Lol. The layoffs weren’t just developers, it was different fields too. Layoffs happened due to over hiring , not because chatgbt 3.5 was made public XD
@@dscsd9198 are you slow? where did I say anything about chatgpt and how does that invalidate what I said? You're just reafirming what I said, that there is no developer shortage. Oh, and lol
You obviously associated layoffs with programmers, to which I replied that other roles were also affected by hyper hiring, rather than what most people state, AI Chatgbt lol You dumb, go ask chatgbt how to reply to my comment XD
Well put, its so fucked.lol My only hope is that its oversold as being more capable than it appears in the demo, which isn't far fetched when it comes to tech demo's.
@@NeilStuddYa, but extend that into the future a bit further, and now we have 10 testers, and if there's an issue, and the AI can't fix it, _no one_ can, because no one understands how the code works.
Wait, CEO barfs code on team and then makes the code reviewer write the tests? Yup, C suited confirmed. Also "Explain this code" needing a billion dollar tool to deal with the fact the code is illegible? I'm getting grumpier by the minute. I really like the partner program stuff though. Will save a lot of time vs logging into dashboards of 100 tools.
12:00 This is why new software is slow. People just don't think or optimize. DO NOT SORT if you only need the minimum and/or maximum element !!! You are unnecessarily using a O(n log(n)) when you only need O(n). The problem is not that programmers will be replaced by AI, but if they don't hire good developers who know not just how things can be bone, but if they should be.
Imagine the potential keynotes for the next few years: 2024: No need to code, just tweak what AI generates. 2025: No need to tweak, simply ensure everything runs smoothly. 2026: No need to intervene, just be there to pay for the service. 2027: You... 💀 2028: ... Hearing about these changes, which frame developers' once-creative tasks as repetitive ones, I have to ponder. Today, we applaud as these tasks are automated. But what if, in a year or three, the entirety of coding - planning, design, testing, and all other tasks we believe require a human touch - becomes automated too? Will we keep applauding? Every year, we fine-tune our current models with data from interactions between human users and tasks that AI hasn’t mastered yet. Then, in a year or two, with enough data, we roll out a new model that can take over. It's as simple as drafting a PR message, sifting through issues, tweaking code, running tests, fixing bugs, and deploying-all without any human involvement. Surely, in a few years, an entrepreneur will be able to transform an idea into a fully operational app using tools like Copilot. The number of Github Copilot users won't just be in the millions; it'll hit a billion if things go as Satya Nadella envisions. And there’ll be applause all around! The writing's on the wall: we're approaching a future where software engineers, developers, and designers, as we know them today, may become obsolete. I'm not knocking the trend - as an entrepreneur, I see the benefits. But isn’t it ironic? Programmers cheer when they're told, "You don't need to do this or that anymore," not realizing that eventually, it might turn into, "We don't need you at all." And to be honest, I even used the ChatGPT to improve this comment grammatically!
these theories has been done and proven already, most of the manual labor these days have been replaced with automation. you sound like those hill billie guys from south park always saying "thEY tOoK mE JoB"
@@Xenon-h9z For generations, individuals manually cut logs using a traditional hand saw. However, the landscape changed when we introduced automation to the process, revolutionizing the way we approach woodcutting.
İ mean thank you for highlighting it. However that doesn't diminish a bit on how great and how exciting the news are. What you say is true, however the world is not a snapshot and this iteration of the software is not sealed in stone. the system was, is, and will be, improving over time. what is true today might not be true tomorrow, and that is the key message! This is dope! is not perfect and doesn't need to be, as is a tool to speed up work, not the producer of final writen in stone, stagnant, untouchable code. And this tool is for adult, intelligent and ground in logic humans, not for brand new untrained people.
Every time I saw things like this makes me fell amazing that IT developers are willing to collectively develop a tool eventually replace majority of them..... and then us. Good job.
😂 all developers here don’t want to say it but they know it in their heart. This presentation shows that companies will be able to grow with 0.1 percent of their engineers. Only very high level positions and research scientists are going to have jobs. This is inevitable
what about the code that can't be described in natural language any shorter than writing the whole step by step when writing the code directly is shorter to write and read, why don't we do the same for mathematics and just express all mathematics in natural language what about not writing boilerplate but abstracting it into library
Couldn't you find any better than those washed-out instagram Hollywood-style presenters. The heck!. I could swear that they never wrote a code before!.
No way. Copilot can only expand on your intentions / stuff it's been trained on. Try asking copilot to do something novel without explaining every step to it, and it'll fumble it every time
We are not nearing the end, we are just getting started. "Programming" will go through a transformation in the upcoming years. The AI might write all the code for you, but you still need to think like a programmer to solve problems. Programming will become as diverse as math in the coming years. We will be entering a whole new dimension of problem solving with all the tedious stuff outsourced to AI
Brudda ion even know how to code gud. I proper feel like a caveman being told how fire works. Well, future is bright so I guess it's exciting having all these fun tools to learn and play with.
@@marek19922209 Did you not see how AI did your entire job with the push of a button? Do you really think you will still be paid to push that button in 2-3 years? 😂
12:47 Its so good that she had to abandon Copilots instructions and use the premade demo code :P Well this has been my experience with copilot since day one. Its either getting dumber or my projects get to complicated for it. Anyways, i think copilot needs to be able to track a bigger picture of the project and get better at understanding the objective from that.
If she did a /fix on this, they'd have a massive gain in terms of public perception, as it would be apparent that they are equally as confident with their tools as their are with prepared demo code.
@@IntentStore yea agreed, on stage like that if anything goes wrong it's hard to improvise. but just pointing out the AI tech (which was being demo'd) was just fine
Soooo boring. AI is still dumb as a rock and needs to be told what to do. Better validate against a human coder if you don't know how to code very well. We already saw what a mess AI has done with Microsoft news.
Ze German developers ;). Early adopter here, it’s been certainly interesting. Good at easy things, deceivingly bad at complex things. But I’m all cases it prevents a writer’s block. Hell, I even enjoy writing documentation with copilot as my partner
it makes you wonder: is writing documentation even needed in the future? If you use copilot to write documentation, then any other engineer could use copilot to write similar documentation aka to understand undocumented code.
Tabnine early adopter here. Co-pilot is pitched as the leading bot pair programmer. That will change very quickly once developers realise they can choose from a range of platform agnostic AI coding assistants. Subscription pricing tiers start low and then the classic corporate enterprise tiers get introduced. Once you can't code without it, you're dependent on the bot.
The coming technology is pretty exciting, and will definitely improve our lives in many ways. There is still a lot of hype that is not yet warranted, but probably will be in the long term. As others have mentioned, we need to be cautious since no technology is perfect, nor are humans and the interface is going to be rough at times. Speaking of imperfect technology, what is going on with all of the video artifacts in this recording? 😅
Apple sure has created the next gen of arrogant nerd product launch speeches! Petty cool use cases! Still concerned about my data being incorporated into the god programmer!
why not develop a tool to instruct in natural language line by line. Let code be generated behind the scenes. Show me just the output for my instructions. In other words run the instructions written in natural language straight away. For ex. 1. Ask user for an input 2. Calculate its factorial 3. Show output. 4. Ask user to press q to quit, otherwise go to Line 1 Why should I see codes, if large language model has learnt it. Afterall, we dont see copiled codes in current system. With LLMs why see source code. Let LLM decide behind the scene which langauge, framework etc it will chooce for implementation. In that way, we will acheive a universal language for coding, and headache of learning every this and that language and framework should be over.
Sorry Github. Co-pilot workspace will never replace the necessity for human centered UX design to recommend improvements for humans. Devs, don't let the machines you utilise start to dictate the solutions for you.
"Huge win for engineers" by replacing the engineering with "tell the AI what to do". They're trying to turn software engineering into no more than a data entry job.
@@KaedysKornah There will be managers that try to reduce down to those levels (as they should for any effective function), but these tools are built to accelerate the value through the full dev pipeline, and remove the drudgery of a lot of low level tool-specific voodoo we have to do today. Even after all this is perfect, you need to know what you’re trying to solve with the tools (all of em) available to you as a developer. A good contractor or artisan needs to keep an eye on maturing tools just as much as their craft. At some point focusing on the minutia and not the actual work and value is just being a hipster developer. “Look at all this stuff I setup for the perfect espresso” :) mkay but there are coffee shops too, but cool.
@@c016smith52 You're analogy of a coffee shop is more astute than you may realize. Another would be woodworking. We can successfully mass-produce coffee, just as we can wooden furniture, and yet any connoisseur of coffee, or woodwork, will sharply contest the suggestion that the mass-produced versions are of equivalent quality. And yet, the tradecrafts of hand woodworking, and hand-made drinks, and all sorts of handmade crafts, are largely a dying art. There's unquestionably a _quality_ advantage to that handcrafting, but the _cost savings_ of the mass-produced stuff means that by-in-large, we're stuck with the inferior, if efficiently-produced, mass produced version. I fear that's exactly where software engineering is going. To an extent, it's already been moving down that road, but the introduction of powerful AIs that can largely negate any need for carefully cultivated engineering skill is going to turn what is currently an area of high skill and at times arguably artistic talent into, well, factory mass-production lines of low-skilled labor. And the result will certainly be cost-effective, but it'll also be noticeably inferior, though it will probably remain at least _serviceable._ Yet another skilled trade sacrificed on the alter of mass production.
i can't wait until this is the level of gpt 4.5, this is cool, but I like saying hey build me this function add that function and having it work without knowing any coding. this is one line at a time, I could print that out with one prompt in gpt...
at what point will c-suites and investors admit that they have no idea what a turing machine is or what their limitations are. but hey, i guess if you can fool enough people into paying for this why not
this demo makes my stomach hurt.... great keynote, but im afraid Ai is close the point where companies can have the choice to lay off developers. I guess we just have to continue to learn new skills and continue to learn how to develop on the cloud and be better cloud engineers.... or are we not safe there either? Bring out the SWARM of AI agents! pooof we are gone.
What real work sir. This is not going to completely replace tech jobs. Only 1% of tech jobs are going to remain. High level research scientists and system designers and 1 software engineer to make clicks for generation and supervision. Nothing else
I promise I’m not a Microsoft shill lol but to be completely fair in the last decade or so Microsoft has been really great for open source. Let’s not forget the contributions that brought us typescript and vscode. I know they have a shit past before then but I’ll acknowledge their recent contributions at least
There was a really good interview with the former CEO of GitHub where he explained what happened after Microsoft bought GitHub. They reprioritized the tasks in their backlog by how highly requested things were from the community. Basically trying to maximize the value provided to the community in features or changes that could be implemented and shipped in a day. And they did that each day for like at least a month. Maybe that's anecdotal but actions speak louder than words
@@aoe2geek is that really something to be thankful for? Typescript just adds complexity many times, and performance is slower than normal JS. VSCode... well if you can find a version without telemetry
Hehehe,,,, no whoohoos/ crowd when silent when he said the Enterprise version is 39 dollars per user per month. Why not just give the Business version these added features and make more sales? Ah, Micro$oft.....you'll never learn.
Mhm. Guess programming languages as such and learning them will be very obsolete in the near future. I say this as a guy born 1970 who used to program assembly language.
Speaking a month ahead of this talk. Generating the app skeleton is not as impressive as getting to a point where default configuration is handled by a CLI prompt from the start.
