does throw a curveball, in syntax and in the way it works. It however doesn't give you same sense of freedom as C++ does. C++ on the other hand allows you to approach any problem in the way you see fit. You can use C++ in very much same way as Python, but you can use C++ in many different ways aswell. Different tools for different jobs.
@@cpK054L I know that and some of it's core functionality relies on c++ not to mention some libraries. They are still very different languages in syntax and in functionality.
I am a freelance developer and I was waiting for these AI tools to start building AI solutions for clients. And here they are, the public LLMs and so forth. I share your view that I can be an AI developer without having to be a machine learning engineer. I really enjoy when you comment on an interesting presentation. Great video !
I took a year of computer programing back maybe before you were born but life took me a different path, retired and thinking I need a winter hobby. thanks, worth a sub...👍✌🖖🥃
Hey folk, i am currently learning js, the concern is by the time i learn full stack with good projects (stuff like nextjs, ShadCN, tailwind, docker, kubernetes, CI/CD), genAI, langchain, deep learning stuff, 2026 or 2027 will come, where the hell will my requirement accomodate in the industry, because all these senior engineers by that time (like you) have already replace junior devs, what value will I provide? Pls help, it's infuriating.....
I love AI, I am a robotics and mechatronics engineer and I hate coding from scratch, so AI is good for that. I prefer to modify code that I get AI to write from scratch and then get AI to check it and optimise it. I like ladder programming and assembly by myself, as I like to run it in my mind step by step.
I have mostly been full stack Java, JavaScript, PHP, Python, SQL but have also done work with Analytics and Data warehouses with Python, SQL, Java, ETL
I think in the big picture, python is in a similar place to where BASIC was during the 80's or PASCAL was during the 90's. But just looking at the differences between these three languages shows just how far we've come.
@@patpat8727 I mean with BASIC we had something that was only really useful for teaching programming and from a pure language standpoint, a pretty bad language (taught a lot of bad habits and hard to scale). Then PASCAL was a step forward since beginners at least got a language which could be used to create large, useful programs even if that didn't happen much in actual industry or with commercial development. Finally today with Python beginners get a language that is easy to scale, has plenty of useful libraries, and is sometimes even used by industry or in the commercial world. So a beginner today is in a far better place than in the 80's in terms of what they can do, the usefulness of what they learn, etc.
I love that he said, "grounding" in Python, LLM and ML... love this pun, not sure how many people got it but I smiled when I heard. If you don't know what "grounding" is....yeah 🙂
beware of entering a position that has a low barrier to entry like "python developer". This means lots of people can become qualified and compete with you, thus driving down job security and pay.
As if there's such thing as "pure python developer" - it's never about python, or any other language, it's about freaking data, design, structure, cloud, and so on... There's NO market for pure python.
Thank you Travis for this great information To be honest with you Learning Python is a great way for beginners get into Tech. I love 3 Field in Tech and they are Software engineering, AI and Machine Learning and Cloud engineering and I will make sure I train myself to be an expert in those fields, Am currently learning Python right now and i started 2month ago
Thank you, Travis! I've been following you for a while and I'd say the content you produce helps me a lot on my programming journey. Thanks again for sharing valuable knowledge which is quite rare to find.
Having worked as a jr dev at a software company, its difficult for me to see an AI tool that is capable of letting the client build professional websites with little to no code required. Pro websites are far too complicated and have problems all the time. One thing ive learned about SWE is that things break all the time and it takes a deceloper to figure out why. Implementing a fix in the existing code base ranges. At times, I'd think something was so easy to fix but it turned into a much more challenging task. Sorry, I just don't see it. There are tools like WIX that build websites for small businesses but these sites are like the bare minimum in quailty and functionality.
I don't think that's the point. As a dev you would take an existing image recognition neural network and train it on a customer's problem(identify bad quality cheese spreading on a pizza) using their photos(hopefully annotated). Then you would give them an interface or automate the process by making an ai prompt like "is this pizza ok?". If the data isn't annotated you could set up a team of inexpensive remote workers that sit there and annotate data for very low wages. If you recycle your interfaces you're talking about very little development work and these projects become affordable to small businesses.
Python is the defacto language for AI/ Machine Learning/ Data Science. But id argue Mojo will most likely be the future as it is basically a superset of Python.
@@vectoralphaSec LLVM IR compilation from Python bindings will be multiple order of magnitude faster than Python and way faster than MOJO, while not requiring new synthax. Mojo is a hype
Am so glad I came across this channel.New subscriber here, and just as you mentioned I went down that Cloud path with AWS and I even got the foundation CCP certification. Focusing on building my Python skills✅
There has always been developer oportunities for small use cases. Think of the plug in writers for all sorts of applications and use cases. Right now, people are slapping .AI on everything like it's monkfruit or stevia
Update I just spent 5 hours with chat gpt4 creating a JavaScript filtering script for an insurance project I'm working on. It's working exactly the way I wanted it to. I can expandt it. I added features and functionality as it was built. It saved me money . I'm learning how the sausage is made while I talk to ( prompt) the sausage maker. I see and correct mistakes thru prompting. I'll deploy it in production. I now see the future. No more rip off developers. I can do this by myself as I teach myself. Greatest thing since 3D printers and oh yes that's next. As this tech progresses the sky is the limit.
This is already the field of embedded software engineering and control systems engineering, using PID controller, and small computers such as arduino and raspberry pi
in one of my jobs the users always wanted the applications to be as simple as possible, I always told my coworkers they want the button "do my job" so until AIs are able to read the mind of people, we devs have work that will look similar to what we have right now, but even then when everyone has a neuralink and AI can read their minds, people need the discipline to follow the logical steps to build a product/solution & if they want the "do my job" button, then they won't even think about the solution.
Andrew ng was pioneer and had wide knowledge of old ai. His videos helped lot of people. I have not heard him talk anything substantial about Transformers and large language models
Travis, thank you very much for all your videos! I feel like you work with the speed of light as well uploading video after video!! 🙏🙏🙏 I started learning following your web dev 6month blue print(HTML,CSS,JS) doing Colt Steele's course on Udemy. Thank you very much! I really love it and I see my progress every day. I really hope to be able to fit in somewhere with my knowledge and be able to evolve further. I just wanted to thank you again, and please believe me your videos inspired me and gave me confidence🙏🙏🙏 Greetings from Romania!! 🤗😘
@@TravisMediaclip said great Azure solution and cert but free blueprint link says AWS. Do you have a free blueprint for Azure? Or does the the AWS blueprint apply to both? Great clip all the same
Andrew Ng: "Ng" is pronounced like "eng" as in the way "Engelbert Humperdinck" is pronounced by Americans. An AI pioneer in his own right, he's also a fellow techno-optimist and effective accelerationist.
I agree with the subject on a macro level but why a developer would be more suited to execute this smaller projects than lets say a data science, statistician or pure mathematician (which usually lack on strong developing skills). From what I am seeing the need to have high levels of expertise in developing is diminishing as no code ML platforms are being used, yet you still need deep knowledge on data analysis, understanding results and deriving conclusions which other fields have stronger. Give trash data to the model will throw trash results, no amount of fancy ML will make a model work without domain expertise & the understanding of the math behind.
Interesting. I have tried, many many times over the years (I recently turned 60 yrs old), to learn coding, most often with Python. I've gotten a few books, watched many tutorial videos and started online classes. I usually get to a point where I'm doing a lesson/exercise/lab and get stuck somehow, something just isn't clicking, and I can't figure it out/get passed it and there's no realtime instructor to ask, even via email. Now, I was just personally interested, I'm retired, and I'm not looking to get back in to work. I don't know if my brain is wired to "get it" when it comes to understanding coding. My career was in high-tech, oddly enough, and I worked with a lot of LINUX/UNIX systems, used regular expressions, etc. But it was more from the systems were vendor provided/supported, and we were the in-house "technicians" that handled software upgrades, module replacements, outage recoveries, etc., it wasn't writing code or anything ...
Where do you get stuck? You should divide the problem into several small problems - steps and solve them one by one and test each of them on the way. That way programming is quite easy. You also shouldn't think much about the design of the program. Just do it and then you will refactor and redesign it to make it more clean after you have workable solution.
@@tongobong1 Thank you! Hmm, typically it's several lessons in, variables I think. I think I end up going too long in lessons at one sitting, like several hours, and just get tired and then frustrated, I think I just followed along, monkey-see, monkey-do, but didnt fully grasp everything. I should take a break/let things soak in, and go back to it. It's the only programming I would know. Thank you for the support, I appreciate it!
@@ph5915I propose that you just set yourself a goal to create some kind of a rather simple app in python. Maybe you can create simple calculator app or simple game... The 1. step is to create a window and then 2. add some buttons and then 3. some code to run when you press buttons. To complete each step you can find tutorials, examples and documents online. The best way to learn is to build something useful or fun because then you learn stuff that you really need to learn at that moment to complete the next step. This is how you become the real programmer and not just a guy that is learning all the "theory" yet he doesn't know how to use it in practice.
This is very interesting and kind of aligns with my thoughts or focuses on applying AI (outside of software and consumer tech) but I hate writing Python lol I learned development writing C++ so I’ve always gravitated towards things like Java, JS, C# (kinda) but I guess I should brush up on my Python.
I found amusing in the digital era, people thinking that learning how to code will became obsolete.. it will not, maybe the work change a litle bit, but knowing code it will be revelant.. because not all bugs are syntax or programming error, a code could work just fine and still be no good.. and write your on code a lot of the time is more productive than reading someone else code and trying to figure out if is working as intendend.. if ai is like 95% accuracy good luck trying to find the 5%.
I am not a programmer but what you saying about the %5 is true in my case because when I use chatgpt to write a complete code for me, in most of times it writes a good code but still I find that it has a crucial missing parts to it that I keep scratching my head into the wall trying to figure it out and most of times I just give up and move on to another task but however I think this will not last for so long as A.I is getting better and better so it made me change my mind to learn a programming language myself like Python or Kotlin as I think I will be just wasting my time while A.I is moving in light speed
This is my thought exactly but that’s still not optimistic. Basically means ai will code and we become the chatgpt. Like a janitor position. I agree the most annoying bugs are often not related to the code and ai will have trouble with that One issue I had recently was an app crash and after debugging it just had to do with using google canary browser. Another issue I had , I solved by downloading the recent code commit from GitHub which has the same code but doesn’t crash for some reason (something to do with dev environment). How would ai or newbie figure out issues like this ? The worst bugs have the stupidest solutions. The equivalent of unplugging a computer and plugging it back in. I can’t imagine an ai suggesting solutions like that because it has nothing to do with the code
@@ahmedsalah7474 as a programmer I had gpt code the bulk of my app and became convinced there’s no way a non coder can do this. You’d end up wasting a lot od time and getting stuck. Many times I had absurd issues that would end a non coder when ai can’t help and especially if you have no thought process of how to prompt further or figure it out on your own I don’t think you’ll waste your time learning the basics and fundamentals. I think nowadays it’s bad idea to learn software engineering but the best time and maybe very important to learn basics and fundamentals of code and engineering. Everyone should know how to code in era of ai . But just not going into software as your expertise
The problem with all these companies is the legacy software. You need a way to enter and edit the data. Where the opportunity is, is creating an AI that can be trained to use these software.
"AI expert" - translation: a person who has never programmed a computer, but knows more than you, because they once did a business course, and they subscribe to the WallStreetJournal
@@TravisMedia in this case … no :) an exception to the rule for sure. As a wild generalisation though, you have to admit it’s a pretty good rule of thumb. We do seem to have an epidemic of AI experts these days, and it’s likely to get worse over the next couple of years. By then, the next misleading buzzword will emerge, and give us a whole army of new experts who still can’t program a computer. Rinse and repeat.
I really appreciate this video. There is so much doomerism in the tech space right now and this is a genuinely uplifting thing to hear. There's still a chance, you just have to position yourself correctly as always
im a software engineer, not a programmer, and haven't written any code at work in a few months when AI is able to replace the actual responsibilities of a software engineer, it will be capable of replacing any job
AI it can't really replace all real sw engineers because many aspects of SW engineering is often accessing proprietary hardware and apps like ChatGPT can't possibly understand proprietary hardware. You could write AI models to understand proprietary hardware but you would need to pay AI SW devs to do it. In which case would require even more higher payed SW engineers.
@@chrischoir3594bro, will you tell your experience in tech, also what dyt about what should a guy who has 2-3 years learn to become kinda secure and keep up with upcoming market situations?
Thank you so much. I own a small SEO AI SaaS bootstrapped to 10k MRR with no marketing and i really am frustrated by my lack of code knowledge. I plan on learning python now.
These opportunities are opening up... but for how long? The way I see it, you basically have to stay on top of AI technology and ride the wave for as long as you can. Maybe it'll be 15 years or maybe 3 but eventually AI will have sufficient capability so that there really isn't any middle man. It will listen to your requirements and create what you want. All you can do is try to be as high up on the ladder as you can as it gets pulled up.
The goal in tech is profit per employee and it always has been. So if that long tail was going to provide job opportunities for software engineers, big tech responds by investing in AI to try to cut out as many potential employees as possible. Then you can retain the profit per employee benefits with your ultra-productive, relatively small group of engineers. In the past, companies maybe thought they needed to hire a lot of engineers to show investors that they are growing. Now, there isn't such an expectation. There is a realization amongst tech leadership that maybe the work can be done with far fewer people.
Thank you for this guide sir, You are one of my "Notification ON" mentor, Your first video I saw was 10 essential certificates... and you are really helping me learn how to approach this new evolution Since last couple of days I was thinking that there is a major change in the way we will be coding and learning things... And I was think how can I learn web Development (learning form THe Odin Project, on foundation) journer more ai driven and take a leap in learning these ai tools and combining it with treditional technologise. Like thare are many no code or low code project using ai tool but in current time we still need unterstand treditional way to code to check weather ai is coding properly.
Addressing practical limitations to simply using AI tools... It will be costly, and it will (for a while, at least) lead to far more carbon being released into the atmosphere. From a societal perspective, we are running way behind on renewable energy adoption relative to the speed of progress in AI, with scaling of data centers that release more carbon.
I don’t like everything built just around python. We can call a python web service from a Java backend and make an Angular/ React front end to consume it as required. Because Java is much faster in receiving and retrieving it would be better to use it as a backend. I’ve worked with Streamlit and it is very very slow in giving the data for outlier analysis.
The trouble with all these online experts talking about software development, especially university lecturers, is that they dont have any real world experience out in industry commercially developing software, and so dont know what they are talking about when it comes to what industry expects. You dont even mention C# or Java which are the worlds industry standards for application development with vast prewritten libraries for reuse. Now why is that? As soon as he said "100s of developers" for measuring the cheese on a pizza it is a giveaway he doesnt know what he is talking about. Feels like all the failed developers out there are salivating on AI taking revenge for their own short comings. lol. Industry expects a design, an architecture, reusable code, meaningful error handling, verifiable automated tests based on code developed, etc, etc., and TO BE IN CONTROL of all parts. AI will have its uses, but not in an informal unstructured language like Python.
Sorry but I’m not as enthusiastic about AI. I believe it will lower the need for software engineers and human design creativity will be taken out of the equation.
Ya Im worried about lowering the need for software engineers too, but i think we are quite a ways from that. Ive found that most people who preach that you will soon lose your code job to AI, are often selling something / inveested in the tech. But also, theres nothing we can do to stop it, so we may as well be ready .
@@hikemalliday6007 the VAST majority of Senior developers do not think AI is going to take all the jobs any time soon. It is really only Junior devs that think this is the case. Which, if you think about it, makes sense. A Junior dev doesn't know much yet and been in the thick of it in development. They get handed tasks like "there is a bug on this page that moves the comment box out of alignment on mobile devices-re-center it on mobile devices without affecting desktop viewing." And that sort of bug is something that can be done pretty slowly by a Junior, possibly fairly quickly by AI, and super fast by a Senior dev. But what people forget is that a Senior dev can't be bothered to realign a comment box on mobile devices. An office drone wouldn't know how to take ChatGPT's output and put it into the test environment without themselves becoming a Junior dev (though without any programming knowledge). So either the Senior dev has to now do the task, or they're going to need to train the office drone to the competency of a Junior dev who doesn't know how to program anything themselves. The point of a Junior dev is to push that task to them. The Senior dev is trying to figure out how to stop the database from crashing when the user numbers spike 200% at intermittent intervals throughout the day, without needing to run the server at maximum all day long that would cost the company $20k/month. Those tasks that get pushed to the Junior is how they become a Senior, so they can replace the current Senior and solve those $20k/month challenges. And the Senior dev needs to be replaced so they can move on to Architect or Tech Lead and identify changes to save $35k/month. There is a huge shortage of Seniors right now, and the Seniors see absolutely no reason to believe AI is going to take their jobs. Eventually, companies will have to settle and hire more Juniors again. The companies who tried to replace 2 juniors with 1 junior and ChatGPT are going to find themselves in hot water when they need 2 Seniors and only have 1 Junior to promote.
@@holdfast5332Not the same man. What you are talking about is evolution throughout decades or even centuries. What's happening to us now is something that is going to happen well within our lifespan
Don’t be, sure the job is not gonna be there, instead it will evolve. I.e. taxi drivers = ubers. Especially with the younger generation of American citizens refraining from having kids. Ai implementation in something like immigration would take a process that takes years and complete it in a couple months. Logistics, economy, advice, medicine and most importantly, actual people behind security to keep ai safe. If you have an understanding of economics, society and how it has evolved every two or three generations, you’ll get the bigger picture. Doors are opening my friend, you just gotta learn how to walk through them.
If ai is cheaper and better even filling senior positions, the only job left will be entrepreneur. Ai can help with that too. So output will go up tremendously as supply constraints vanish. Our only limitation would be time to consume.
Until we reach AGI (which is still a big if), we will always have a need for coders. AI can only regurgitate data that it was trained on. It can't reliably create anything outside of its training data set. It also suffers from hallucinations, causing inaccuracies. The more you take a Jesus take the wheel approach to AI coding, the more you will experience this kind of unreliable, and unhelpful scenario. Once we have AGI, then I see our jobs going the way of the dinosaurs. That doesn't bother me though, because when we get AGI, it will be able to replace most jobs, and at that point, governments will be forced to change the nature of economics or they will have a revolt on their hands due to all of the unemployed workers with no opportunity for work. We will end up with UBI, or some other economic model in a post-labor economy.
well, Python was made for non-developers, for researchers and such, so is kinda fitting, however all those python libraries are just wrappers on C libraries to get the speed, & any other programming language with a native interface can get access to those as well.
These opportunities evaporate when/if AGI comes along, though. Maybe that's a year our. Maybe 50 years out. One of the concerning dynamics. If we stick with the current AI capabilities, then there will certainly be plenty of work for devs. I'm not really expecting any plateaus in progress just yet...
@@shugyosha7924Yes, because manual labour does not scale very well the average earning ceiling will be very low. Outside of that there is owning real estate and businesses...
@@reyne2077Think from the business perspective of the developers. They want to create a software that makes mental human labor obsolete, so they can profit from or even monopolise the new AI driven workforce. Since most physical labor not yet automated requires expensive robotics (ex: robot plumbers) most usually doesn't count those. So AGI would have skills that are on par or even better than your average worker. You can argue that this could be a lot of small AI's rather than one hughe one, which could be the case, but the end goal doesn't change.
Im 30 years old, doing masters in tax and learning to code. But I had passion for soccer my whole life . I always wondered if i would make a good coach. Guess now is a good a time as any...
That's a really odd combination. Tax and code has almost nothing in common, unless you're planning to work for one of the tax software companies like Turbotax. I think the future will see increasing levels of specialization and people with odd combinations of skills could do very well though.
I sincerely doubt that ai will take any software jobs. It will definitely raise the floor for engineers but it won’t ruin any job market. If anything it will bolster job markets given that ai can be used for many infustries for very specific reasons. If your job is in threat of being stolen by ai you should rethink how simple your job is. Ai can not handle edge cases with software due to its incapability to think for itself.
You are describing the current publicly available crop of "Expert Systems". Very akin to an extremely autistic idiot savant. If we could freeze the technology here your assessment would be correct. But large blocks of talent and money are and have been pushing for higher. And which among those talents ever predicted that, among the most easily replaceable jobs with first-out-the-gate, Turing-Test positive but not quite AGI souped up chatbots...... Would be the ART majors???? I don't think anyhuman can predict where this all ends up.
@@NullHand Exactly. All the folks I see claiming AI is not going to take any jobs keep thinking about the current publicly availabe generalized AI they are interacting with. They are ignoring the explosion of investment and development that is now going into AI and the specialized use cases being created that will operate on a more advanced level. It's only a matter of a few years before the AI of tomorrow makes today's generalized AI look like an old flip phone compared to a smart phone.
R is mainly used for data analysis python is better. Python is much more ubiquitous and has an enormous amount of libraries and modules already built on top of the base language. Moreover everything that you can do in R you can do in python and more
Coding is really interesting. My sister used to be a coder now she’s a BA. she gets paid more than coders, has way less stress and for the past 3 years been getting huge jump in salaries. I’ve seen her on a few zoom calls and I always hear devs and testers under so much pressure I feel it’s an industry that’s not as fun as it sounds.
Just remember, Tinder's business is money making, not match making... I guess that is every business, business, but that leads to real undesirable outcomes when there is conflicts of interests like this. *Single* people, mostly use Tinder.
Thank you for this video. However, even if Python is going to be the industry standard, it won't be that relevant to be really, really good at python since ChatGPT and other LLMs are very proficient at writing Python code already and the speed at which they are getting better will increase with more CPU resources and additional Tweaks coming in the next two years. I think it is far more valueable to understand customer needs and be able to transfer that knowledge into what is needed to realize solutions that serve these needs. Hereby LLMs and other AI tools are good advisers but there is still someone needed to validate that they implemented the correct thing and correct them if needed and at some point also the knowledge that in case you need it to dive deeper into the code.
I am no expert. However, it seems to me that learning those things, and coding in general, is still a prerequisite for being a "software engineer". I think a better title would be "software solutions professional/expert". At some point, you will not need to code anymore, but you will still need to know how all the pieces work and fit together, and learning the fundamentals together with domain specific knowledge is still entirely necessary. Take someone who knows nothing about programming or software technologies, give them an LLM and see how far they get.
look devs are still needed when the cases are complex or require optimization or low latency or complicated messaging or large architecture or ML which needs all of that , knowing the intricacises will help you spot issues or improve or solve more complex issues however that is 10% of the job market so highly skilled people will be more valued than ever but lower skilled will no be so important
Lol ive started learning cybersecurity too just out if defense for the future but its nice to know this too. My career is trading and its made me good income but i debate whether or not AI could render trading useless
AI is the biggest example of creative destruction in recent memory. It will wipe out many and replace with much much more...unless some party poopers get in the way.
Imagine all of the assembly line workers who were worried when robots started taking their jobs away; they would say "Don't worry, just make yourself more skilled and no robot will be able to replace you on the line." Well, we see how that worked out don't we?
I'm looking into completing, and engineering degree but very tempted to pursue a career into AI have I missed my opportunity as the field appears to be advancing at an alarming rate, would I be better doing a different engineering degree such as mechanical or electronics in a hope that I could integrate AI into those at a later date?
By definition, if we get actual AGI, "web developers" and the like will become extinct. If you're one of these developer people your best bet is to leverage your technical ability/experience and apply it into any role that will directly contribute to the expansion of LLM and any other related tech that will continue to power AGI. I'd also advise taking this very seriously, as AGI seems to be around the corner. Don't listen to folks telling you it's all going to be alright. It simply won't be.
@@simpsimperson73synthetic data will just pollute any model created after gpt3.5. Models today don’t have a complete world model. It’s entirely plausible that gpt4 is essentially the peak as it was already trained on all of the easily accessible data humanity had, and data quality will just be going downhill from here. Different results will require both a different model architecture and alternative sources of data (besides “all of the internet”).
As a guy who runs an AI consultancy firm, this is an illusion. Coding is like 10% of the job. The main problem is the training data (to acquire, label, prepare, etc), and I don't see much improvement in the last 3 years.
would we ever see an improvement from now?
Apparently AI is now creating it's own training data
@@JimMilton1 that is the problem, if you use to prepare the data the quality will not improve…
exactly
@@PracticalAI_
@@JimMilton1 LOL! i'm so excited to see the outcome of this! A super powerful bullshit generator that eats it's own excrement ...what could possibly go wrong?
A lot of non-tech companies, including large ones, don't even have good data governance. They don't have policies regarding the collection, storage, maintenance, and disposal of data. It's sort of a hodge podge of different departments doing their own thing with data. So, even before AI can make a positive impact, these companies need to get their data governance policies in place.
Yes but this is a part of the AI consultant job to advice them how to collect good data.
@@tongobong1 Far easier said then done. There will be a ton of office politics that a good consultant would have to manage in order to leverage the full capability of AI. Department heads won't give up their data (power) easily .
@@Yavin4 yes but as a programmer I can tell you that being good at office politics is more important than being good at programming unfortunately.
@@tongobong1 Most definitely. That's been my experience as a Team Lead. AI is a different animal entirely though. Adaption will lead to people losing their jobs/careers.
Although Im totally with you on data and its importance, I still call to listen to client more. Right now I have a client with absolutely garbage data infrastructure, they have several offers from IT companies that give a perfect plan (step by step, all is really good etc). The only problem is that the client needs quick wins to justify AI/ML investments, and they are not ready to launch a project when they will get feasible results only in like 9 months or more.
So my advise to them now - “ok, lets try some basic cheap AutoML stuff, if you’ll value in, then we do everything right”. And yes, i told them all pros and downsides of such approach
From a Senior Software Engineer at a huge bank: AI is the future... but the future is pure garbage. We are currently having a problem with people even understanding the basics of JSON and YAML and when you throw in an LLM engine that just regurgitates stupidity... well... we are all doomed. Good luck, everyone.
Having worked in tech I agree with the original comment. I don't see how someone with a background in AI learning rudimentary full-stack development helps with the issue @@MARTIN-101
I’m also a senior (cloud) engineer at a large financial company… I don’t think the problem you describe is an AI problem. 🤭
@@MARTIN-101no. It’s the same way a really good set of tools won’t make a bad carpenter any better at building a good house.
That's one of the many issues: AI cannot be double checked for consistency and accuracy in it's answers/outcomes.
So, unless you are a very skilled professional that knows what is BS and what is not, many entry level professionals could potentially get stuck in something that doesn''t work and that they are unable to make it work.
@@MARTIN-101 he will definitely not answer you, because naysayers are always like this.
I've been a data scientist for 20 years and I've never seen Andrew so far off the mark. I have delivered dozens of these solutions in the tail. The model is typically you get a data scientist like myself to find what the thing is. Build out a prototype using open source libraries. Plenty of code there and then you get some data engineers and machine learning engineers in to harden up the code base and scale it
How did you find those tail end jobs?
@@HaiLeQuang They are the most common because most of the ones in the bulk have been commoditized
Can you share what you mean with building „the thing“ as in is there a practical example?
@@john.8805these would be your standard regression and classification models which have to be bespoke to the use case of the specific client. Lead generation. Price prediction. Churn likelihood. Probability to fail.
This is the whole job of a data scientist. I don't understand how this is news to anybody
Actually now you say "Claude, you are data scientist with 20 years experience. Find what the thing is, build out a prototype."
I’ve been thinking about this lately. I told my friend that I was getting into tech and I wanted to start by learning software development. I was surprised by what he told me. He said that’s a basic skill rn and what I need is to either get into cybersecurity or cloud. All these areas use Python as their main language. I started learning python and I’m impressed by the versatility of the language. I mean you can use this language to do almost anything. From data analysis (Python is now in excel), data visualization, network automation you name it!
Yes but you shouldnt run applications off it.
You dont learn programming languages. You just learn "programming" and most languages look the exact same.
python is a jack of all trades, it can do absolute everything but is a master of none. The trade-off for a company is fast deployment and low cost (low cost due to fast deployment, need to pay less hours of working to employees) vs generall "poor" performance. By "poor" it means for very specific situations where speed and stability is critical, like a mobile or web app with millions access, , trillions rows of data, then an AI in Rust for example would perform better at the cost it takes a bit more time to fine tuning it. Or just get Azure, GCP, AWS, and their top tier engineers already did all of this hard work 😄. For small applications, self projects, prototypes, python all the way.
But! Python is getting closer and closer in terms of performance. Getting rid of GIL is a feature is something will be a game changer for python to support the speed of true parallelism. Is already on the table but may take few years to me implemented. A glimpse of this would be Mojo language.
Python is big daddy of all you can go into data science anything you want after python
Getting it into software development you should know Python, Javac C++ and the applications of all of those
Python is a great programming language. It's straightforward, consistent and very intuitive. But occasionally it does throw a curveball, in syntax and in the way it works. It however doesn't give you same sense of freedom as C++ does.
C++ on the other hand allows you to approach any problem in the way you see fit. You can use C++ in very much same way as Python, but you can use C++ in many different ways aswell.
Different tools for different jobs.
Python was based off C++, take that as you will
@@cpK054L I know that and some of it's core functionality relies on c++ not to mention some libraries. They are still very different languages in syntax and in functionality.
I am a freelance developer and I was waiting for these AI tools to start building AI solutions for clients. And here they are, the public LLMs and so forth. I share your view that I can be an AI developer without having to be a machine learning engineer. I really enjoy when you comment on an interesting presentation. Great video !
I took a year of computer programing back maybe before you were born but life took me a different path, retired and thinking I need a winter hobby. thanks, worth a sub...👍✌🖖🥃
I'm a junior dev and i just got a job at a small AI consulting company. Maybe this guy is onto something
Hey folk, i am currently learning js, the concern is by the time i learn full stack with good projects (stuff like nextjs, ShadCN, tailwind, docker, kubernetes, CI/CD), genAI, langchain, deep learning stuff, 2026 or 2027 will come, where the hell will my requirement accomodate in the industry, because all these senior engineers by that time (like you) have already replace junior devs, what value will I provide? Pls help, it's infuriating.....
I love AI, I am a robotics and mechatronics engineer and I hate coding from scratch, so AI is good for that. I prefer to modify code that I get AI to write from scratch and then get AI to check it and optimise it. I like ladder programming and assembly by myself, as I like to run it in my mind step by step.
I have mostly been full stack Java, JavaScript, PHP, Python, SQL but have also done work with Analytics and Data warehouses with Python, SQL, Java, ETL
Ok?
indeed, it appears so
I think in the big picture, python is in a similar place to where BASIC was during the 80's or PASCAL was during the 90's. But just looking at the differences between these three languages shows just how far we've come.
both obsolete and Python will be as well
Can you explain what you mean here?
@@patpat8727 I mean with BASIC we had something that was only really useful for teaching programming and from a pure language standpoint, a pretty bad language (taught a lot of bad habits and hard to scale). Then PASCAL was a step forward since beginners at least got a language which could be used to create large, useful programs even if that didn't happen much in actual industry or with commercial development. Finally today with Python beginners get a language that is easy to scale, has plenty of useful libraries, and is sometimes even used by industry or in the commercial world. So a beginner today is in a far better place than in the 80's in terms of what they can do, the usefulness of what they learn, etc.
I love that he said, "grounding" in Python, LLM and ML... love this pun, not sure how many people got it but I smiled when I heard. If you don't know what "grounding" is....yeah 🙂
beware of entering a position that has a low barrier to entry like "python developer". This means lots of people can become qualified and compete with you, thus driving down job security and pay.
pretty much all of the IT market is inflated by new developers right now
As if there's such thing as "pure python developer" - it's never about python, or any other language, it's about freaking data, design, structure, cloud, and so on... There's NO market for pure python.
@@trykozmaksymHow does one get a job with python knowledge
These are good videos thank you, I am a 2nd degree CompSci career change guy in my 30s; these vids help w optimism and entrepreneurship
Thank you Travis! I'm writing this comment as a commitment to myself that I am going to start and complete your blueprint 🤝
Thank you Travis for this great information
To be honest with you Learning Python is a great way for beginners get into Tech. I love 3 Field in Tech and they are Software engineering, AI and Machine Learning and Cloud engineering and I will make sure I train myself to be an expert in those fields, Am currently learning Python right now and i started 2month ago
Keep at it 👍👍
Thank you, Travis! I've been following you for a while and I'd say the content you produce helps me a lot on my programming journey. Thanks again for sharing valuable knowledge which is quite rare to find.
Having worked as a jr dev at a software company, its difficult for me to see an AI tool that is capable of letting the client build professional websites with little to no code required. Pro websites are far too complicated and have problems all the time. One thing ive learned about SWE is that things break all the time and it takes a deceloper to figure out why. Implementing a fix in the existing code base ranges. At times, I'd think something was so easy to fix but it turned into a much more challenging task. Sorry, I just don't see it. There are tools like WIX that build websites for small businesses but these sites are like the bare minimum in quailty and functionality.
I don't think that's the point. As a dev you would take an existing image recognition neural network and train it on a customer's problem(identify bad quality cheese spreading on a pizza) using their photos(hopefully annotated). Then you would give them an interface or automate the process by making an ai prompt like "is this pizza ok?". If the data isn't annotated you could set up a team of inexpensive remote workers that sit there and annotate data for very low wages. If you recycle your interfaces you're talking about very little development work and these projects become affordable to small businesses.
Python is the defacto language for AI/ Machine Learning/ Data Science. But id argue Mojo will most likely be the future as it is basically a superset of Python.
Definitely possible
Lol tell me you know nothing about python without telling me you know nothing about python
LLVM Python bindings will win the race for performance.
@@alivepenmods Not against. Mojo. Mojo is faster and more performant than Python.
@@vectoralphaSec LLVM IR compilation from Python bindings will be multiple order of magnitude faster than Python and way faster than MOJO, while not requiring new synthax.
Mojo is a hype
Am so glad I came across this channel.New subscriber here, and just as you mentioned I went down that Cloud path with AWS and I even got the foundation CCP certification. Focusing on building my Python skills✅
There has always been developer oportunities for small use cases. Think of the plug in writers for all sorts of applications and use cases. Right now, people are slapping .AI on everything like it's monkfruit or stevia
Gotta experiment.
Update I just spent 5 hours with chat gpt4 creating a JavaScript filtering script for an insurance project I'm working on. It's working exactly the way I wanted it to. I can expandt it. I added features and functionality as it was built. It saved me money . I'm learning how the sausage is made while I talk to ( prompt) the sausage maker. I see and correct mistakes thru prompting. I'll deploy it in production. I now see the future. No more rip off developers. I can do this by myself as I teach myself. Greatest thing since 3D printers and oh yes that's next. As this tech progresses the sky is the limit.
This is already the field of embedded software engineering and control systems engineering, using PID controller, and small computers such as arduino and raspberry pi
in one of my jobs the users always wanted the applications to be as simple as possible, I always told my coworkers they want the button "do my job" so until AIs are able to read the mind of people, we devs have work that will look similar to what we have right now, but even then when everyone has a neuralink and AI can read their minds, people need the discipline to follow the logical steps to build a product/solution & if they want the "do my job" button, then they won't even think about the solution.
Andrew ng was pioneer and had wide knowledge of old ai. His videos helped lot of people. I have not heard him talk anything substantial about Transformers and large language models
Travis, thank you very much for all your videos! I feel like you work with the speed of light as well uploading video after video!! 🙏🙏🙏 I started learning following your web dev 6month blue print(HTML,CSS,JS) doing Colt Steele's course on Udemy. Thank you very much! I really love it and I see my progress every day. I really hope to be able to fit in somewhere with my knowledge and be able to evolve further. I just wanted to thank you again, and please believe me your videos inspired me and gave me confidence🙏🙏🙏 Greetings from Romania!! 🤗😘
🙏 Thank you!
@@TravisMediaclip said great Azure solution and cert but free blueprint link says AWS. Do you have a free blueprint for Azure? Or does the the AWS blueprint apply to both? Great clip all the same
@@TravisMedia Andrew’s last name is pronounced “ung”, with as little emphasis on the u as possible
Andrew Ng: "Ng" is pronounced like "eng" as in the way "Engelbert Humperdinck" is pronounced by Americans. An AI pioneer in his own right, he's also a fellow techno-optimist and effective accelerationist.
But the real question is why does he say piza??????
@@BGivo In Japanese people say Piza. I think Ng is a Vietnamese name though...
I thought the initial vowel was a schwa /ə/, like the vowel in the first syllable of 'Umbridge'
@@BGivo cuz he's not from Chicago :-)
Google Maps is not efficient finding routes, but we get your point Travis.
I agree with the subject on a macro level but why a developer would be more suited to execute this smaller projects than lets say a data science, statistician or pure mathematician (which usually lack on strong developing skills). From what I am seeing the need to have high levels of expertise in developing is diminishing as no code ML platforms are being used, yet you still need deep knowledge on data analysis, understanding results and deriving conclusions which other fields have stronger. Give trash data to the model will throw trash results, no amount of fancy ML will make a model work without domain expertise & the understanding of the math behind.
Since everyone is moving to python, it's the good time to move away from python. 😂😂😂
Interesting. I have tried, many many times over the years (I recently turned 60 yrs old), to learn coding, most often with Python. I've gotten a few books, watched many tutorial videos and started online classes. I usually get to a point where I'm doing a lesson/exercise/lab and get stuck somehow, something just isn't clicking, and I can't figure it out/get passed it and there's no realtime instructor to ask, even via email. Now, I was just personally interested, I'm retired, and I'm not looking to get back in to work. I don't know if my brain is wired to "get it" when it comes to understanding coding. My career was in high-tech, oddly enough, and I worked with a lot of LINUX/UNIX systems, used regular expressions, etc. But it was more from the systems were vendor provided/supported, and we were the in-house "technicians" that handled software upgrades, module replacements, outage recoveries, etc., it wasn't writing code or anything ...
Where do you get stuck? You should divide the problem into several small problems - steps and solve them one by one and test each of them on the way. That way programming is quite easy. You also shouldn't think much about the design of the program. Just do it and then you will refactor and redesign it to make it more clean after you have workable solution.
@@tongobong1 Thank you! Hmm, typically it's several lessons in, variables I think. I think I end up going too long in lessons at one sitting, like several hours, and just get tired and then frustrated, I think I just followed along, monkey-see, monkey-do, but didnt fully grasp everything. I should take a break/let things soak in, and go back to it. It's the only programming I would know. Thank you for the support, I appreciate it!
@@ph5915I propose that you just set yourself a goal to create some kind of a rather simple app in python. Maybe you can create simple calculator app or simple game... The 1. step is to create a window and then 2. add some buttons and then 3. some code to run when you press buttons. To complete each step you can find tutorials, examples and documents online.
The best way to learn is to build something useful or fun because then you learn stuff that you really need to learn at that moment to complete the next step.
This is how you become the real programmer and not just a guy that is learning all the "theory" yet he doesn't know how to use it in practice.
@@tongobong1 Hmm, that's a great idea! I'll have to start over again into Python though, it's been many months. Thank you again!
Also i highly recommend Python crash course book by eric m
I use AI to make Python scripts all the time. I love Python :) So easy!
This is very interesting and kind of aligns with my thoughts or focuses on applying AI (outside of software and consumer tech) but I hate writing Python lol I learned development writing C++ so I’ve always gravitated towards things like Java, JS, C# (kinda) but I guess I should brush up on my Python.
Thank you man. I working as web dev 10 years (js). How I understood - time to learn new field
AGI chips is the new gold
Thanks Travis, great video, and I've been grinding my Python skills for quite some time, hope it all pays off soon!
Hello, how much good are you in python? I've just started and learnt basics. Now I'm about to learn Oop and DS. where should I go next?
I found amusing in the digital era, people thinking that learning how to code will became obsolete.. it will not, maybe the work change a litle bit, but knowing code it will be revelant.. because not all bugs are syntax or programming error, a code could work just fine and still be no good.. and write your on code a lot of the time is more productive than reading someone else code and trying to figure out if is working as intendend.. if ai is like 95% accuracy good luck trying to find the 5%.
I am not a programmer but what you saying about the %5 is true in my case because when I use chatgpt to write a complete code for me, in most of times it writes a good code but still I find that it has a crucial missing parts to it that I keep scratching my head into the wall trying to figure it out and most of times I just give up and move on to another task but however I think this will not last for so long as A.I is getting better and better so it made me change my mind to learn a programming language myself like Python or Kotlin as I think I will be just wasting my time while A.I is moving in light speed
This is my thought exactly but that’s still not optimistic. Basically means ai will code and we become the chatgpt. Like a janitor position. I agree the most annoying bugs are often not related to the code and ai will have trouble with that
One issue I had recently was an app crash and after debugging it just had to do with using google canary browser. Another issue I had , I solved by downloading the recent code commit from GitHub which has the same code but doesn’t crash for some reason (something to do with dev environment). How would ai or newbie figure out issues like this ? The worst bugs have the stupidest solutions. The equivalent of unplugging a computer and plugging it back in. I can’t imagine an ai suggesting solutions like that because it has nothing to do with the code
@@ahmedsalah7474 as a programmer I had gpt code the bulk of my app and became convinced there’s no way a non coder can do this. You’d end up wasting a lot od time and getting stuck. Many times I had absurd issues that would end a non coder when ai can’t help and especially if you have no thought process of how to prompt further or figure it out on your own
I don’t think you’ll waste your time learning the basics and fundamentals. I think nowadays it’s bad idea to learn software engineering but the best time and maybe very important to learn basics and fundamentals of code and engineering. Everyone should know how to code in era of ai . But just not going into software as your expertise
Very good. Just as it has been foretold in the book "The Cyber AI of the Law".
The problem with all these companies is the legacy software. You need a way to enter and edit the data. Where the opportunity is, is creating an AI that can be trained to use these software.
all it needs is a dataset
Everyone is an expert in youtube.
Really enjoying your content. I'm buying you a cup of coffee for the guide.
Not all heroes wear capes!
"AI expert" - translation: a person who has never programmed a computer, but knows more than you, because they once did a business course, and they subscribe to the WallStreetJournal
Have you looked at his credentials?
@@TravisMedia in this case … no :) an exception to the rule for sure.
As a wild generalisation though, you have to admit it’s a pretty good rule of thumb. We do seem to have an epidemic of AI experts these days, and it’s likely to get worse over the next couple of years.
By then, the next misleading buzzword will emerge, and give us a whole army of new experts who still can’t program a computer. Rinse and repeat.
I think this sounds more optimistic than most. Thanks
I really appreciate this video. There is so much doomerism in the tech space right now and this is a genuinely uplifting thing to hear. There's still a chance, you just have to position yourself correctly as always
Inshort we have to do more projects to get more money , one niche will not enough we have to discover several niche.
im a software engineer, not a programmer, and haven't written any code at work in a few months
when AI is able to replace the actual responsibilities of a software engineer, it will be capable of replacing any job
AI it can't really replace all real sw engineers because many aspects of SW engineering is often accessing proprietary hardware and apps like ChatGPT can't possibly understand proprietary hardware. You could write AI models to understand proprietary hardware but you would need to pay AI SW devs to do it. In which case would require even more higher payed SW engineers.
@@chrischoir3594 It is not a case of replacing all SW engineers, its about creating less jobs for them cuz they wont be as competetive as AI
@@KG-ut7kl hence what I stated
@@chrischoir3594bro, will you tell your experience in tech, also what dyt about what should a guy who has 2-3 years learn to become kinda secure and keep up with upcoming market situations?
Thank you for this!!! Just rejuvenated me :)
Thank you so much. I own a small SEO AI SaaS bootstrapped to 10k MRR with no marketing and i really am frustrated by my lack of code knowledge. I plan on learning python now.
Let’s chat. I’m a senior python dev and been wanting to launch an AI saas. We may be able to work with each other
These opportunities are opening up... but for how long? The way I see it, you basically have to stay on top of AI technology and ride the wave for as long as you can. Maybe it'll be 15 years or maybe 3 but eventually AI will have sufficient capability so that there really isn't any middle man. It will listen to your requirements and create what you want. All you can do is try to be as high up on the ladder as you can as it gets pulled up.
I don't like python really much, but Mojo has arrived.
I am learning so much from watching your videos. Thank you so much for sharing this amazing source of information. This one is quite helpful.
The goal in tech is profit per employee and it always has been. So if that long tail was going to provide job opportunities for software engineers, big tech responds by investing in AI to try to cut out as many potential employees as possible. Then you can retain the profit per employee benefits with your ultra-productive, relatively small group of engineers. In the past, companies maybe thought they needed to hire a lot of engineers to show investors that they are growing. Now, there isn't such an expectation. There is a realization amongst tech leadership that maybe the work can be done with far fewer people.
Thank you for this guide sir,
You are one of my "Notification ON" mentor, Your first video I saw was 10 essential certificates... and you are really helping me learn how to approach this new evolution
Since last couple of days I was thinking that there is a major change in the way we will be coding and learning things...
And I was think how can I learn web Development (learning form THe Odin Project, on foundation) journer more ai driven and take a leap in learning these ai tools and combining it with treditional technologise.
Like thare are many no code or low code project using ai tool but in current time we still need unterstand treditional way to code to check weather ai is coding properly.
AI ain’t gonna replace crap. I don’t know AI but as an enterprise developer, I know what we are expecting AI to replace… not gonna happen 😂
I just really want to develop and deploy my own AI assistant. Jarvis!
Great video spot on earned a subscriber
Addressing practical limitations to simply using AI tools... It will be costly, and it will (for a while, at least) lead to far more carbon being released into the atmosphere. From a societal perspective, we are running way behind on renewable energy adoption relative to the speed of progress in AI, with scaling of data centers that release more carbon.
Exactly. When people find out it costs $17/hr to run an LLM they will think twice.
I don’t like everything built just around python. We can call a python web service from a Java backend and make an Angular/ React front end to consume it as required. Because Java is much faster in receiving and retrieving it would be better to use it as a backend. I’ve worked with Streamlit and it is very very slow in giving the data for outlier analysis.
AI Blueprint is not presented in the video, just a link to the blueprint
The trouble with all these online experts talking about software development, especially university lecturers, is that they dont have any real world experience out in industry commercially developing software, and so dont know what they are talking about when it comes to what industry expects. You dont even mention C# or Java which are the worlds industry standards for application development with vast prewritten libraries for reuse. Now why is that? As soon as he said "100s of developers" for measuring the cheese on a pizza it is a giveaway he doesnt know what he is talking about. Feels like all the failed developers out there are salivating on AI taking revenge for their own short comings. lol. Industry expects a design, an architecture, reusable code, meaningful error handling, verifiable automated tests based on code developed, etc, etc., and TO BE IN CONTROL of all parts. AI will have its uses, but not in an informal unstructured language like Python.
Sorry but I’m not as enthusiastic about AI. I believe it will lower the need for software engineers and human design creativity will be taken out of the equation.
Ya Im worried about lowering the need for software engineers too, but i think we are quite a ways from that. Ive found that most people who preach that you will soon lose your code job to AI, are often selling something / inveested in the tech. But also, theres nothing we can do to stop it, so we may as well be ready .
Think the same but do look into new careers that are already here. Like ai engineers and ai operations ( not the very high skilled ml engineers)
@@hikemalliday6007 the VAST majority of Senior developers do not think AI is going to take all the jobs any time soon. It is really only Junior devs that think this is the case.
Which, if you think about it, makes sense. A Junior dev doesn't know much yet and been in the thick of it in development. They get handed tasks like "there is a bug on this page that moves the comment box out of alignment on mobile devices-re-center it on mobile devices without affecting desktop viewing." And that sort of bug is something that can be done pretty slowly by a Junior, possibly fairly quickly by AI, and super fast by a Senior dev.
But what people forget is that a Senior dev can't be bothered to realign a comment box on mobile devices. An office drone wouldn't know how to take ChatGPT's output and put it into the test environment without themselves becoming a Junior dev (though without any programming knowledge). So either the Senior dev has to now do the task, or they're going to need to train the office drone to the competency of a Junior dev who doesn't know how to program anything themselves.
The point of a Junior dev is to push that task to them. The Senior dev is trying to figure out how to stop the database from crashing when the user numbers spike 200% at intermittent intervals throughout the day, without needing to run the server at maximum all day long that would cost the company $20k/month.
Those tasks that get pushed to the Junior is how they become a Senior, so they can replace the current Senior and solve those $20k/month challenges. And the Senior dev needs to be replaced so they can move on to Architect or Tech Lead and identify changes to save $35k/month.
There is a huge shortage of Seniors right now, and the Seniors see absolutely no reason to believe AI is going to take their jobs. Eventually, companies will have to settle and hire more Juniors again. The companies who tried to replace 2 juniors with 1 junior and ChatGPT are going to find themselves in hot water when they need 2 Seniors and only have 1 Junior to promote.
@@holdfast5332Not the same man. What you are talking about is evolution throughout decades or even centuries. What's happening to us now is something that is going to happen well within our lifespan
Don’t be, sure the job is not gonna be there, instead it will evolve. I.e. taxi drivers = ubers. Especially with the younger generation of American citizens refraining from having kids. Ai implementation in something like immigration would take a process that takes years and complete it in a couple months. Logistics, economy, advice, medicine and most importantly, actual people behind security to keep ai safe. If you have an understanding of economics, society and how it has evolved every two or three generations, you’ll get the bigger picture. Doors are opening my friend, you just gotta learn how to walk through them.
If ai is cheaper and better even filling senior positions, the only job left will be entrepreneur. Ai can help with that too. So output will go up tremendously as supply constraints vanish. Our only limitation would be time to consume.
Demand. Which is finite, and will go down when people will no longer have jobs.
Thank you very much. I will do what you said as my mentor
I see the blueprint link for AWS, but not the one for Azure talked about in this video?
Until we reach AGI (which is still a big if), we will always have a need for coders. AI can only regurgitate data that it was trained on. It can't reliably create anything outside of its training data set. It also suffers from hallucinations, causing inaccuracies. The more you take a Jesus take the wheel approach to AI coding, the more you will experience this kind of unreliable, and unhelpful scenario.
Once we have AGI, then I see our jobs going the way of the dinosaurs. That doesn't bother me though, because when we get AGI, it will be able to replace most jobs, and at that point, governments will be forced to change the nature of economics or they will have a revolt on their hands due to all of the unemployed workers with no opportunity for work. We will end up with UBI, or some other economic model in a post-labor economy.
yep. what I think too.
well, Python was made for non-developers, for researchers and such, so is kinda fitting, however all those python libraries are just wrappers on C libraries to get the speed, & any other programming language with a native interface can get access to those as well.
Already saw Ng’s talk. Hope this video helps disseminate
the fact that dude thinks 'relationship' advice can come from Tinder just truly shows how out of touch he is with humans && the real world lol
😆😂
Thx for the blueprint
I am loving the interpretation for “C” students 😂
These opportunities evaporate when/if AGI comes along, though. Maybe that's a year our. Maybe 50 years out. One of the concerning dynamics. If we stick with the current AI capabilities, then there will certainly be plenty of work for devs. I'm not really expecting any plateaus in progress just yet...
Yes, but when AGI comes, there will probably be very few profitable professions at all.
@@shugyosha7924Yes, because manual labour does not scale very well the average earning ceiling will be very low. Outside of that there is owning real estate and businesses...
People cant even agree what this AGI thing is and what defines it, yet everyone so sure "its coming and gonna change everything"...
@@reyne2077Think from the business perspective of the developers. They want to create a software that makes mental human labor obsolete, so they can profit from or even monopolise the new AI driven workforce.
Since most physical labor not yet automated requires expensive robotics (ex: robot plumbers) most usually doesn't count those.
So AGI would have skills that are on par or even better than your average worker. You can argue that this could be a lot of small AI's rather than one hughe one, which could be the case, but the end goal doesn't change.
Each problem has different solution AI can have industry specific templates or libraries that can be can be tailored to solve a problem
Learning python n ML , hope I will land somewhere in 2024
short and good clear opinion
I feel you didn't really distinguish between software development and AI careers
Very helpful. I am grateful
Really nice info thnks..
Nice talk.
Jobs are going.. face it.
subscription mate
your content is nice!
Im 30 years old, doing masters in tax and learning to code. But I had passion for soccer my whole life . I always wondered if i would make a good coach. Guess now is a good a time as any...
why masters in tax and learning code. you could have done masters in AI or something!
That's a really odd combination. Tax and code has almost nothing in common, unless you're planning to work for one of the tax software companies like Turbotax. I think the future will see increasing levels of specialization and people with odd combinations of skills could do very well though.
I am all for job opportunities. Lets see what is it.
I sincerely doubt that ai will take any software jobs. It will definitely raise the floor for engineers but it won’t ruin any job market. If anything it will bolster job markets given that ai can be used for many infustries for very specific reasons. If your job is in threat of being stolen by ai you should rethink how simple your job is. Ai can not handle edge cases with software due to its incapability to think for itself.
You are describing the current publicly available crop of "Expert Systems".
Very akin to an extremely autistic idiot savant.
If we could freeze the technology here your assessment would be correct.
But large blocks of talent and money are and have been pushing for higher.
And which among those talents ever predicted that, among the most easily replaceable jobs with first-out-the-gate, Turing-Test positive but not quite AGI souped up chatbots......
Would be the ART majors????
I don't think anyhuman can predict where this all ends up.
@@NullHand Exactly. All the folks I see claiming AI is not going to take any jobs keep thinking about the current publicly availabe generalized AI they are interacting with. They are ignoring the explosion of investment and development that is now going into AI and the specialized use cases being created that will operate on a more advanced level. It's only a matter of a few years before the AI of tomorrow makes today's generalized AI look like an old flip phone compared to a smart phone.
Do you have an opin ion on the R language vs Python for AI?
R is mainly used for data analysis python is better. Python is much more ubiquitous and has an enormous amount of libraries and modules already built on top of the base language. Moreover everything that you can do in R you can do in python and more
@@chindianajones3742 thank you, i appreciate it
good video!
Coding is really interesting. My sister used to be a coder now she’s a BA. she gets paid more than coders, has way less stress and for the past 3 years been getting huge jump in salaries. I’ve seen her on a few zoom calls and I always hear devs and testers under so much pressure I feel it’s an industry that’s not as fun as it sounds.
Problem solving has its own fun. It is not for the week. Architects and Leads need to be mentally tough
BAs getting paid more than devs? Lol, she’s lying/confused
@@threlonmusketeers9371 it’s true
What is BA?
@@itsmedekaBusiness Analyst
Just remember, Tinder's business is money making, not match making... I guess that is every business, business, but that leads to real undesirable outcomes when there is conflicts of interests like this. *Single* people, mostly use Tinder.
Thank you for this video. However, even if Python is going to be the industry standard, it won't be that relevant to be really, really good at python since ChatGPT and other LLMs are very proficient at writing Python code already and the speed at which they are getting better will increase with more CPU resources and additional Tweaks coming in the next two years. I think it is far more valueable to understand customer needs and be able to transfer that knowledge into what is needed to realize solutions that serve these needs. Hereby LLMs and other AI tools are good advisers but there is still someone needed to validate that they implemented the correct thing and correct them if needed and at some point also the knowledge that in case you need it to dive deeper into the code.
Very good at writing Python?
I see 5 comments saying how to pronounce Ng and all 5 recommend a different pronunciation. 😅
Surprised you didn't mention Mojo!
I've not spent any time hands-on with Mojo, but thanks for the mention as I need to take a look at it. Do you have experience with it? Thoughts?
So do you think that Learning HTML CSS JAVSCRIPT is waste of time?
I am no expert. However, it seems to me that learning those things, and coding in general, is still a prerequisite for being a "software engineer". I think a better title would be "software solutions professional/expert". At some point, you will not need to code anymore, but you will still need to know how all the pieces work and fit together, and learning the fundamentals together with domain specific knowledge is still entirely necessary. Take someone who knows nothing about programming or software technologies, give them an LLM and see how far they get.
look devs are still needed when the cases are complex or require optimization or low latency or complicated messaging or large architecture or ML which needs all of that , knowing the intricacises will help you spot issues or improve or solve more complex issues however that is 10% of the job market so highly skilled people will be more valued than ever but lower skilled will no be so important
Lol ive started learning cybersecurity too just out if defense
for the future but its nice to know this too. My career is trading and its made me good income but i debate whether or not AI could render trading useless
Brother, will you tell how much you make, apologies for directly asking, but from where i can learn how to trade?
It looks like the Machine Learning Career Path link is invalid…
AI is the biggest example of creative destruction in recent memory. It will wipe out many and replace with much much more...unless some party poopers get in the way.
lol. you are such a naive man.... do you believe Santa Claus and the Easter Bunny too?
Imagine all of the assembly line workers who were worried when robots started taking their jobs away; they would say "Don't worry, just make yourself more skilled and no robot will be able to replace you on the line." Well, we see how that worked out don't we?
We do see how that worked out. They have other, better jobs today.
I'm looking into completing, and engineering degree but very tempted to pursue a career into AI have I missed my opportunity as the field appears to be advancing at an alarming rate, would I be better doing a different engineering degree such as mechanical or electronics in a hope that I could integrate AI into those at a later date?
Mat welsh is wrong in many thing. HIs calculation of coding cost by LLM as $1200 is so simplistic and appears very deliberate.
By definition, if we get actual AGI, "web developers" and the like will become extinct. If you're one of these developer people your best bet is to leverage your technical ability/experience and apply it into any role that will directly contribute to the expansion of LLM and any other related tech that will continue to power AGI. I'd also advise taking this very seriously, as AGI seems to be around the corner. Don't listen to folks telling you it's all going to be alright. It simply won't be.
You got me in first half, "AGI seems to be around the corner" lol
@@howl404 If the Q* rumours are true then it's game over. Self-improving AI with limitless synthetic data.
@@simpsimperson73 From the creator of "self driving cars this year", i believe in it
@@howl404 surely this year is the one.
@@simpsimperson73synthetic data will just pollute any model created after gpt3.5. Models today don’t have a complete world model. It’s entirely plausible that gpt4 is essentially the peak as it was already trained on all of the easily accessible data humanity had, and data quality will just be going downhill from here. Different results will require both a different model architecture and alternative sources of data (besides “all of the internet”).
Game over!!!!!!!