14 ms·
Programming won’t be automated, or it already has been
- endergen 10y agoProgramming doesn't need automation to reduce a companies needs for programmers. Cloud computing, better open source projects, and basically anything we are excited about reduces how many programmers you need in order to build a set of product features in comparison with say even 5 years ago, go 10 years back and it gets worse.
- adrianN 10y agoLuckily, there seems to be a near endless amount of things that can be improved with software, so that more automation doesn't mean that programmers lose their job, instead we just solve more problems with software.
- astrodust 10y agoIt's not programming that needs to be "automated" but the project process. If you could shave 30% off of meeting times you can probably boost productivity by 50% or more.
- Michielvv 10y agoAt the same time, what is expected from those solutions in terms of features, usability, security and reliability also grows. So I'm not sure it will get worse, it will however change the skill set required. As it has been changing since programming exists.
- braveo 10y agoThat's not programming, that's devops. I think we can all agree that the more we automate deployments and the like, the better, but when people talk about automating "programming", they are absolute not talking about automating pushing new versions to the cloud, etc.
- solomatov 10y agoIMO, devops is just a special kind of programming.
- craigvn 10y agoIMO, devops is a special kind of sysadmin, one where you get rid of the sysadmin and tell the developer to do it.
- solomatov 10y agoDevops is a highly automated sysadmin job. When we automate something, we write code. When we write code, we want to make it maintainable, which developers are usually good at, but sysadmins are not. Most of good devops who I worked with came to this field from development, not from sysadmin.
- braveo 10y agoAnd yet, we don't continue calling a sysadmin turned programmer a sysadmin, so why would you insist that we call a programmer who turns sysadmin a programmer? It doesn't really matter where they came from, or what's involved in the job, they're sys admins. No one calls a physicist a programer because the simulation software they wrote could've been written better by a professional developer, they just call them physicists.
- solomatov 10y agoI think, the difference between devops and programmer is similar to the difference between a programmer in OOP and a programmer in a functional language. Or a programmer in a commonly used programming language and programmer in DSL.
- braveo 10y agoI think the difference between devops and a programmer is that they're 2 different titles that describe two different responsibilities in the long term maintenance of software inside a company.
- openasocket 10y agoYes and no. Yes, with those things the productivity per programmer is much higher. But the trend seems to be that increased productivity per programmer simply increases the work demanded to create products. The easier it is to add features to a project, the more features customers (and thus managers) will demand. This whole "everything will be automated and we'll all be out of work" argument I think is similar to the mistake made by Malthus, who famously predicted the world would become overpopulated and be unable to feed itself by the 20th century. There are several mistakes in his work, but one of the most prominent is not accounting for the fact that agricultural technology, and thus yields, also increase exponentially.
- cortesoft 10y agohttps://en.wikipedia.org/wiki/Jevons_paradox https://en.wikipedia.org/wiki/Jevons_paradox
- HillaryBriss 10y agoas efficiencies are introduced to the ecosystem, human demands on the ecosystem grow too
- _audakel 10y ago>To completely remove programmers from the equation would require essentially a human level artificial intelligence. And if I start seeing near sentient robots walking around, my first thought is certainly not going to be, “oh no, it’s going to take my job!” Even with all DeepMind can do we still are very far from anything remotely human intelligence. On the extreme low end, the tiny worm C. Elegans has only 302 neurons in its nervous system. The full circuit has been completely mapped out for over a decade, and still no one knows how it works. Bits and pieces are understood, but that is all. The fruit fly has a meager 100,000 neurons in its nervous system
- olewhalehunter 10y agoMost human intelligence is used to solve problems created by other humans who are doing tasks that can be automated.
- cookiecaper 10y agoEven human intelligence is not always correct. Serious misunderstandings and miscommunications occur all the time. If we get to the level where instructing a computer is just as simple as instructing a human, we're still going to need people who have the job of ensuring that the instructions were correctly received and are being correctly executed and who fix things when they go wrong. What's the saying? Something like: "To err is human; to err a million times in one second, you need a computer." Computers are going to need caretakers and, if we're at all serious about usage, we have to recognize that computers will need a special hyper-specific dialect to define what we need done with adequate precision. Human language is not designed to handle such specificity. That dialect will need its own experts. Today, we call those who wield such dialects "programmers".
- BrailleHunting 10y agoThere is no reason to assume human programmers are somehow special apart from all other disciplines and tasks, except maybe to delude oneself with a false sense of (job) security. Computers using AI, deep learning and natural language processing will eventually be able to program themselves. This is the next, big revolution that is inevitable.
- heynowletsgo 10y agoFinally some damn sense on this topic. The whole "programmers will be automated away" is just a scare tactic to slow down the automation of what can be automated. People scared and afraid and attacking what they perceive as the enemy. Lots of things will be automated, automating will not be one of them.
- bleachedsleet 10y agoI've often found that making broad statements about what will or won't happen in a hypothetical future often leads one to embarrassment. Example: "mankind will never set foot on the moon...how absurd!"
- HillaryBriss 10y agook, yes, but cut me some slack. when i said that, i was stuck on the side of the road next to my model T with three flats and i was reflecting on the poor quality of leather they'd used in the tires.
- heynowletsgo 10y agoThat's true, but I draw the line at people creating a higher life form, and that's what it would take to replace programmers.
- goatlover 10y agoAlternatively, we'll have flying cars and fusion in 20 years.
- jimbokun 10y agoAs programmers, we are the automaters. Programming is precisely taking ill-specified tasks and automating their completion by a computer.
- beaconstudios 10y agotechnically, that's business analysis and it's part and parcel of most day-to-day programming jobs. Automating processes does not inherently require typing logical statements into a text file - only our current conception of "how to tell a machine what to do" requires this.
- danso 10y agoFrom my perspective, programming is the work of automating the mechanical parts of human work and experience, whether it be something as simple as FizzBuzz, or a web scraper, to DeepFace. Novice, and even experienced programmers seem to forget that the appeal of a system like DeepFace or ReCAPTCHA is the ability to apply it at a massive automateD scale, not the bespoke ability itself to detect or classify human faces/behavior. I think by definition, programming would seem to be among the final group of human activities to be automated.
- solomatov 10y agoI think, deep learning is only one part of universal AI. What these system are good at are visual processing, we don't have really good (comparable to what a human can do) models for language, logical reasoning etc. Concerning programming automation, consider modern IDEs, they automate so much, that for me it's very hard to program without them. All this automation is quite intelligent but not human level. If we consider latest experimental programming languages, for example, Agda, its environment has limited ability to complete code based on its type. If we can make this functionality more efficient, it's theoretically possible to find small programs which satisfy specification, though writing correct specification isn't easy.
- dang 10y agoDavid Parnas in 1985: Throughout my career in computing I have heard people claim that the solution to the software problem is automatic programming. All that one has to do is write the specifications for the software, and the computer will find a program [...] The oldest paper known to me that discusses automatic programming was written in the 1940s by Saul Gorn when he was working at the Aberdeen Proving Ground. This paper, entitled “Is Automatic Programming Feasible?” was classified for a while. It answered the question positively. At that time, programs were fed into computers on paper tapes. The programmer worked the punch directly and actually looked at the holes in the tape. I have seen programmers “patch” programs by literally patching the paper tape. The automatic programming system considered by Gorn in that paper was an assembler in today’s terminology. All that one would have to do with his automatic programming system would be to write a code such as CLA, and the computer would automatically punch the proper holes in the tape. In this way, the programmer’s task would be performed automatically by the computer. In later years the phrase was used to refer to program generation from languages such as IT, FORTRAN, and ALGOL. In each case, the programmer entered a specification of what he wanted, and the computer produced the program in the language of the machine. In short, automatic programming always has been a euphemism for programming with a higher-level language than was then available to the programmer. Research in automatic programming is simply research in the implementation of higher-level programming languages. http://web.stanford.edu/class/cs99r/readings/parnas1.pdf http://web.stanford.edu/class/cs99r/readings/parnas1.pdf
- return0 10y agoWhat happens when that 'higher level language' is the everyday language? Everyone knows it
- aianus 10y agoHow many people can describe an algorithm properly in English? Probably the same set of people that can program a computer.
- Zeebrommer 10y ago
- michaelbuckbee 10y agoThings don't have to be fully automated to mean big changes. The factory that makes Ikea bookcases employs twice the number of people it did in the 1980s but they make 37x as many bookcases [1] We're looking at a similar situation in the programming, ops, and IT support areas. A friend of mine Tweeted this the other night: "I'm sleepless so I just created a voice-based health tracker using Google Assistant in about 20 minutes that took me wks to write 3 yrs ago." [2] I've heard numerous anecdotes of "A Zapier account and 20 minutes of fiddling with things just replaced a SAAS app/consultant, etc. we were paying for." This is what automation looks like: the top 10% in a profession doing 100x the work previously done. To date the appetite for more and more software development (eating the world) has sufficed, but there is no guarantee that will always be the case. 1 - http://www.bbc.com/news/business-38747485 http://www.bbc.com/news/business-38747485 2 - https://twitter.com/Raelshark/status/843776684407111680 https://twitter.com/Raelshark/status/843776684407111680
- jordwest 10y agoI don't understand the constant focus on the upcoming automation crunch. Efficiency improvements (through automation and otherwise) have been happening over the past several decades - more output with less jobs - where the majority of the benefits go to the owners of the clockwork. Now we're in a situation where we produce most goods that society needs (and a lot that we don't need) yet there are fewer jobs. This is absolutely going to get worse, but we shouldn't be thinking of robots as any different to the market forces that have been going on for years. Something has to give.
- macawfish 10y agoPeople like Ray Kurzweil prophecy about a time when human "intelligence" is replaced in value. But what about human work? It is something deeper and at least as precious, our freedom to be physically integrated with one another and and with the physical environment we are embedded in. Karl Marx, etc already wrote about a kind of devastating "singularity" that is here and now at the roots of many social, physical and spiritual ills.
- craigvn 10y agoHaving been programming for a long time, automation has been talked about for a long time and never really happened. We have had CASE tools, RAD tools and so on, but users demand for new features and capabilities increases at the same rate as developer tool productivity increases meaning the developers are still required.
- glangdale 10y agoThe idea that "automating away programmers" == "building some ridiculous tool that allows the pointy-haired boss to program with 'plain English'" is a strawman. Programmers don't have to be fully automated away for automation to happen. Suppose I have a "Really Good Compiler" for a "Pretty Nice Language" (let's not start a language war by discussing which language is closest to this, or which exact features are in it). This "Really Good Compiler" has a great optimizer, the "Pretty Nice Language" has a garbage collector that's really good for the task, etc. The PNL standard library has a huge collection of well-tuned data structures and a generics system that actually works etc. The PNL toolchain also has a ton of really effective static and dynamic analysis tools to find lots of bugs that its type system didn't. Meanwhile the folks down the hall in a hypothetical competing startup are programming in C using pcc, like they just teleported from the 80s. How many staff do they need to do the same job, even assuming that they are people just as smart as we are (despite their strange choice of tools?) We have managed to be largely blind to this because the amount of work that's available to programmers has expanded hugely to compensate for increased productivity - plus our tools aren't quite as nice as the hypothetical I described. On a related note, I found a great code sequence for something I'd been thinking about for a while, by specifying what the sequence was meant to do, and specifying the semantics of a few SIMD instructions, and turning Z3 loose. Not quite ready to fire myself though.
- learc83 10y ago>We have managed to be largely blind to this because the amount of work that's available to programmers has expanded hugely to compensate for increased productivity The increased complexity allowed by increasing levels of abstraction is a huge part of that increase in demand. When we move up the abstraction chain we can generally build more complex systems. People then come up with new ways to use the increased capabilities to solve new problems. I don't think this is going to stop unless we develop strong AI.
- mseebach 10y agoQuoting PG used to be all the rage around here, but it seems to have gone somewhat out of fashion. Anyway, he basically credits the power of Lisp with the success of Viaweb, in exactly the way you suggest: If other companies didn't want to use Lisp, so much the better. It might give us a technological edge, and we needed all the help we could get. When we started Viaweb, we had no experience in business. We didn't know anything about marketing, or hiring people, or raising money, or getting customers. Neither of us had ever even had what you would call a real job. The only thing we were good at was writing software. We hoped that would save us. Any advantage we could get in the software department, we would take. http://www.paulgraham.com/avg.html http://www.paulgraham.com/avg.html
- derefr 10y agoI've always understood "the automation of programming" to refer to essentially what happens in this scene of Star Trek TNG: https://www.youtube.com/watch?v=lPLiaYD1lUo https://www.youtube.com/watch?v=lPLiaYD1lUo In other words: constraint-based design, in an interactive context-sensitive dialogue, with the machine-agent able to resolve change requests by—among other strategies—discovering and consuming new third-party components it was previously unaware of. Such a tool essentially presents the same interface that a contract programmer presents to their client: tell me what you want, I'll build it; tell me how it's wrong, I'll fix it. The article brushes the concept of such a level of automation aside as requiring near-human intelligence. But, unlike a human programmer, said tool wouldn't need do the hard part—requirements-gathering—for you, or have any sense of what "sensible" requirements would be. Where human software architects guide their clients into specifying requirements, this sort of tool would be much like the compilers of today: it would take you literally and give you what you asked for, rather than what you wanted, until you ask for exactly the right thing. It would just do it very fast, such that this could be the basis of an effective feedback loop.
- jimbokun 10y agoSure, once we have a holodeck that can understand arbitrary requests and make them physical reality, I might be worried about finding work as a programmer.
- goatlover 10y agoBut even on Star Trek the crew is often portrayed as writing some form of code. I don't know how accurate that would be once you can have computers that do such a fantastic job of extrapolating simple requests into extremely realistic world settings, but then again, stuff seems to go wrong often enough on the holodeck, or with the transporter, or some spatial anomaly requiring tweaking of the deflector dish, so maybe you still need programmers in the 24th century. I do like how just telling the holodeck computer to make an opponent challenging for Data in the Sherlock setting was enough to grant sentience to the Moriarty character. That's pretty impressive.
- cortesoft 10y ago
- brilliantcode 10y agoI feel like at the rate we are going we will eventually reach the Visual Basic days where a monolithic IDE will lower the barrier for front-end development so much that anyone could tinker with it and produce stellar applications with little to no programming knowledge. The only difference from Visual Basic days is that now we have ML and AI to further simplify and do the thinking for us. Designers will be hit the hardest. Followed by developers.
- petra 10y agoWe are getting close to that futureat least for business apps) with the low-code trend in the Enterprise.
- brilliantcode 10y agoany examples?
- visarga 10y ago> Designers will be hit the hardest. Followed by developers. They have already been hit hard, by Wordpress and Facebook.
- goatlover 10y agoLotus Notes was a thing at one point. It didn't seem to stick around. Excel would sort of fall in that category.
- agreedasdfs4 10y agoThese discussions usually conflate "programming, a viable career option" with "programming, an intellectual human endeavor". The viable career option will, IMO, unquestionably vanish for the majority. Manufacturing and working at McDonalds were once seen as a viable career path. They've been streamlined and automated such that it requires only a fraction of the human labor it did a generation ago. Google and the like will make this happen for computing as well. We'll only need so many Evernote like apps. I'd wager we'll have reduced GUI interaction in general, such that a lot of JS/HTML/CSS, and similar coding work will be gone. We're abstracting things into higher level languages. TODAY I can, with my voice alone, tell my phone to perform a truly fascinating amount of work compared to a decade ago. When programming as a career is reduced to filling out text files in something like YAML and having ML go to work, only the bleeding edge, the mathiest of the mathheads, will have skills to command a living wage.
- LeanderK 10y agoAren't compilers and runtimes automating programming (well...parts of it)?
- EGreg 10y agoI would like for more people to talk about automation as REDUCING the need for as many workers, than an all-or nothing REPLACING workers. Then it won't be a straw man of "it will never replace what I do" or "well, people will find other jobs." Some will, some won't. It's not all or nothing, but the AVERAGE DEMAND FOR HUMAN LABOR will go down, which means wages will progressively become worse as a means of delivering money to the masses. You already see it now, but few people state it in such clear terms. We need to stop thinking in terms of "well, why doesn't everyone just buy a ticket out of the ghetto" and realize that some will, and many won't. For those many, we will need single payer healthcare, single payer food, rent, movies, etc. You can call it basic income. Whatever you like. But it's inevitable.
- goatlover 10y agoCan't imagine the fierce debates that are coming over that, particularly in countries like the US, which have strong ideological aversions to such things.
- mdpopescu 10y ago> wages will progressively become worse Given that history - especially recent history - has shown the opposite trend, I wouldn't worry too much about it.
- EGreg 10y agoLink backing up your claim? Ummm https://www.dougsguides.com/content/whats https://www.dougsguides.com/content/whats Have you even read Piketty
- partycoder 10y agoA job is role responsible for conducting productive activities. Today, some of those activities can be automated, some others can't. The tasks that cannot be automated usually require human level performance. As AI advances, when it achieves human level performance in a task, that task can then be automated. This means that fewer people is required to meet the same productivity goals. Some people argue that automation allows humans to focus in higher level tasks, using bank tellers as an example. Bank tellers remain relevant even after the deployment of automatic teller machines. I think this is just temporary. When every activity expected of a job is automated, the job can be fully automated and the human becomes redundant. In this category you have: elevator operators, telephone switchboard operators, human calculators, etc. Soon: truck drivers, dog walkers, etc. Automating programming is going to be a thing. Just enumerate the activities a programmer performs and see how many of them can be automated. My hypothesis is that our jobs will first become negotiating requirements with a computer, and then the computers will fully take over. Once it's done once, then it's about serializing their trained state and deploying that same system many times. Then, it's over for humans: A professional programmer requires 20+ years to train. Then, can only work 8 hours, gets distracted and has lots of benefits/compensation and rights, including the right to leave you at any time. In contrast, AI will do whatever you say with no objections, and if your project is late then you can spin up 30 more AI programmers that know exactly the same to work.
- visarga 10y ago> Just enumerate the activities a programmer performs and see how many of them can be automated. I decide the instructions and the robot pushes the buttons for me?
- partycoder 10y agoMore like requirement analysis, design, implementation, maintenance, testing, deployment... Those can also be broken into activities, and some of them can be automated.
- bcheung 10y agoI assume they mean in the general artificial intelligence sense? Most people don't really think of it in these terms, but aren't functions and macros just automation for programming? You define it once and then it handles it from there. You "taught" it how to handle something.
- markbnj 10y ago>> Moreso, if the goal is to produce reusable applications, we need this to be even more abstract: “Create a plugin that downloads bank statements, compares to the expense statements, and produces a standard report”. But if the computer could do that, who would need a plugin?
- toss1941 10y agoI'm with the "or it already has been". Automation is always happening with programming. Fully automatic programming has no meaning, since you have to define parameters for some set of algorithms to act upon. Once you define parameters, you are already using automation because today we can already with a single line effectively "open file x for appending, and create it if it doesn't exist" and thousands or tens of thousands of lines of code will execute behind the scenes to make this happen.
- mdpopescu 10y agoThis reminds me of the time when I saw the C code required to display a window in Windows, with all the things we've come to expect from one (close "x", system menu and so on). At the time (Windows 95 I think) it was around 80 lines of code. I was surprised because in Delphi 2 that was a File / New Form followed by a Show or something like that... I never even thought of the API calls required to do all that.
- Animats 10y agoWe used to have more automated web design than we have now. Remember Dreamweaver and Front Page?
- mackan_swe 10y agoDo you all mean to say that nowhere in the near future will it be possible for a business person to speak out loud to their computer, asking it to show him on screen the latest sales figures regarding the new line of products he released yesterday? "Give me the latest sales figurs for the new XLine line." We coudn't parse that into "exec sales_report '2017-03-23', 'xline-123' thus rendering the data analyst job you used to have obsolete? EDIT: Google will easily solve the problem of speach-to-code, or someone else will if Google's not interested once they "crack" NLP, and they will. But what does it mean to crack NLP? Well, they already have the means to build the perfect model. I love the word2vec idea and I think there can be innovation still, standing on the shoulders of that discovery. What would a perfect model mean? With a perfect model you would be able to spot concepts with unflawed precision and be able to translate those concepts, with unflawed precision, into whatever language you have. It's perfectly doable.
- goatlover 10y agoHow is that much of an advance over what we currently do? If Excel isn't rendering data science jobs obsolete, then being able to tell the computer you need some sales figures isn't going to either. To get rid of data science jobs, you would need the computer to be able to translate from business speak to generating code for all the things data scientists actually do, which probably involves something more than Excel. For starters, that would be cleaning and wrangling data into a format that a tool like Excel could be meaningfully used on.
- paganel 10y ago> "Give me the latest sales figurs for the new XLine line." What does "the latest" mean? The latest 24 hours? The last 7 days? The last financial quarter? Next thing you know you'll be asking the user (the business person from your example) to spell out pseudo-code to the machine.
- marmshallow 10y agoSurprised nobody has mentioned the halting problem [1] yet, which logically proves programming can never be automated. [1] https://i.imgur.com/HVeO7ek.png https://i.imgur.com/HVeO7ek.png (CH 8, Algorithms - Sanjoy Dasgupta, Christos H. Papadimitriou, and Umesh V. Vazirani)
- josephg 10y agoThe halting problem doesn't prove that at all. It just shows that there exist programs who's output can't be predicted except by running them (potentially forever). It obviously doesn't prove that all programs have that property. I mean, we write programs which provide useful output then halt every day. And if humans can write code, a sufficiently advanced AI will be able to write code too.
- Longwelwind 10y agoYou can use Rice's theorem[1], which is a generalized version of the Halting problem. I don't believe this proves that programming can't be automated, though. Humans too aren't able to prove completely that what they write is correct, it doesn't stop them from making programs. [1] https://en.wikipedia.org/wiki/Rice%27s_theorem https://en.wikipedia.org/wiki/Rice%27s_theorem
- BrailleHunting 10y agoSorry, wrong. Deep learning can and will take over everything, given enough computing power and profit motive. http://www.ted.com/talks/jeremy_howard_the_wonderful_and_terrifying_implications_of_computers_that_can_learn http://www.ted.com/talks/jeremy_howard_the_wonderful_and_ter... Eventually (30-150 years), there will be AI pottery artists selling kitsch, easily-offended toasters and corporations without wetware upper management.
- goatlover 10y agoMarvin Minsky has pointed out that modern AI methods lack common sense, which is one of the requirements for general purpose AI. Rodney Brooks agreed with him.
- victor9000 10y agoIf programming will be automated then who writes and maintains the automation software?
- thewhitetulip 10y agoI have been hearing this for a long time now, "automation will take away lot of jobs", I don't understand the people who make such statements. Do they realize how much manual work is done in projects in even the biggest companies? Do they realize that the internal tool which caused AWS downtime last month was not built in one go, it was built eventually, that's why it is hardly perfect. If we move to a more automated env where we have high quality softwares like the author mentions, it'll just enable us to deliver projects faster, bring better software in the world. Plus, who is going to build these automation tools? surely every company is going to have a different requirement, considering that for the last 30yrs we haven't yet discovered how to build OTC softwares, where the main software is a generic one which can be customized across domains, it remains to see the "automation scare" and who'll sell robots who do everything that humans do. Plus, I don't know why this doesn't get discussed, the day we have a general AI which can think for itself, why would it work for us?
- braveo 10y ago> What they are, indeed, is dynamic. The lookups you are talking about are done in runtime, so the result of these lookups can vary depending on the state in memory at a given time. You mean... gasp The lookups are ambiguous? WHAT!?!?! no way.... there's no way, it's impossible to refer to that as ambiguity, the word ambiguity can ONLY be used to describe the syntax, everyone knows that, just as everyone knows it's impossible to have an ambiguous syntax... only... don't go ask the RoR people about ambiguity between their libraries and the runtime, you'll find that they've experienced exactly such a thing... but what do I know, right?
- pvg 10y agoApparently what you don't know is what 'ambiguous' means. Take a look at: https://en.wikibooks.org/wiki/Introduction_to_Programming_Languages/Ambiguity https://en.wikibooks.org/wiki/Introduction_to_Programming_La...
- braveo 10y agoThat's called an ambiguous grammer, we're discussing ambiguity in the language. Stop being a neckbeard.
- pvg 10y agoIt not what ambiguity means in a programming language. Sure, you can make up your own terminology and call whatever you're talking about 'mauve'. You'll probably have a hard time convincing people dynamic programming languages are mauve as well. It's 'grammar', incidentally.
- braveo 10y agoAlright, you're young. No reasonable person is going to buy that the following quote was in the context of formal grammar's. > Todays dynamically typed languages are already too ambiguous for most people to write correct software without tests This is the quote you responded to initially, by the way. If you read that as grammatical ambiguity, you read it wrong, and this entire misunderstanding is really a testament to your reading comprehension. But whatever, it's not my problem, it's yours, you'll fix it or you won't.
- _pmf_ 10y agoThe focus of software engineering research should be on requirements elicitation, analysis and transformation. There has basically been no progress at all.
- UweSchmidt 10y agoExactly. But this is not fun, and being the next level requirements elicitator is not a prestigious career goal. Authoring a new javascript framework is.
- _pmf_ 10y agoIt's completely incomprehensible to me why research pussyfoots around the large elephant in the room. Commercial requirements tools (that are amenable to partial transformation into inputs for data driven code generation) do exist (DOORS, ReqIF), but there's a distinct lack of FOSS communities, which is something I cannot understand. The rise of the agile methodology has IMO worsened the problem by basically making throwing in the towel the default approach. (Throw in some "but, but, you're doing agile wrong!" here.)
- UweSchmidt 10y agoPeople don't want to do requirement engineering for their FOSS project, of course - that's the boring part. To actually provide value the requirement world should offer a higher abstraction level - say, map requirements on "business objects" that are already implemented and configurable, and build software out of those. Workflow descriptions, relations between entities should result in actual code. Otherwise, you know why doing agile sounds attractive to many developers. Otherwise
- samblr 10y agoHow many 'DO NOT' believe that programming will be brain interfaced in distant (25 yrs) future ? where 'program-thinking' brains 'can' result in new 'products' ? edit: 25yrs
- ilaksh 10y agoI agree with most of that, but I think based on current trends we should anticipate that human-like ("human level") artificial general intelligence will be created within less than 10 years. And not too long after that happens, we must assume that there will be AIs that are tuned or turbo-charged so that they can perform better than humans in specific areas like programming. And not too long after that, there will be competition between multiple such AIs being sold to replace humans. And then prices will be reduced and you will be able to buy 10 of these 50X AI programmers for the cost of one human programmer. Then no one will have human programmer employees except for the novelty of it.
- paol 10y ago> I think based on current trends we should anticipate that human-like ("human level") artificial general intelligence will be created within less than 10 years Why? People have been saying this since the 1960s, so much so it's become a running joke. And some of them were very smart people too, like Marvin Minsky. What is it about AI that creates this illusion of progress (or at least, wild overestimation of progress)?
- ilaksh 10y ago"Wild overestimation of progress" has been part of the story for many technologies. Minsky was very optimistic, and then he got burned by the AI winter, and became very pessimistic about AI. There is quite a lot of very serious investment aimed at this type of 'real' AI now. For example Deep Mind, DARPA L2M, recent $100 billion Vision Fund that Softbank's founder said was intended to bring about the singularity. I believe it because I can basically see how to handle most of the challenges using embodiment or virtual embodiment and interactive learning/L2M etc., by combining or taking inspiration from various techniques that have been published. Stuff like this: Overcoming catastrophic forgetting in neural networks (adding plasticity to NN models) http://www.pnas.org/content/early/2017/03/13/1611835114.full http://www.pnas.org/content/early/2017/03/13/1611835114.full "Decoupled Neural Interfaces using Synthetic Gradients" https://arxiv.org/pdf/1608.05343.pdf https://arxiv.org/pdf/1608.05343.pdf, somewhat similar to autoencoder/decoder pair use in "Feynman Machine: The Universal Dynamical Systems Computer" https://arxiv.org/pdf/1609.03971.pdf https://arxiv.org/pdf/1609.03971.pdf "Safe Baby AGI" http://agi-conf.org/2015/wp-content/uploads/2015/07/agi15_bieger.pdf http://agi-conf.org/2015/wp-content/uploads/2015/07/agi15_bi... (cognitive architecture) "OpenPsi: Realizing D¨orner’s ”Psi” Cognitive Model in the OpenCog Integrative AGI Architecture" http://goertzel.org/OpenPsi_agi_11.pdf http://goertzel.org/OpenPsi_agi_11.pdf
- mpg33 10y agoProgramming seems like a thing that would be ripe for automation...
- threatofrain 10y agoThe ultimate programming interface is a declarative one, where an authority tells an AI to build a program for them. One might say that abstractly, from the business perspective, the programmer is an expensive AI licensed for use by government. The executor or authority declares what they want, and the machine attempts to build a program with what specifications it was given.