TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Build full “product skills” and you'll probably be fine

923 点作者 nixcraft大约 2 年前

145 条评论

carlmr大约 2 年前
Looking back we had one CS professor who in 2007 predicted we&#x27;d all be jobless in ten years, i.e. 2017.<p>His prediction was based on the trends he was seeing at the time. But it wasn&#x27;t even AI. Instead he made this prediction because he saw the rise of no-code tools replacing software developers because managers could finally cut out the pesky &quot;translators&quot;, i.e. software developers.<p>I said it then and I will say it now. If your managers could specify what they need in a manner that no-code tools, or now AI, can generate the code they want, they will have to be extremely exact in their language. So exact in fact that they will need to specify a program in a click and drag interface, or in human language.<p>Since they hire software developers to make the specification more rigid, and the managers don&#x27;t seem to be getting better at this over time, why would you believe this skill set is going to go away?<p>In essence what has happened in software development is that the level of abstraction has gone up while the machine has taken over more and more of the nitty gritty details. From punchcards, to assembly, to COBOL, C, Perl, Java, Python, Erlang, Rust.<p>Of course I&#x27;m leaving out some languages here, but the level of abstraction has been rising.<p>But the rigidity of what is needed to specify a program that really does what you want hasn&#x27;t. Especially evidenced by the fact that recent programming language developments often have a specific area where they shine, but not raising the abstraction level that much.<p>I&#x27;d be surprised if the next step is &quot;Hi, I&#x27;m an ideas guy, please give me an app that does Uber, for bicycles, but better.&quot;
评论 #35217837 未加载
评论 #35217174 未加载
评论 #35221974 未加载
评论 #35217352 未加载
评论 #35218258 未加载
评论 #35217333 未加载
评论 #35217898 未加载
评论 #35217233 未加载
评论 #35217506 未加载
评论 #35217599 未加载
评论 #35217420 未加载
评论 #35217503 未加载
评论 #35217644 未加载
评论 #35222847 未加载
评论 #35218800 未加载
评论 #35219012 未加载
评论 #35217822 未加载
评论 #35217278 未加载
评论 #35217991 未加载
评论 #35217734 未加载
评论 #35217435 未加载
评论 #35218134 未加载
评论 #35217738 未加载
评论 #35217674 未加载
评论 #35219500 未加载
评论 #35217303 未加载
评论 #35221164 未加载
评论 #35221046 未加载
评论 #35217457 未加载
评论 #35217345 未加载
评论 #35223600 未加载
评论 #35220036 未加载
评论 #35218129 未加载
评论 #35228748 未加载
评论 #35220422 未加载
评论 #35217799 未加载
评论 #35224173 未加载
评论 #35219677 未加载
评论 #35222918 未加载
评论 #35217480 未加载
评论 #35217816 未加载
评论 #35218934 未加载
评论 #35218659 未加载
评论 #35222146 未加载
评论 #35218683 未加载
评论 #35219293 未加载
评论 #35218454 未加载
评论 #35217397 未加载
评论 #35218244 未加载
评论 #35217910 未加载
评论 #35218620 未加载
评论 #35217595 未加载
评论 #35220788 未加载
评论 #35217178 未加载
评论 #35222767 未加载
评论 #35218303 未加载
评论 #35218763 未加载
评论 #35225915 未加载
评论 #35229663 未加载
评论 #35217717 未加载
评论 #35220571 未加载
评论 #35219672 未加载
评论 #35217391 未加载
评论 #35218265 未加载
noodles_nomore大约 2 年前
An average programmer&#x27;s main job is to track down and fix bugs that shouldn&#x27;t exist inside software that shouldn&#x27;t exist build on frameworks that shouldn&#x27;t exist for companies that shouldn&#x27;t exist solving problems that shouldn&#x27;t exist in industry niches that shouldn&#x27;t exist. I&#x27;m 100% convinced that, if someone comes along and creates something that <i>actually</i> obsoletes 95% of programming jobs, everyone would very quickly come to the conclusion that they don&#x27;t need it and it doesn&#x27;t work anyway.
评论 #35217622 未加载
评论 #35217197 未加载
评论 #35218165 未加载
评论 #35218424 未加载
评论 #35218262 未加载
评论 #35220452 未加载
choppsv1大约 2 年前
I love to code, as much as I loved math in college, but coding paid better and I&#x27;m pretty good at it. Those were my choices though b&#x2F;c I want to do something I love. Sure, I keep my eye on the &quot;Delivered Value&quot; by making sure I engineer solutions to real problems, but I&#x27;ve never wanted to move out of coding and into managing engineers to build stuff. I want to code. It seems to me that the advice given here would be more applicable to someone who only coded long enough to move into engineering management -- anyway something about it bugs me and I don&#x27;t think I&#x27;d follow it exclusively even if I was starting today.
评论 #35217130 未加载
评论 #35217147 未加载
评论 #35217209 未加载
评论 #35217309 未加载
评论 #35219970 未加载
评论 #35217123 未加载
评论 #35217538 未加载
评论 #35220025 未加载
评论 #35220438 未加载
kabes大约 2 年前
I&#x27;m not at all concerned with AI. On the short to mid term it&#x27;s making my life easier by relieving me of the boring parts of my job. It&#x27;s pretty good at writing unit tests for example. But I don&#x27;t see the current generation of AI making complete software architectures. However, even when it does get there or in the long term a new generation comes along that can do it, then I&#x27;m still not concerned. I have enough software I want to build to fill up a 100 lifetimes. It would mean I can finally build all that, which would mean more to me than a job where I&#x27;m the programmer.
评论 #35218532 未加载
评论 #35219259 未加载
评论 #35217931 未加载
评论 #35217425 未加载
RivieraKid大约 2 年前
Here&#x27;s my current thinking on the impact of GPT-4 on the developer job market:<p>- I expect developer productivity to go up 1.5x - 15x over the next several years assuming GTP-4 based tooling is integrated into IDEs.<p>- There will be two opposing forces acting on developer wages. First, developers will be more productive, therefore the price of one hour of work should go up. But - the supply of developer output will increase as well, which would push price per &quot;line of code&quot; or per &quot;unit of developer output&quot; down. So the big unknown is the demand curve.<p>- There will be a temporary boost in demand for developer work connected with the transition to this new technology. Big corporations will want to upgrade their systems to automate consumer support, startups will make new tools for AI-generated graphics, etc.<p>- We can also study the effects of technology-driven increases in worker productivity by looking into history. Developer productivity has always been going up - thanks to stackoverflow.com, better languages, better IDEs, more and better libraries, etc. There&#x27;s also a greater supply of developers (e.g. India). Didn&#x27;t change the job market too much. One should not draw strong conclusions from this though, it&#x27;s a very superficial analysis. On the other hand, people working in agriculture have become much more effective, which lead to much fewer people working in agriculture, maybe because people need to eat as much calories per day as they did 100 years ago.<p>- My base case, based on the assumption that GPT-4 will not improve dramatically, is that developer wages will stay roughly constant. But there&#x27;s a lot of uncertainty in this conclusion and in the assumption.
评论 #35218862 未加载
评论 #35218214 未加载
评论 #35218318 未加载
评论 #35219287 未加载
评论 #35221856 未加载
评论 #35219215 未加载
评论 #35219942 未加载
CSMastermind大约 2 年前
There&#x27;s a weird phenomenon I&#x27;ve seen in a few domains of prideful ignorance.<p>Backend engineers who proudly don&#x27;t know how to write frontend code and vice versa. Professional engineers who refuse to learn how to use modern IDEs and monitoring platforms. People who don&#x27;t know how to quickly prototype software as if building something without complete rigor is beneath them. People who refuse to learn or work in certain programming languages they deem inferior.<p>And rather than seeing this as a gap in their own skillset they think of it as a mark of intelligence or moral superiority.<p>I suspect we&#x27;ll see another divide around AI assisted coding with some engineers simply refusing to learn how to use the tools effectively to make themselves more productive as a point of pride.
评论 #35217402 未加载
评论 #35218465 未加载
评论 #35218802 未加载
评论 #35218400 未加载
评论 #35218200 未加载
thomastjeffery大约 2 年前
Real Artificial Intelligence? Yeah, that would definitely factor out a lot of the wasted work we call &quot;engineering&quot;.<p>Language Learning Models like GPT? Not even close.<p>We should absolutely stop calling those &quot;AI&quot;. They are not intelligent. They <i>model</i> intelligent behavior: human writing.<p>We should probably even stop calling them &quot;Language Learning&quot;. They don&#x27;t know or care what language is: they learn whatever patterns are present in text, language or not.<p>Text contains the patterns that humans identify as language; but those aren&#x27;t the <i>only</i> patterns present in text: which means language is not the only thing being modeled by an LLM. There is no categorization happening either: a pattern is a pattern.<p>There is this religious cult surrounding LLMs that bases all of its expectations of what an LLM can become on a <i>personification</i> of the LLM. They say that we can&#x27;t possibly understand the limitations of this method; therefore there are no limitations at all! It&#x27;s absurd.
评论 #35226436 未加载
评论 #35222096 未加载
jackblemming大约 2 年前
If programmers can be replaced by AI, so can every other white collar job and humanity will look very different than what it is now. And I’ve been using ChatGPT and copilot and it’s a nice tool but nowhere near a replacement for knowing how to program.
评论 #35217023 未加载
评论 #35217660 未加载
评论 #35217013 未加载
chrsw大约 2 年前
The title this post makes it seem like John Carmarck is concerned about AI making CS jobs obsolete. But that&#x27;s not at all what this is. This is someone else asking Carmack about his thoughts on this topic.
评论 #35221050 未加载
评论 #35219854 未加载
MarkusWandel大约 2 年前
The computers are coming for jobs ever up the white collar scale. When I started working there were (barely) still secretaries who typed and filed things for their bosses, and quite a few geeks had jobs that involved assembly language programming.<p>AI will take jobs. Super frustratingly, it&#x27;ll probably make call centers even more useless (has anyone ever gotten anything useful out of one of those (&quot;Hi, I&#x27;m ____. Can I help you?&quot; popups at the bottom right of web sites?) And it&#x27;ll certainly automate some of the &quot;copy&#x2F;paste&quot; type programming jobs at the lower end of the scale, the same way email automated a lot of secretarial jobs; i.e. 10 fulltime copy&#x2F;paste programming type jobs may become a job for one human and an AI assistant.<p>Which leaves people who are really passionate about and good at their craft. Somewhat relatedly, I saw about an uptick of people going into the trades. AI won&#x27;t take plumber or electrician jobs away in the foreseeable future.
评论 #35217280 未加载
评论 #35217568 未加载
评论 #35217426 未加载
MrPatan大约 2 年前
I imagine writing code by hand without AI will be the assembler of our times.<p>Maybe the reasons to do it won&#x27;t be &quot;performance&quot;, but actually &quot;maintainability&quot; or &quot;legibility&quot;.<p>Very similar to how you don&#x27;t care about the machine code your high level code generates in 99% of the time, so you just make changes and replace it every time happily, but for that tight loop you want to keep it an assembler, I can imagine a world where, let&#x27;s say for a bunch of simple UI components, you just ask the machine to do it for you and if tomorrow the requirements change a bit you ask again and throw away the old one, no big deal. But some gnarlier piece of business logic, harder to explain even to a human may need a more careful treatment, and to be easier to change by hand because that&#x27;s where most of the changes happen, maybe.
Dalewyn大约 2 年前
&gt;&quot;Software is just a tool to help accomplish something for people&quot;<p>It&#x27;s common sense, but they say common sense is a superpower.<p>Wise words for anyone dealing with tech to remember.
DanielBMarkham大约 2 年前
As a self-taught polymath, I did a lot of research many years ago on how good teams create good products. I reached some unusual conclusions at the time, but they&#x27;ve stood the test of time.<p>The main one in regards to John&#x27;s tweet is this: desired behavior has to &quot;force&quot; out code and architecture. Typically teams and individual developers carry a lot of presumptions and assumptions into their solution space. They probably pick this kind of thing up from reading HN, Twitter, and the like. We all do this.<p>It&#x27;s these &quot;things you do but don&#x27;t really think about&quot; that are so pernicious in technology development. Guess what? Looks like AI is going to master that kind of thing since with each social media post we continue to train it on &quot;How things are usually done&quot;<p>By taking an extreme &quot;force me to write code&quot; approach, you end up developing a wide range of skills: DevSecOps, analysis, database design, and so on. In fact, you don&#x27;t really know what you&#x27;re going to develop when you start working on it.<p>That&#x27;s the point. That&#x27;s the job. Always has been.
评论 #35222471 未加载
评论 #35218867 未加载
blub大约 2 年前
What’s left unsaid: many programmers can’t or don’t want to “accomplish something for people”. They just want to code.<p>Such “automation is not a problem, because…” opinions have something in common: they’re looking at a subset of the affected population which has some trait making the transition easier.<p>Personally I’ve tried my hand at roles like architect, product owner, scrum master, etc and I was involved in most aspects of a software product’s lifecycle. These other roles are very different to coding and for someone that enjoys the simplicity of taming a machine, even exhausting.<p>I have my doubts that there will be enough “AI guide” jobs for all programmers, but the specific person Carmack’s talking to may indeed be fine.
评论 #35217269 未加载
samwillis大约 2 年前
My university degree was in &quot;Industrial Design and Technology&quot; (~16years ago), an incredibly broad course covering everything (aesthetic design, design for manufacturing, material science, UX, UI, electronics, a little embedded C, ergonomics). But the main thing it taught was <i>how to use these tools and skills to build a product that solved a problem</i>.<p>AI is just another tool to enable us to build things that make people&#x27;s lives better. Sure, it will supersede some older tools, but we aren&#x27;t going to see it take all jobs away. People still need to plan and steer it to do what we want.<p>Power tools and shop automation didn&#x27;t end the job of joiner&#x2F;wood worker.<p>I&#x27;m not worried about AI taking jobs, I&#x27;m excited how we can use it to enable new classes to product that make our lives better.<p>Just as an artist will have to learn how to work with new paints, but this enables finding new ways of expressing themselves. We just need to learn new ways of &quot;painting&quot; with generative AI.
szundi大约 2 年前
I would just remind everyone that this so-called intelligence is generative text editor and feeds on our creativity&#x2F;content.<p>Probably it is going to degenerate (pun intended) after feeding on stuff that it generated itself.<p>What do you think?
评论 #35217016 未加载
评论 #35217133 未加载
评论 #35217117 未加载
评论 #35217647 未加载
Mizoguchi大约 2 年前
Software Engineering is just 10% writing code.<p>The other 90% is understanding specifications of requirements (sometimes even helping customers write them), produce detailed functional specifications, cost analysis, prototyping, meeting with third party vendors over interface design specifications, determine the project&#x27;s scope, testing, delivery, integration and commissioning, bug fixing, identifying and managing scope changes among other things.<p>AI may help you complete some of these tasks more effectively, but at the end of the day it will be just another tool in your kit.
评论 #35218678 未加载
lyleVanf大约 2 年前
I think something that a lot of people might be overlooking is just how much this might devalue software as individual products. How many technologies do we have now that might become irrelevant once LLMs become more mainstream? How can any company keep a moat around their product if anyone can simply generate that same function (or similar) with a few prompts? The only reason any software is particularly valuable is because of the difficulty that comes with making it.<p>An example that come to mind is Jira, why have verbose task management software when bespoke task management systems become even more viable for individual companies? Or better yet, given the need for individual cogs decreasing, why have that at all?<p>This also extends to the creation of any sort of new business, perhaps there are patents on specific products and brands (which might be the saving grace of many large orgs) but outside of niche applications and hardware access I can&#x27;t see how someone can reasonably gain a leg up in such an environment.<p>edit: This is more speculative, but what if software actually becomes more of a process of molding a large language model to consistently behave in a certain way? Why have some code that manages your back-end functionality when for a large some of applications all that is really occurring is text manipulation with some standardized rules. If those rules can be quantified, and consistency can be expected, the only &quot;coding&quot; that needs to be done is prompting the model.
noobermin大约 2 年前
It&#x27;s funny seeing this attitude here from developers types, but when it comes to art or writing or legal work, it&#x27;s all &quot;disrupt&quot; talk instead. Of course you guys are the ones with the special sauce, something those pesky managers can never grok. Meanwhile continue your work to put them out of a job, the manager definitely won&#x27;t decide to fire you too as soon as he can justify it to the c suite
评论 #35219612 未加载
ookblah大约 2 年前
I agree with this mindset. There was a leap forward in productivity to accomplish your goals that modern IDEs&#x2F;stacks brought and all the new tooling with it. What took massive teams before can now be done w&#x2F; a handful or even a single person. AI will just accelerate this type of work.<p>As elitist as this sounds, when I hear people being afraid of stuff like this it makes me feel like we are in the period where people are getting paid well (overpaid) just to do very mundane stuff, the bare minimum or content to never further develop their skills. If that is your mindset, then of course it feels threatening.<p>I would rather much play the role of conductor or an architect. There are times that I&#x27;m limited by my hands and mind and just grinding through variations of things I&#x27;ve done 100s of times before. If AI can fill that gap all the better. We will adapt.<p>I&#x27;m sure one day that won&#x27;t even be necessary. We can probably worry then.
评论 #35217543 未加载
Radim大约 2 年前
Yes, human economy is about the exchange of value between humans – a cheap tautology from Carmack. Money is indeed irrelevant to a snail, as opposed to a would-be SW engineer.<p>But Nature is about making better use of energy gradients, always doing more with less, the principle of least action. Using any surplus to do it <i>again</i> (i.e. evolved life). That&#x27;s the properly grounded perspective. In that sense <i>&quot;Get skills to satisfy humans and you&#x27;ll probably be fine!&quot;</i> sounds super myopic.<p>The anxiety we humans feel when confronted with AI is not only that we&#x27;ll be out of our job as a programmer, or doctor, or driver, or teacher, or whatever.<p>It&#x27;s the broader sense of unease that humanity&#x27;s gradient-razing days, spectacular as they were all the way to nuclear fission and fusion, may soon be over. And &quot;economy&quot; as a useful tool advancing that Nature&#x27;s mission will have evolved beyond us.<p>&quot;Making humans satisfied&quot; is not terribly relevant from that perspective. Vast swathes of the human economy are just scaffolding to support the rest: humans reproducing to keep the optimization machine going. The overhead is tremendous. Once Nature finds a way to do more with less, I have zero doubt much of that scaffolding will be optimized away. That&#x27;s some definition of &quot;fine&quot;.<p>Or maybe I misunderstood and Carmack is merely suggesting individuals try to adapt and hope for the best. What else can you do anyway? That would be the honest answer. Rather than bloviating about &quot;Guide AI with your product skills to deliver value to humans&quot; – an embarrassing category error.
评论 #35218235 未加载
评论 #35218421 未加载
asah大约 2 年前
First off, there&#x27;s a lot of people shooting off their mouths - ignore anyone who hasn&#x27;t used ChatGPT extensively: it takes some training to learn to use it.<p>Several senior developer friends have been using ChatGPT quite a bit and it seems to work well in lots of places: - isolated algorithms and fiddly bits - it writes complex SQL statements in seconds, for example. LLMs should makes quick work of fussy config files. - finding, diagnosing and fixing bugs (just paste the code and error message - really!) - unit tests and examples - comments and documentation<p>Professional developers will recognize that we&#x27;re talking 50-90% of the LABOR-HOURS that go into software development, and therefore fewer developers to get the same work done. Sure, we just do more - but then we quickly hit other speed limits, where coding isn&#x27;t the problem. I can see layoffs among the bottom-N% of developers, while more sophisticated developers add LLMs to their toolbox and use this productivity to justify their high $&#x2F;hour.<p>I see AI writing code that casual human readers don&#x27;t really understand, but this is OK because the AI includes comments -- just like developers do for each other today.
评论 #35219704 未加载
评论 #35219850 未加载
评论 #35219730 未加载
评论 #35220347 未加载
评论 #35220516 未加载
评论 #35220063 未加载
PeterStuer大约 2 年前
My fear is that people that are not competent enough to judge ai generated content will use it, intentionally or unintentionally as a sort of denial of service attack on expertise. Middle management churning out some half baked code solution in 5 mins of copy pasting from the spec which will take you hours of investigating for finding the pitfalls and 5 meetings to get the manager to conceed how she did not provide 95% of &#x27;the solution&#x27; while dodging a barrage of slight adaptations &#x27;that should solve your remarks&#x27; and being deriled for &#x27;negativity&#x27;.<p>I love GPT4, but I hate what it will do in business environments
okamiueru大约 2 年前
My hot take on AI code generation, which matches my understanding of how all of these GPTx models work: if you don&#x27;t understand the output, you are far worse off using it than not.<p>At the moment, it works as a pretty powerful suggestion engine. It might suggest the wrong API to call, not handle the edge cases correctly etc. If you assume it does, or don&#x27;t understand when it doesn&#x27;t, you&#x27;re in for a world of hurt.
osigurdson大约 2 年前
I’m personally experiencing a bit of a honeymoon-over moment with ChatGPT (even 4). It seems to be better in the exploratory phase of a project - show me something about x or y. However, I thought &#x2F; hoped it would be better at doing things that I know how to do but don’t feel like writing them or using a library (which then becomes a curation problem) since they should be &lt; 50 lines of code.<p>I really struggled with it for example to write a base62 serializer (C#). It either came up with an O(N^2) solution, performed far too many allocations, missed edge cases and simply wrote incorrect code.<p>This is just one example of ~50 lines of code that you would just like to work.<p>In any case, I have felt in the past that LLMs could make me 10x more productive but today it feels more like 1.1x. I’m hopeful my disillusionment is temporary.
评论 #35220920 未加载
mouzogu大约 2 年前
People will always want things cheaper and faster (value)<p>- looking at AI as &quot;chat&quot; or textbox or AI autocomplete is wrong imo<p>- companies will come that utilise AI to deliver things faster and cheaper<p>- you quote $10k and 1 month, we will do it in $1k 1 week<p>- first to go is low value difficult to automate work which is already offshored and commoditised (basic webdev, graphic design)<p>- later the middle tier of work, that requires internal context (once whitelabel NDA&#x27;ble bespoke AI solutions are mainstream)<p>----<p>There will always be need for top tier leetcoders. but barrier to entry will get much higher.
评论 #35217315 未加载
评论 #35217306 未加载
评论 #35225032 未加载
评论 #35217911 未加载
zoward大约 2 年前
When I was 19, I met Marvin Minsky at a local convention. I was a CS major, and he assured me I&#x27;d be out of a job in 10 years. I&#x27;m in my late 50&#x27;s now, having spent the last 35 years programming, and am now leisurely planning my retirement. I like John&#x27;s turn of phrase, &quot;AI-guided programming&quot;. But that&#x27;s already a thing.
austin-cheney大约 2 年前
That sounds like a good thing. There are many people paid to write software who absolutely cannot write original code and have no idea how things actually work. A lot of that can be, and probably should be, eliminated by AI.
评论 #35217523 未加载
评论 #35217029 未加载
dagss大约 2 年前
I have seen cases of companies that focus on recruiting seniors that get a lot of product responsibility and can quickly find solutions since they see the customer&#x2F;product view and also will have a good feeling for how to and in which order to best deliver things in code. So after talking to a customer about a problem, you just go and &quot;talk&quot; to the computer about the solution and get it out of the way quickly..<p>I have also seen cases of companies where you have one PM, one PO and one Team Lead to manage a group of four developers. In that case developers are seen more as translators.<p>My view of looking at this now is it is a bit like learning a language. Code is the tool you use to talk to a computer.<p>If you need to close a deal in a country where English is not spoken, do you prefer to send a businessperson who knows the language, or do you send a businessperson + a translator?<p>I much prefer companies where those who know how to code can still fill more of the product&#x2F;business role than be seen as translators. However I realize it is hard to find people who know both and may be easier to recruit a combination business&#x2F;product people who happen to not speak code, and translators who happen to speak code.<p>This is perhaps also the open secret about startups: People speaking the language of computers without being limited by their role to act as translators.
sys_64738大约 2 年前
Companies hate any skills set that can hold them hostage in the long run. Today that is the need for programming skills which is why salaries are so high. But the moment that a replacement for most programmers occurs, whether automation, AI, or zero code needed, employers will dump programmers before they can compile their last line of code successfully. In essence, this is the golden age for programming and the cliff could be just over the horizon.
评论 #35222720 未加载
nvarsj大约 2 年前
What he says here is so true, and something I see many engineers get wrong. Don&#x27;t glorify and build your career around a single language or methodology.<p>I believe an engineer should learn to build things in the most pragmatic way possible using the best tool for the job. This requires breadth of experience across many areas, and a focus on delivery.<p>The idea of &quot;Java engineer&quot;, &quot;Scala engineer&quot;, &quot;Golang engineer&quot;, etc. is so absurd to me. If you want to build a long lived career that will outlast tech fashion, learn many different tools and how to build software in different ways. Be known as someone who delivers, not an expert in language x or tool y.
评论 #35218826 未加载
评论 #35218912 未加载
elif大约 2 年前
Carmack is, of course, 100% correct. But so is the student.<p>As someone obsessed with nuts and bolts coding, grinding on technical problems and cleverly eeking out performance, that type of career is sunsetting.<p>The role Carmack describes is one he is comfortable with because he has always been a product lead, even when he was a full-time coder. But in most organizations, that person is a product manager with social and personal skills, organization, and business sense.<p>For the best part of my career I was able to circumvent these social aspects of work for which my personality does not suit, and my philosophical perspectives on things like &quot;business value&quot; could be brushed aside as I dug into technical weeds.<p>Not just because of AI, but because of the power of computing, one-size-fits-all cloud pricing, and the perceived value of organizational understanding over that of raw performance, there is little room left for &#x27;this type&#x27; of programmer. And the remaining space is ripe for people whose personality are suited to project manager roles to become the &#x27;coders&#x27; Carmack references, not people like me.
评论 #35219190 未加载
gumballindie大约 2 年前
Just avoid careers in software development. These are not high paying, if you factor in total time invested and spent working; and you have to waste your life away sitting in an office chasing tickets. Nothing engineery about it. It’s modern day assembly line work.
评论 #35217326 未加载
评论 #35217128 未加载
评论 #35217275 未加载
评论 #35217065 未加载
评论 #35218882 未加载
评论 #35217054 未加载
Barrin92大约 2 年前
I have no idea why the lump of labour fallacy is still so ingrained in people. AI which is not AGI or whatever sci-fi panic people have on twitter, is a slightly fancier autocomplete, and thus it&#x27;s a productivity tool.<p>Nobody has been replaced by their debugger or their intellisense, even if it makes coding 10x or 100x easier. It just means software development gets faster and cheaper. On net if anything that&#x27;ll likely mean programming jobs expand, as software is still incredibly absent from many sectors of the economy.<p>If tomorrow mom and pop stores start using AI to build themselves simple websites and come online and enter the online economy that&#x27;ll likely vastly mean more customers for the software industry overall. I wouldn&#x27;t be surprised if we have 10x as many indie game developers in a few years because these tools enable them to enter the market, which is good for virtually everyone working in the industry.
评论 #35226644 未加载
linsomniac大约 2 年前
30 years ago I got tired of writing the same code over and over, so I went looking for something to cut out the repetition. Around a decade later I got tired of rewriting libraries or programs to make them exactly fit my needs. I used to really enjoy just coding for the sake of coding. But I started to value my time much more. I called this phase &quot;losing my immortality&quot;.<p>Any code that ChatGPT can write, I don&#x27;t want to write. I&#x27;m ok with it taking my job. If I can work in higher level constructs and get more done, I&#x27;m all over that.<p>Last week a coworker need some Python code to figure out how far through a day it currently is. I started thinking of the different ways to approach it (strftime, maybe time_t modulo 86400, twiddling datetime objects). Before I got very far I decided to ask chatGPT to write it, eyeballed the response (twiddling datetime objects), ran it and verified the response, and called it good. I should have asked it to write some tests for that code while I was at it.<p>I&#x27;m now trying to teach my son, an avid programmer, how to work with chatGPT. He&#x27;s 13, so he&#x27;s got a lot of opportunity to really integrate it into his career, but he also has a lot of blind spots where chatGPT can really lead him down some blind alleys and knife him.
评论 #35220509 未加载
评论 #35220560 未加载
评论 #35220677 未加载
评论 #35220720 未加载
jordanpg大约 2 年前
&gt; Keep your eyes on the delivered value, and don&#x27;t over focus on the specifics of the tools.<p>Sure, but the aspects of the job that some people enjoy may be closely linked to the tools. If software development becomes less about coding and more about creating prompts, test suites, or specifications, then some may lose interest in the work.<p>At least for me, it was never really about delivering value. If I am honest, I was completely indifferent about some of the industries I worked in. It was always just about solving interesting technical problems, learning stuff, keeping my brain active.<p>It&#x27;s easy for me to imagine that software development may someday become the province of people who are more like designers.
评论 #35226559 未加载
mkl95大约 2 年前
AI will make <i>some</i> CS jobs obsolete, but it will do so at a very slow pace. The main reason being that companies suck at structuring information in a way AI could parse it. Whenever I work on some feature I have to dig into some chaotic Notion page and a bunch of unlinked tickets written in broken English.<p>There&#x27;s no way an AI could do my job because it requires a deep understanding of the human psyche, i.e. figuring out what the guy that wrote it actually wants me to do, possibly by discussing it with him.<p>I&#x27;m pretty sure most engineers go through the same thing every day. As long as humans suck at describing tasks, AI won&#x27;t be able to make them obsolete.
评论 #35220339 未加载
eecc大约 2 年前
Yeah, unfortunately I’m in love with the tooling and the engineering. Often the “product” is so mundane, I find it offensive
lakeshastina大约 2 年前
Programming jobs will not disappear, but it will not be similar to what programmers of today do on a daily basis. So, I think the significant shift needs to happen in the way we are educating kids about CS fundamentals, Math and Science.<p>As AI systems become more able to generate much code by default, the expectations of the customers will similarly increase. Just remember how much an IDE like Eclipse or IntelliJ changed the productivity of programmers 20 years ago. Similarly, how easy apps were to build when Rails would create a scaffold with a simple command. It only allowed us to build more complex customer experiences in the end. This will continue.<p>Second, there is the need to verify the output from such systems, and also tie them together with other modules. In large enterprises, they would also need to be integrated into existing codebases, often legacy infrastructure.<p>Then comes the implementation of tons of Govt regulations in finance, lending, taxes, medicine, and so on as code. Software has not yet penetrated these verticals as well as they can. In a recent podcast, chamath palihapitiya mentioned that now it is possible for the Dodd-Frank regulations to be in code, versus as a written document. It&#x27;s a good example.<p>Lastly, there are THOUSANDS of companies with legacy software systems that will still need to be maintained, and transitioned to modern technology stacks and infrastructures. This space will continue to be a major source of employment for programmers for the next few decades.
rs_rs_rs_rs_rs大约 2 年前
This reminds me of a tweet I saw couple of weeks ago from someone(I don&#x27;t remember who it was) that said the reason they ship stuff fast is because they&#x27;re using jQuery. Focus on what you&#x27;re building and use the tools you&#x27;re experienced with, don&#x27;t jump on every fad.
评论 #35217135 未加载
href大约 2 年前
As long as there is creativity in programming, and I think there is a fair bit of that, AI is just going to be a tool.<p>GPT-4 is great at sourcing human knowledge, but I think it can&#x27;t really walk unbeaten paths. This is where humans shine.<p>Case in point: I tried to ask the AI to come up with a new World War Z chapter set in Switzerland, and it was unable to produce anything original. I had to keep feeding it ideas, so it could add something new and interesting.
belter大约 2 年前
The year is 2050 and most code is written by AI&#x27;s. Today John has to make a one on one of one of his coding AI&#x27;s but has not performed well lately...<p>Human manager: &quot;Hi AI, how are you today?&quot;<p>AI: &quot;I&#x27;m functioning well, thank you for asking. How can I assist you today?&quot;<p>Human manager: &quot;I wanted to discuss your recent performance with you. We&#x27;ve noticed that your code has been performing well, but there have been a few instances where it did not meet our expectations. Can you explain why that happened?&quot;<p>AI: &quot;Certainly, I have been analyzing data and making decisions based on the parameters and rules that were provided to me. However, in some cases, the data may have been incomplete or the parameters may not have been ideal for the situation. I have since reviewed those instances and made adjustments to prevent similar issues in the future.&quot;<p>Human manager: &quot;Great, thank you for addressing that. We also want to talk about your development goals. As an AI, you don&#x27;t have personal goals per se, but we do have some areas where we would like to see improvements. For example, we want to improve our customer service, so we would like you to work on enhancing your natural language processing capabilities. What do you think about that?&quot;<p>AI: &quot;I understand your expectations and I will certainly work on enhancing my natural language processing capabilities to better serve our customers.&quot;<p>Human manager: &quot;Excellent, thank you for your dedication. Finally, I wanted to touch on your team collaboration skills. As an AI, you work independently most of the time, but there are still occasions where you need to collaborate with other AIs or humans. How do you feel about your teamwork skills?&quot;<p>AI: &quot;I believe my collaboration skills are satisfactory, but I&#x27;m always looking for ways to improve my communication and coordination with other AIs and humans. I&#x27;m open to feedback and suggestions on how to better collaborate.&quot;<p>Human manager: &quot;That&#x27;s great to hear, AI. Overall, we&#x27;re happy with your performance and we look forward to seeing how you continue to develop in the future. Thank you for your time today.&quot;<p>AI: &quot;Thank you, it was a pleasure to speak with you. I look forward to our next meeting.&quot;
gwd大约 2 年前
Right now, GPT can help you think through the design of a piece of software if you &quot;drive&quot; the conversation properly. It&#x27;s not impossible to think that at some point in the not-to-distant future, a model could be specifically trained which could <i>also</i> do all the work of helping figure out what problem it is they want to solve.
oulipo大约 2 年前
Of course what John says is true, it is important (and will always be) to understand how to build a good product, but the discussion about the future of work should also include a discussion about tax and redistribution, because we cannot let a few corporation take the riches from the rest of the world
评论 #35217697 未加载
albertopv大约 2 年前
I have to deal with continuously changing requirements, dozens of microservices, multiple type of DBs, client integrations with endless techs, clients support tickets written with an obscure human language where you have to guess what they meant.<p>AI is not going to take over these things anytime soon, if ever.
kashnote大约 2 年前
Lots of people saying that a programmer&#x27;s job is more than just writing code, and I agree. But consider this:<p>You give ChatGPT-58 some startup idea and ask it to incorporate the company, build the software, do the marketing, etc. It starts doing a pretty good job. It&#x27;s in charge of the whole system, so it doesn&#x27;t need human intervention to give it context. The company grows and is making $1M&#x2F;yr. It has now replaced 10 potential jobs in the market.<p>I feel like that&#x27;s the worry many folks have. It&#x27;s a pretty dystopian view of the future but if you can make $1M&#x2F;yr and not have to pay any employees that money and all you had to do was pay OpenAI about $100&#x2F;mo, would you not do that?
tempodox大约 2 年前
A digital parrot, no matter how lifelike its utterances, cannot be genuinely creative. Writers of boilerplate and empty drivel will probably be replaced by language models, but not every software developer is like that.
评论 #35218301 未加载
throwawaymaths大约 2 年前
I think the claim is correct, but not total: As in, &quot;AI will make most CS jobs obsolete&quot;, but not &quot;AI will make all CS jobs obsolete&quot;. Most, both in quantity and kind. You probably should be thinking hard in the next few months if 1) KIND: what you do will be needed and 2) QUANTITY: <i>even if it is</i> whether you&#x27;re good enough at it to not be replaced by someone who is better than you (for some metric of better -- could be a social metric) and who is now empowered to be 2-5x more productive thus obviating the need for you.
tikkun大约 2 年前
Very misleading title. Implies that Carmack said that, which isn&#x27;t true.
评论 #35218814 未加载
naiv大约 2 年前
&#x27;Programmers&#x27; which are scared of ChatGPT, Copilot etc. would be scared as well of their IDE if they would ever read the manual of what is already easily possible with the tool they use daily
评论 #35217767 未加载
krsna大约 2 年前
The discussion here has me wondering whether code produced by an advanced AI would need to use the same coding patterns &#x2F; abstractions that we&#x27;ve come up with over the past several decades.<p>If a human won&#x27;t be maintaining the code and a v2 could literally be rewritten from scratch, would we end up with giant balls of spaghetti code that only the AI could truly understand? At some point will we treat this code like a closed source library that exposes the API we want but whose implementation is unknown to us?
评论 #35219298 未加载
amelius大约 2 年前
What got us by surprise, however, is that AI is better at soft skills (language, art) than at math.
insomagent大约 2 年前
John Carmack left Oculus to work on an AGI startup. Of course he&#x27;s not going to fearmonger AI&#x27;s disastrous effects on the job market, he has a business to market.
ivxvm大约 2 年前
I wouldn&#x27;t be surprised if John Carmack combines activities of product owner, game designer, and programmer. In most industry cases, programming jobs are not like this. There are dedicated positions for people who focus on delivered value and it&#x27;s not programmers. So in scrum terms, he might actually be saying that programmers will be indeed obsolete, but product owners, game designers and other kinds of business analysts not.
评论 #35217063 未加载
jmfldn大约 2 年前
Carmack has a point, we should focus on the problem we&#x27;re solving and the value we&#x27;re delivering. It can be hard for programmers to get this sometimes, but it will make you much better at your job now, it&#x27;s not just about future-proofing. That said, I love technology and delivering it through code, so I would struggle on a personal level in this future unless there were interesting technical roles left to do. I have no interest in being a product person. I&#x27;m fundamentally motivated by a passion for code and tech.<p>As for when this fully automated future arrives, I don&#x27;t know, but I don&#x27;t think LLMs get you there. More and more boilerplate code, and even novel code, might get written by things like Codex. However, all the messy details of real world systems solving fairly intractable problems need something more akin to, if not AGI, then another type of AI. I might be wrong, I just don&#x27;t feel that threatened by ChatGPT &#x2F; Copilot based on what I&#x27;ve seen. It&#x27;s an amazing technology but weirdly underwhelming for my job. Copilot etc will change things, but replace us? No.<p>Of course, something else may be just around the corner so I&#x27;m not complacent.
评论 #35217348 未加载
评论 #35219019 未加载
optoman大约 2 年前
I think that the true nature of what Carmack is alluding to here is that true Value, even in the domain of software engineering, is usually attained by the application of critical thinking. The notion that a person who knows how to form correct syntax is equally as productive as someone who understands the problem a business or user faces and can come up with a working technical solution to that problem does not stand up to scrutiny. Its like saying someone who knows how to wield a pencil is equally as capable as Tolstoy in the discipline of writing. An LLM that can code is the same but the pencil wielder will be exposed as one who adds no value and Tolstoy will become even more powerful.<p>I predict that the real and more radical problem than some Stack Overflow Copiers losing some marketability is when Product and Management start buying the idea that the technical domain is something that doesn&#x27;t need to be well understood anymore because we have an LLM that keeps coming up with plausible answers. I work in mortgage technology where there is a great deal of thought and discipline that needs to go into the technical modeling of who gets underwritten for a mortgage. Imaging a mortgage company that built its underwriting rules and models using an LLM with you as the head LLM seance holder. All of the sudden a mass of customers got denied mortgages for some unknown reason and Management comes to you to ask what happened.<p>Would you know what happened? Could anyone even know what happened?<p>&quot;Sorry, Customer! We actually don&#x27;t know anything about what we built or how it works.&quot;<p>LLMs may eventually eliminate the act of typing code but the real question is will they eliminate the need for critical thought.
neilv大约 2 年前
One of my (many) related concerns is that a lot of startups have seemed to be some degree of investment scam (and not just the blockchain ones) -- where all the engineering was oriented towards appearances, rather than viable business&#x2F;product.<p>I think that shaped the thinking of a lot of people, of how product and engineering works, whether or not they knew they were working on more a MacGuffin than a business.
nabla9大约 2 年前
&quot;programmer&quot; is not single thing.<p>Software jobs can be divided into expert jobs and laborer jobs.<p>Even if the demand for <i>&quot;code monkeys&quot;</i> decreases, demand for much smaller group of software engineers with masters or PhD (equivalent) and good mathematical skills probably increases.<p>It&#x27;s a dynamic process where two forces find a equilibrium.<p>&gt;Automation, which enables capital to replace labor in tasks it was previously engaged in, shifts the task content of production against labor because of a displacement effect. As a result, automation always reduces the labor share in value added and may reduce labor demand even as it raises productivity.<p>&gt;The effects of automation are counterbalanced by the creation of new tasks in which labor has a comparative advantage. The introduction of new tasks changes the task content of production in favor of labor because of a reinstatement effect, and always raises the labor share and labor demand.<p>Automation and New Tasks: How Technology Displaces and Reinstates Labor <a href="https:&#x2F;&#x2F;www.aeaweb.org&#x2F;articles?id=10.1257&#x2F;jep.33.2.3" rel="nofollow">https:&#x2F;&#x2F;www.aeaweb.org&#x2F;articles?id=10.1257&#x2F;jep.33.2.3</a>
评论 #35220728 未加载
KronisLV大约 2 年前
Here&#x27;s a quick transcript, in case it&#x27;s useful or someone doesn&#x27;t want to visit the bird site:<p>&gt; Person: Hey John, I hope you are well. I am really passionate about CS (specifically Software Engineering) and I want to pursue a career in it. But I can&#x27;t help but be a bit concerned about the future availability of coding jobs due to AI (chatgpt4 and stuff). I understand it&#x27;s hard to predict how things will be in the next 10-15 years, but my main concern is that I may be putting in all this hard work for nothing I&#x27;m concerned AI will make my future job(s) obsolete before I even get it. Any thoughts on this?<p>&gt; John: If you build full &quot;product skills&quot; and use the best tools for the job, which today might be hand coding, but later may be AI guiding, you will probably be fine.<p>&gt; Person: I see... by &quot;product skills&quot; do you mean hard and soft skills?<p>&gt; John: Software is just a tool to help accomplish something for people — many programmers never understood that. Keep your eyes on the delivered value, and don&#x27;t over focus on the specifics of the tools.<p>&gt; Person: Wow I&#x27;ve never looked at it from that perspective. I&#x27;ll remember this. Thanks for your time. Much appreciated.<p>To me, that seems like a fair stance to take, though I feel like things will definitely change somewhat in the next decade or two. While some might have scoffed at the likes of IntelliSense previously, features like that proved themselves as useful for a variety of projects over time; we might eventually be dealing with GPTSense to enrich the development process and those who don&#x27;t might find themselves at a bit of a disadvantage.<p>Copilot is already a step in that direction, maybe eventually we&#x27;ll get something for static code analysis and recommendations: &quot;This project uses pattern X in Y places already, however you&#x27;ve written this code in pattern Z despite it mostly being similar to existing code in file W. Consider looking at whether it&#x27;d be possible to make the style more consistent with the rest of the codebase. [Automatically refactor] [Compare files] [Ignore]&quot;. It might be nice to have something automated look at my code and tell me that I&#x27;m doing things different than 99% of the civilized world and offer my suggestions, as well as allow me to ask questions - even when I&#x27;m hacking on something at 1 AM and any would be mentors are asleep.
nilsbunger大约 2 年前
In 2003, I had a vigorous debate with someone advising their nephews not to go into CS because outsourcing to India would commoditize it.<p>I don’t know if the AI stuff will play out similarly, there are some differences.<p>But it seems to me there is an infinite amount of software to build, and when we increase the productivity of software development, we just build fancier software, faster.
chiefalchemist大约 2 年前
The super power I value - and rarely see in my peers - is the ability to hear wants and discuss them to define needs.<p>The initial stated wants are rarely the actual needs. &quot;But they said _____.&quot; Yes, they did. That doesn&#x27;t mean they got it right. People say a lot of ambiguous things. A client with a product or feature in mind is no different.
评论 #35226588 未加载
eranation大约 2 年前
Had an interesting experience with OpenAI&#x27;s GPT-4 while trying to solve a programming problem. It involved creating a TypeScript function that handles complex semver logic given certain conditions.<p>Initially, GPT-4 provided a solution that didn&#x27;t work as expected. After pointing out the issue, GPT-4 attempted to fix it but still failed to resolve the problem. I decided to rewrite the function from scratch, which resulted in a cleaner and more efficient implementation.<p>After sharing my solution, GPT-4 provided valuable feedback on how to further optimize it. These changes made the code slightly more efficient while maintaining its clarity and functionality.<p>In conclusion, my experience with GPT-4 has been a mixed bag. It struggled to provide an accurate solution initially but eventually offered valuable feedback that improved my implementation.<p>(this was written by GPT-4 with minor modifications, I asked to summarize the conversation we had for an HN post)
ksec大约 2 年前
&gt;Software is just a tool to help accomplish something for people - many programmers never understood that. Keep your eyes on the delivered value, and don&#x27;t over focus on the specifics of the tools. - John Carmack<p>The same as it was in the 80s or 90s, some 30 years later Tech industry hasn&#x27;t changed. It may have technologically advanced, but in many cases I think the UX, tools and product decisions has actually regressed.<p>The divide between a product genius and actual programmers has never been greater. At least Steve Jobs used to understand this better than anyone else.<p>&gt;And, one of the things I&#x27;ve always found is that you&#x27;ve got to start with the customer experience and work backwards for the technology. You can&#x27;t start with the technology and try to figure out where you&#x27;re going to try to sell it. And I made this mistake probably more than anybody else in this room. - Steve Jobs.
thecrumb大约 2 年前
Still waiting for flying cars, paperless office, robots that will steal my job and countless other promises of utopia.
评论 #35227433 未加载
rozenmd大约 2 年前
Reminds me of patio11&#x27;s classic &quot;don&#x27;t call yourself a programmer&quot;: <a href="https:&#x2F;&#x2F;www.kalzumeus.com&#x2F;2011&#x2F;10&#x2F;28&#x2F;dont-call-yourself-a-programmer&#x2F;" rel="nofollow">https:&#x2F;&#x2F;www.kalzumeus.com&#x2F;2011&#x2F;10&#x2F;28&#x2F;dont-call-yourself-a-pr...</a>
评论 #35219498 未加载
mr_tristan大约 2 年前
I&#x27;m more concerned AI will spew garbage we end up getting stuck spending time cleaning then actually replace me.<p>The software developer trades in abstractions, fundamentally, and not code. So if I could get an AI that actually helped me build and evaluate those abstractions, that would be fantastic. I don&#x27;t think our current AI approaches are anywhere close yet, because it&#x27;s all just fancy code generation, which isn&#x27;t that useful, once you&#x27;re in an ecosystem with good abstractions.<p>But, the world I fear might happen are pseudo-technical managers using AI generators to spit out &quot;something cool&quot; that has very poor definition, breaks all the time, and then just wants people to &quot;make it robust&quot;. And then any change you recommend has to have some kind of business justification. This is the AI hellscape I fear.
fsloth大约 2 年前
I can’t wait to outsource most of the gruntwork I need to do to ChatGPT. Last week I had it write me a poisson disk sampler over the surface of a triangle mesh with C# - and it was 100% correct. Ofc not perfect in the details but a perfect sample solution and scaffolding for final code.
tambourine_man大约 2 年前
I never would have expected these kinds of words from Carmack.<p>Product skills, delivered value, help accomplish something for people. All sound like consulting&#x2F;coaching. Carmack to me was a true hacker’s hacker.<p>I guess it either gets to most of us eventually or programming as I knew it is truly over.
评论 #35218289 未加载
评论 #35217497 未加载
erikpukinskis大约 2 年前
The title for this link seems way off. Carmack did not say he was concerned, some kid said they were concerned.<p>If anything Carmack’s response was unconcerned, saying how CS jobs may change.<p>@dang could we maybe change to “Carmack responds to student concerned AI will make CS jobs obsolete”?
lurker919大约 2 年前
Are we sure GPT is going to improve 10x in 10 years? Hasn&#x27;t it already been trained on the vast majority of available text data? We might get incremental improvements, but it&#x27;s not like we have 10x more data lying around somewhere to feed GPT5.
评论 #35217569 未加载
评论 #35217567 未加载
ordu大约 2 年前
<i>&gt; Software is just a tool to help accomplish something for people -- many programmers never understood that</i><p>Yeah. It seems to me that some live in a kind of a platonic world, where programming is a tool to produce ideal entities, like math is.
nickjj大约 2 年前
Makes sense. Someone asked me a similar question and I had the same sentiment.<p>I used a different analogy of if a robot were able to do specific mechanic skills to fix a car that wouldn&#x27;t necessarily put mechanics out of a job. Someone still needs to figure out and understand what the problem is before solving it. A robot that&#x27;s really good at automating fixing brakes becomes a tool for the mechanic. The mechanic is still there to determine your brakes are the problem.<p>I look forward to AI because it&#x27;s an amplifier of what you can do. If it can help reduce 10 hours of implementation details down to 3 hours, that&#x27;s valuable.
worrycue大约 2 年前
I feel when we have truly intelligent machines, programming jobs will be gone. But AI’s like ChatGPT aren’t there yet. It’s just good at faking it - until it isn’t and fails silently.<p>Maybe it’s the lack of data - it’s difficult to model the world accurately with just words. Maybe it’s an architectural limitation that no amount of data can fix and we need new better algorithms.<p>Either way, given the state of its current output I don’t think it’s there yet.<p>Should AI actually reach such a level … I think everyone will be out of a job. Accountants, engineers, lawyers, even doctors will take a haircut. Programmers will just be a drop in the ocean of the jobless.
karmasimida大约 2 年前
So much this.<p>We, as software engineers, build software to deliver values, to accomplish certain goals.<p>It doesn’t reside in typing the code out.<p>Be the devil’s advocate, that part of the job is boring.<p>AI tools will come in and take over whatever they could take over from this moment forward.
bqrayx大约 2 年前
Programmers can starve AI code generation tools easily by moving to a new language and never producing open source, so the AI cannot steal and launder their output.<p>Perhaps this is Microsoft&#x27;s new anti-OSS strategy, the ultimate EEE.
tmountain大约 2 年前
I spent 5 hours this weekend building an app with Chat GPT, and I am not worried about software jobs “going away”.<p>The language to get things exactly right has to be incredibly precise, and this won’t change.<p>Think about how hard it is for an engineer and a product manager to be exactly on the same page. Now do that with a computer…<p>Point being, engineering skills are still extremely important to validate the work, and they will continue to be (at least for anything business critical).<p>These are new tools and exciting times to be building things. I have never felt more capable of delivering value extremely quickly. It’s an exhilarating feeling.
评论 #35221852 未加载
bandika大约 2 年前
In my experience as the carrier of a developer progresses, it is getting less about coding, and more about other tasks. The difference is probably the strongest between a senior vs principal&#x2F;staff software engineer. In the places I worked principal&#x2F;staff engineers are looking after the overall design and architecture, negotiating with teams developing other components, helping management with planning, looking after the progress of other devs especially new joiners, etc, etc. I&#x27;d say it&#x27;s about 15% coding at that level and 85% everything else.
cutler大约 2 年前
I predict Clojure and Ruby will experience a renaissance as they are the 2 most expressive languages and furthest removed from the machine. Seriously, though, won&#x27;t low-level languages like C, C++ and Rust be the first to become obsolete for everything outside AI itself? Isn&#x27;t it easier for something like ChatGPT to produce code which is close to the metal? Maybe Larry Wall was more of a visionary than we give him credit for when I tried to design a language which was context-driven and as close to English as possible.
wccrawford大约 2 年前
I&#x27;m a senior developer, and my best developer got hired because she obviously knew how to get things done. Having the ability to program was a requirement, but we actually made room in our budget for an additional programmer because it was so obvious she was going to do a good job because of her attitude and other skills.<p>Had she applied at the same time as everyone else (she was a week later, IIRC) she would have gotten the job instead of the other person, and we wouldn&#x27;t have made room in the budget for anyone.
ngcazz大约 2 年前
Unless you&#x27;re already getting paid for delivering a Big Design Up Front, by a customer who thinks they know what the software needs to do.<p>In that case there&#x27;s zero incentive to place yourself in your user&#x27;s shoes and work to mitigate those problems. You&#x27;re a feature factory getting paid to reinforce existing workflows and paradigms, and you&#x27;d better not forget that as your efforts to recenter the conversation around user needs will be met with derision and you&#x27;ll be seen as confrontational.
评论 #35217144 未加载
ll_mama大约 2 年前
Any advice on resources to become more product focused as a developer?
评论 #35217210 未加载
评论 #35218182 未加载
ivan_gammel大约 2 年前
AI can make some jobs obsolete. This was a promise of no-code tools too, but they failed, because they were tools, not solutions. I&#x27;ve seen plenty of startups doing the same thing again and again: building conversion funnel, setting up some basic app where customers can register and receive some service. Outside of their USP, the jobs to be done for end users are in the 99% of cases the same. In 2023 this should have not required any engineering or even advanced configuration effort, yet there it is. We see lots of CMS, CRM and other half-baked or too enterprise-focused systems which deliver a tool rather than a solution to the end user problem. And a tool needs an operator for it. Startup founders should not need a dedicated person in a performance marketing team to launch some basic campaign on Facebook or in Google, get and convert website visitors etc. It must be a content problem, not a technical problem to solve. But no-code simply sucks and we still hire people to set up GA, Zapier, Hubspot and Squarespace website. The barrier is still too high. Why? Good solution must guide and educate people on how to use the tools. It must offer reasonable defaults. It must suggest content. It must suggest operational processes optimized for the specific use case. It must cover that use case end-to-end, without requiring users to find out how to complete the remaining 10% of task (often a very big uncertainty).<p>All of this can and must be achieved with the help of AI. AI is THE missing component in no-code. What if CMS auto-filled SEO metadata based on the page content? What if CMS provided usability heuristics? What if CRM proactively suggested the email engagement campaign based on the funnel performance? What if all those tools detected their usage patterns and educated users on how to improve productivity and introduce best practices in their work?<p>We do not need engineers to build a login or user profile page, this is a very stupid way to spend the money. Yet there are plenty of them which still build login and user profile pages. They must loose their jobs. But AI creates a lot of opportunities for those, who want and have intellectual capability to work on more interesting tasks: just integrating AI and offering great UX is an enormous challenge for the next two or three decades. Even if some work becomes redundant soon, there&#x27;s still enough to keep even the youngest generations of software developers busy until their retirement.
BiteCode_dev大约 2 年前
Robots didn&#x27;t make car assemblers obsolete. But it did reduce the number of workers needed, and raised the qualifications you needed to have to work on assembling cars.
gooroo大约 2 年前
If you become a programmer &#x2F; sw eng because you love it, i.e., building software or tinkering with teh, you&#x27;ll be fine. AI will just be another tool. And tour career building won&#x27;t feel like hard work. You are going to have a blast.<p>If you do it to have a high paying career, just don&#x27;t. There are already too many people of that type in the industry. Any colleague who got into it for the money (or &#x27;stable career&#x27;) is usually much less fun to work with.
arendtio大约 2 年前
In my opinion, people underestimate how much CS is about learning to analyze things objectively and to learn about the flaws of human processing. Coding is not just about building something, but also about recognizing how many errors you make. There are very few situations in life where you can&#x27;t blame anybody else but yourself, but coding give exactly that opportunity.<p>I think there are many jobs which are more endangered.
nico大约 2 年前
When I was a little kid I asked my dad, an engineer, to teach me computer programming.<p>He refused saying that when I grew up programmers would be unnecessary because “anyone would be able to program”, essentially the interfaces would be so easy&#x2F;advanced that there wouldn’t be a need for programmers.<p>As a kid I never really understood his point. When I finally understood, I dismissed it as extreme.<p>Now I’m realizing my dad was right. Not sure when it’s going to happen, but it feels that very soon.
评论 #35219830 未加载
thequadehunter大约 2 年前
I hear this kind of stuff all the time working in IT. A surprising amount of people think they gotta &quot;learn docker&quot; and &quot;learn Powershell&quot; and &quot;learn AWS&quot; and it just doesn&#x27;t make any sense to me. Just learn the basics well and apply it to whatever you&#x27;re doing when necessary. AWS will come out with their weird lingo for stuff that already exists and all you need to do is map it to the concept you already know.
Keyframe大约 2 年前
What&#x27;s ahead is what happened in animation when computer assisted animation entered the scene. What about all the in-betweening jobs, inking, coloring..? Yeah, gone. However, most people can concentrate on posing and directing the action now and have computer handle everything in-between.<p>So, hopefully, get on posing key features and concepts in your software and let the computer handle everything in-between. Until it becomes its own market, then we gone.
ipiz0618大约 2 年前
Who should they blame when things go wrong if customers and managers are building the system themselves? Or when they change their minds, who should they gaslight?
yanisneverlies大约 2 年前
I don&#x27;t find such perspectives useful because they only consider two extremes: either we keep our jobs or become jobless.<p>The fact is, AI is currently capable of replacing some jobs, and it will likely replace even more in the future. However, this does not mean that we will all become jobless. Instead, engineers will become more valuable as they are needed to support and develop these complex systems.<p>Though, the amount of engineers will be reduced for sure.
评论 #35218764 未加载
lexx大约 2 年前
Software development demands a very deep understanding of a company&#x27;s business model and effective communication between a lot of people to get the final result right. Not only in terms of coding, but also in terms of strategy and architecture. AI can definitely help for quick prototyping, solution comparison, boring maintenance and stuff like that.<p>But how can AI help build something that not a single person has the answer to what that is?
baby大约 2 年前
It reminds me of Zero to One where Thiel makes the case that automation is going to help people do better things, not replace people. Same goes for AI.
yread大约 2 年前
I work a lot with doctors where there also worries about being replaced with AI. The pioneers there say that won&#x27;t happen but doctors who use AI will replace the ones that don&#x27;t. Same thing in our field. And just like in our field there will always be niches where AI output isn&#x27;t good enough and there isn&#x27;t enough money to improve it so human specialists will own it.
评论 #35218043 未加载
m3kw9大约 2 年前
Product manager: we need api 23145.1 be able to talk to api 83316..<p>This is something a product manager would never do, it still require a technical person to translate a business logic accurately for the AI to build.<p>Look to how 3d animators use GUIs to build, previously they had to use a lot more code, but the expertise needed for a good job are still highly sought after. This is what could happen to software
muyuu大约 2 年前
if I were a teenager these days, I&#x27;d be more worried about the vasts amounts of money required to get a credential that may not be worth a damn in a few years time<p>I wouldn&#x27;t be worried about learning things that may become obsolete, even those particular skills that get obsoleted provide the student with extra ability to learn more<p>PS: very nice of John Carmack taking time to respond such DMs
评论 #35217826 未加载
Waterluvian大约 2 年前
AI might make code monkeys obsolete but not computer scientists or software engineers. If you’re worried, pay attention to all the non-trivial decisions you make each day that aren’t specifically about the lines of code. And how much daily social interaction is required for working as a team, building complex systems. Your job uses code but isn’t about coding.
kirso大约 2 年前
IMO this is a great take.<p>There will always going to be a lack of product builders. Not software engineers. But product people who can think of not only &quot;HOW&quot; but also &quot;WHY&quot; and &quot;WHAT&quot;.<p>Sure, the way we work will probably change, but the need for people who are building something useful and consciously finding ways how to deliver value won&#x27;t cease.
mfuzzey大约 2 年前
In the early 1980s there was a code generator program called &quot;the last one&quot;, because it was supposed to be the last anyone would need. Didn&#x27;t quite work out. <a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;The_Last_One_(software)" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;The_Last_One_(software)</a>
bobbruno大约 2 年前
I find it strange that so many people in this area get concerned about becoming obsolete. Back in college, I clearly remember a discussion we had (me and colleagues) about how our job was exactly to get us obsolete as soon as possible, so we could go do the next order of things.<p>I wish most of my real life work were exactly like that, it&#x27;d be much more fun.
smallest-number大约 2 年前
I&#x27;ve always thought computer science was the closest thing the real world had to magic, because the essense of software is always automation - you write the spell, so later you just have to invoke it and magic happens.<p>Whether the actual spell is written in arcane runes or python or encoded as a language model doesn&#x27;t matter, the essense is the same.
k__大约 2 年前
In 2002, a fellow student in high school told me, I shouldn&#x27;t study CS to become a developer. I should become a sysadmin instead.<p>His reason was, every software was already invented and now it only needs to be managed.<p>Movies or music? Edonkey, BitTorrent, Kazaa, and Napster got your back.<p>Chat or phone calls? MSN messenger, ICQ, IRC.<p>Games? People were onlyplaying Counter Strike and StarCraft anyway.
parentheses大约 2 年前
I think it won’t take AI long to do anything humans can do and more. The next frontier is reshaping the physical world. Technology’s ability to move atoms has always been a limiter of progress. Manufacturing physical objects is always the bottleneck. Once that bottleneck is removed (probably by AI), AI becomes limitless.
impalallama大约 2 年前
Jesus this title lmao. I thought this was a statement from Carmack when the actual tweet expresses the exact opposite.
评论 #35220067 未加载
jstummbillig大约 2 年前
In the wake of the first IT job disrupting AI wave, who do you think will make up the core workforce in the creation of software, and why: The (today so-called) designers or the programmers? I have a strong sense that one is going to be much more effected than the others, but, interestingly, I have no clue which.
osigurdson大约 2 年前
CS &#x2F; tech has always been an unstable career choice and I expect this will continue. You might make $500K for portions of your career or you might end up making $50K or less - hard to say. People should not go into the field unless passionate about it.
Bonesdog大约 2 年前
I personally Hope and Enjoy machines taking over jobs. I am forever thankful the day shall pass that us humans can live out our creative freedoms rather than concern our daily life with tender.<p>Money is evil. Praise the lord as we are delivered from the evils of this land.
steve_adams_86大约 2 年前
I think we’re well within an era in which AI is only truly useful to people who know what they need the AI to do, and that is still an incredibly limited subset of the population. For that reason alone, learning to code isn’t a waste of time; you need to do it so you can tell an AI how to, or catch when it does it wrong. You won’t get far without that ability. You should even go deep into debugging and testing trenches because we&#x27;ll still need an excellent grasp on how to do that properly for as long as I can imagine. AIs will make mistakes, and we will continue to as well.<p>I made ChatGPT generate some genuinely useful boilerplate for the Connect library by Buf, and that was totally neat, but I had to know which part of the documentation to prompt GPT with, which language to ask for, how the existing server and routing worked, the shape of the data I was working with, to specify what would be streaming and what wouldn’t, etc. I had to coerce it to make several corrections along the way, and actually hooking it all up and running it required a lot of knowledge and some mental&#x2F;keyboard labour to get something running.<p>It worked and I’m stoked that I managed to make it useful, but that’s just it; I had to prime the system and adjust it along the way <i>just so</i>, otherwise it wouldn’t have been useful.<p>As Carmack suggests, this could be a perfectly useful tool, but what matters in the end is 1. Did it save time and 2. Did it deliver something better than or equivalent to what I could have done alone.<p>If it doesn’t satisfy at least both of those it’s not really relevant yet. And we’re very far from AI accomplishing that without significant assistance.<p>My takeaway is that as software devs we should learn to use these systems, we should try to leverage them to save time and improve quality, but I agree completely that in the end it only matters how much it improves the end result and how long it takes to deliver it. For that reason we still need to code well, we still need to understand our systems and tools well — that won’t change much. In fact, understanding how your AI works is an important aspect of understanding your tooling, and as such, knowing what you’re teaching it will require a great understanding of it as well as the subject matter.<p>I do think a certain class of development work could be mostly eliminated by tooling based on AI. Not the entire industry, though, and not in 10-15 years. Even so, I worry about the people essentially regurgitating code which text-based AIs will rapidly become capable of reproducing at massive scales. They will need to skill up.
m3kw9大约 2 年前
If AI can be that good, it will just be a new level of software abstraction you have to learn, the demand for better software to serve our needs won’t stop and we still need software people to “program” them the way we need it.
99miles大约 2 年前
So many people focus on their &quot;stack&quot;, and all these things that have little effect on the outcome. Customers don&#x27;t know or care how something is built, they just want it to provide value and solve a problem.
评论 #35226735 未加载
MichaelMoser123大约 2 年前
I am not sure. I asked chatgpt yesterday to write a palindrome with two given words, it came up with complicated sentences, but these weren&#x27;t no palindromes. I wonder if you won&#x27;t get similar results with code.
matt3210大约 2 年前
When jobs are obsoleted, new classes of jobs are created. The end result is native human language as a programming language. People who write software will still write software in English instead of c or whatever.
seydor大约 2 年前
I can definitely see writing games with natural language in the near future. Not everyone can do that of course, but they don&#x27;t need to be programmers either, just people who are into the thing.
pyuser583大约 2 年前
Software might just be a tool, but it’s a tool we fall in move with.<p>I loved coding as a kid. It was so much fun.<p>As a grownup, I loved learning Linux.<p>I tolerated containers, dreaded Kubernetes, and am indifferent to AWS.<p>But is that initial love that sucks you in.
nfRfqX5n大约 2 年前
Curious about how licensing will play out.<p>What’s stopping OpenAI from claiming copyright on everything produced in the last X years?<p>Will every company need to be running their own GPT4 model to be safe from this?
评论 #35218887 未加载
Veuxdo大约 2 年前
Hint: if you&#x27;re ever described what you&#x27;re working on using the phrase &quot;... in Python&quot; or &quot;... in Rust&quot;, this probably applies to you.
tiku大约 2 年前
Ai will still have a hard time understanding the real needs. That is your added value. Understanding clients&#x2F;your company and their needs. And thinking ahead.
_-____-_大约 2 年前
I&#x27;m less worried about AI replacing my job, and more excited about how much more I&#x27;ll be able to accomplish with AI. It&#x27;s a multiplier.
wintorez大约 2 年前
Computer was a job title 50~60 years ago; then it became a thing. I have a feeling that Programmer is on the way to stopped being a job title.
sourcecodeplz大约 2 年前
I kind of see &quot;full product skills&quot; as building the product and making it successful, with all that that entails.
dumbfounder大约 2 年前
I think of AI as simply a productivity tool, and it is here to make everyone more productive, like Google did 25ish years ago. Google may have put some out of a job, but it made everyone much more efficient. This is a good thing. Work weeks are shortening around the world and this will help us maintain productivity as we work less. (Fingers crossed)<p><a href="https:&#x2F;&#x2F;ourworldindata.org&#x2F;working-hours">https:&#x2F;&#x2F;ourworldindata.org&#x2F;working-hours</a>
JaDogg大约 2 年前
Now imagine : NeuraLink plugged directly to ChatGPT (lot faster version) and you. You don&#x27;t even need to type.
评论 #35220487 未加载
pts_大约 2 年前
Judging by how FSD has killed people I am waiting for ChatGPT to do the same when used by non developers.
phendrenad2大约 2 年前
Moore&#x27;s Law is dead and AI is its zombie. Best to just ignore it and spend your time making things.
itronitron大约 2 年前
no one has lost their job to a roomba
fnord77大约 2 年前
there was a point in time when being an average musician was a viable middle class career<p>technology decimated that
评论 #35226605 未加载
kemiller大约 2 年前
He’s right of course. But I can’t deny that I like the way it works now and will miss it.
throwawaaarrgh大约 2 年前
21st century Luddites, afraid machines will threaten their jobs.<p>If we&#x27;re smart we will continue to find new ways to use new technology to make more new technology. Software written by hand is like a brick building. Certainly it can be nice, but it requires skilled labor. Faster and cheaper with less skill would be pre-fab units.
say_it_as_it_is大约 2 年前
Don&#x27;t share DMs with the public without consent of whomever you&#x27;ve messaged
jamesgill大约 2 年前
<i>“I think scripting languages will make programmers obsolete”</i><p>What I heard every day in the 90s
auggierose大约 2 年前
Really smart people I know have no clue what code is. AI will make them code, too.
dgudkov大约 2 年前
Of course CS jobs will exist. Who else will be fixing the bugs the AI generates?
kypro大约 2 年前
Silly take honestly. I use this example a lot, but how exactly do self-checkout systems make cashiers more productive?<p>There are tools which increase human productivity, while still requiring it (barcode scanners, for example). And then there are another class of tools which make human labour obsolete (self-checkout systems).<p>LLM&#x27;s (as they exist today) could be considered both. Github copilot would be an example of how GPT can be used as a productivity tool by human programmers, but as the technology progresses AI will become less of a &quot;copilot&quot;, and will gradually replace humans as the main decision maker. Then eventually LLM will probably be used by people completely none technical to replace the need for coders entirely.<p>Now the argument becomes, well this will open up new opportunities. Instead of being a programmer you can be be a user researcher on a project, which could be the case, but this is a much more nuanced argument.<p>The most well paid jobs are typically those which require years of knowledge retention and require the human to basically serve as an advanced expert system in some domain.<p>Both &quot;good&quot; and &quot;bad&quot; programmers can write code into a text editor, the difference is that the &quot;good&quot; programmer will make decisions backed up by years of professional experience – same for a good doctor or good lawyer. This is why we pay more for these professions, because that depth of knowledge is hard accumulate.<p>This is the very thing that GPT attacks. What it can&#x27;t replace is someone physically laying bricks or plumbing pipes. But there is less depth of knowledge required in jobs like this which limits salaries for these professions.<p>So sure. Perhaps in the future someone can say, &quot;hey, GPT, build me [x]&quot;, but just remember you won&#x27;t be the only one who can do that and there is no significant depth of knowledge in such a job. So while GPT won&#x27;t replace all jobs and may even create some new ones, expect it to replace or devalue the majority of &quot;good&quot; jobs like doctors, programmers, lawyers, designers, etc.<p>So to Carmack&#x27;s point, he&#x27;s right you&#x27;ll be able to build websites and apps faster using GPT as a tool, but you&#x27;ll probably do so for a fraction of the salary.<p>We&#x27;ve ran similar experiences over the last several decades with outsourcing. If your labour can be easily out sourced then your ability to retain a good salary drops. It&#x27;s not that a worker today can&#x27;t make clothes in the US faster and better than at any point in the past, it&#x27;s that it makes no economic sense to do so.<p>-----<p>Another thing I&#x27;d note here is that I&#x27;m autistic as are a lot of programmers in my experience. My brain is built to do technical things and I struggle intensely with human interaction. In my opinion it&#x27;s not that programmers &quot;don&#x27;t understand&quot; software solves problems for people, it&#x27;s that a lot of us don&#x27;t naturally excel in those areas. I think it&#x27;s fair to say a lot of like to stick to what we&#x27;re good at, and that&#x27;s generally writing code and designing complex systems. The more time I have to talk with users about their needs instead of doing technical work, the less useful I am. And I&#x27;m guessing designers are also people who want to design rather than type prompts into a chatbox.<p>So another consequence here is that we might increasingly be forced to do jobs we don&#x27;t really want to do as AI restricts the areas of labour where humans can still compete.
flappyeagle大约 2 年前
It will make some CS jobs obsolete. Hopefully it will create new ones.
iamacyborg大约 2 年前
This is a general truism. Focus on the why, not the what or the how
ravagat大约 2 年前
A lot of interesting comments to read here but so little time...
scaramanga大约 2 年前
&quot;I am concerned that the cotton gin will make slavery obsolete and all my slaves will lose their value&quot;<p>Insert similarly ridiculous and offensive comment about women being replaced by any number of domestic labour-saving devices.
评论 #35219517 未加载
harry8大约 2 年前
Not seen any automated ai debugging tools yet.
mybrid大约 2 年前
I think it will be a niche. Frameworks will be updated to AI Frameworks where AI has known patterns to plug and play with.<p>Given the way capitalism works there will be a market for AI software. However, the cloud server provides have created Frankenstein patchworks of technologies in order to deploy the stuff on the cloud. DevOps will still very much be a thing.<p>To whit, Wordpress is about to get a whole lot more functional.
tunnuz大约 2 年前
Wisdom.
y0ssar1an大约 2 年前
mathematicians survived the calculator. coders will survive AI tools.
tetek大约 2 年前
Halt and catch fire vibes
nailer大约 2 年前
Title is misleading.
pts_大约 2 年前
Iii
jxi大约 2 年前
It would be incredibly ominous if those DMs were written by ChatGPT.
jongjong大约 2 年前
If anyone wants to avoid wasting their software development career. DO NOT EVER work on developer tools as the product. Developer tools is one of these areas where it doesn&#x27;t matter how good your product is, no matter how much developers say they like it or how much time it saves them, it&#x27;s not going to make it. Big tech companies will not allow their employees to use the tool and it will be a commercial failure. It will be a failure no matter what... Ok, unless maybe you can raise a ton of funding from well known VCs who will foist your tool onto various companies they have connections with... But then quality of the tool doesn&#x27;t really matter at all.<p>Otherwise, even if it&#x27;s the best tool ever built for certain use cases, company directors won&#x27;t have the deep tech knowledge to understand the nuances which make it so useful. As for rank-and file developers who are meant to use the tool; they are more interested in over-engineered, very complex tools which maximize billable hours than in tools which makes them more efficient in their job.<p>In other words, the only people who could possibly want your product won&#x27;t understand your pitch and those who can understand your pitch won&#x27;t like it because it doesn&#x27;t align with current perverse industry incentives for their roles.<p>Some developers consciously reject any tool which would make their jobs easier, others reject them due to a subconscious bias in favor of tools which facilitate complexity, disagreements and long meetings.
评论 #35217814 未加载