@gulli72
I'm a freelance translator and copywriter, and I'm very impressed by AI, not because it makes my work faster, but because most businesses that replace me with AI return six months later and ask me to unfuck their marketing.
@adrianbardan782
we are in the worst of both timelines 1) the AIs can't replace skilled labor 2) skilled labor still losing jobs to AIs 😅
@maximilianorios
I've been in this industry for 32 years, and I cannot agree more with you. I've seen the hype before, but this is the biggest hype by far, claiming to do things almost at the science fiction level. I've been using AI to code for one year by now, and even when I find it very useful, I cannot tell it made me more than 20%-25% faster. In most cases, the gain was 0. In repetitive tasks, and silly bugs finding, it is great.
@adambickford8720
Sharpen your skills for the impending over-correction... and know your worth. There will be a vacuum that needs to be filled and nobody to fill it because of the misdirection.
@stefanoperfetti1977
Brilliant metaphor, I’ll quote you on that. “Competitive programming is like solving a riddle. Software engineering is like solving a murder. A riddle is set up so that it’s possible to answer, or at least to recognize the right answer. A murder is messy and involves real people with agendas”.
@stephenpaul7499
The reason why AI does so well on maths olympiads and competitive coding is because the problems are described perfectly. Describing a problem is the really difficult part.
@PlasticCant
Forget hype, this is a death cult now.
@jonathan.watson
I broke into the industry a little under 2 years ago, firmly junior territory and i've had enough of the slop about AI on youtube i've started clicking ignore to just get it out of my feed. Tired of all doom and gloom/negativity and bull by influencers, it's getting tiring..... Thanks for keeping it real.
@fernandogoncalves7297
Be proud of yourself man, people like you are the reason we still have hope on this industry. Good job on the video and keep it up!!!
@manifold84
Since I am a professor here is my analogy: It is like solving exam questions vs. working a job.
@dadsplains
As a fellow former physicist, maybe an appropriate analogy would be the difference between solving textbook problems versus performing research. Like riddles, the textbook problems have right and wrong answers, and can even be solved with multiple approaches. Research, however, requires judgement on when a problem is "solved". As for your point about companies justifying layoffs with AI, I believe those layoffs were always going to happen. They just allow the myth to live because they have a product to sell. I don't know that leadership really believes in it, unless they are a LOT stupider than I think.
@cindr1353
I don't know how much of these you read, but you hit the nail on the head. This is the same thing being thrown around in medicine right now, to replace doctors. AI can beat doctors at riddles, but medicine is murder. Literally, word for word to what you said, all of the challenges exist there as well.
@nebula1100
I have spent the week using Cursor as it's been encouraged by my company and, for better or for worse, the expectation is that AI is now a part of our daily toolkit (remains to be seen whether the expectation extends to increased productivity). I have to say that I've noticed that prompt engineering has an extremely detrimental effect on my ability to write quality software. First, if I use it to write a class, I then have to read through the whole class myself to proofread everything. I don't know if other programmers experience this, but a lot of times if I didn't write the software myself, I have to really spend a lot of time thinking about what I'm reading since I didn't write it. It's MUCH more difficult to find bugs reading through it, rather than writing the logic out myself. I find it very frustrating, and I find myself getting more frustrated because I don't have productivity gains. Overall, I'd say that using prompt engineering to write all of my code (which, albeit, hasn't been the way I normally use AI to help me program) is extremely frustrating and doesn't save me time. If anything, I have more bugs and more messy code to clean up.
@xtremelinux
I have worked in the computer world since 1992 and AI since around 2008. And everything you have said is spot on. But since we, the ones that know, do not voice as much as the hype from so called influences and other non techy or some that say they understand tech but have no idea what even linux is. Basically a clash between people that don't know and want something to give them the edge as a drug, instead of actually studying, and the ones that know.
@sheldoncooper3966
Software developement is so much more than just generating code snippets for very well defined problem statements. It's about understanding concepts like socket connections, multi threading and load balacing. It's about communicating with managers, customers and other developers. It's about building sophisticated long lasting architecture by following principals such as loose coupling and strong cohesion. It's about fulfilling the customers needs that he is not even able to articulate precisely by himself. It's about using experience and intuition. I am quite sceptical whether all this can be replicated by a mechanism that is just predicting the next token.
@afai264
I like the analogy. If you've ever worked on a long-running project with a huge codebase that many people have worked on over the years a lot of the effort is trying to understand how it works and not break anything when you add new functionality. Are there benchmarks for this kind of task - I guess you could fine-tune a model with the entire codebase, ask it to create a regression test suite first , give it some new functionality to create and see how well it does.
@vsots
My analogy with competitive coding and AI: winning the spelling bee doesn't make you a great novelist.
@skulver
I've been using the analogy that implying the best competitive programmer is the best programmer is as silly as implying the person who has solved the most crosswords is the world's greatest writer, or that the person who has solved the most sudoku's is the world's greatest mathematician. I find the endless hype exhausting, partially because of it's obvious falsity, like the Codex demo. A technology so slow that you have to schedule it to run later and so error prone that in their own demo it introduced a build breaking bug extracting a magic number into a constant - something which had has been trivial to do, instantly, with a right click refactor now for over 20 years - but also that somehow is the end of programming as a career.
@slmille4
The metaphor I like to use is that coding algorithms is like creating a puzzle piece but architecture is like putting together a puzzle. There are lots of puzzle pieces you can make to solve a specific problem, but very few that cleanly fit together with the rest of the puzzle.
@CatarinaSantos-h2q