Slashdot Mirror
Dry.io Wants To Democratize Software Development Using AI (venturebeat.com)
An anonymous reader writes: We've seen companies big and small build everything from AI-driven developer tools to AI-powered developer environments. But what if instead of having AI merely help developers write code, it did all the heavy lifting? Dry.io, a developer playground that helps you write web apps using just a few lines of code, began accepting signups today for its first wave of external testing. The programmable software platform lets you set the parameters of what you want to build, "and the AI takes care of the rest."
That is all.
Simple webapps are just barely software development. If your use case is this simple, Ruby on Rails pretty much does this already and you don't really need an experienced software developer involved. More AI buzzword crap.
"Founded in April 2018, Dry.io has not raised any money" - Shouldn't that be kind of a red flag or something? Maybe they should let the AI run the business side too?
I have heard about the death of programming for years. Since the mid 90's people have been telling me that software that can write software better than programmers can is just around the corner. I'm still waiting. Development tools have gotten better and newer languages are certainly easier and faster to develop on, although they don't result in faster code. Right now AI is little more than an industry buzzword. It isn't real yet, not in the way its marketed at least. Don't expect anyone to change this soon.
I can't get to the page that allows me to create a login/password.
I've been coding for 30 years, and for the last 5 years I could see this coming. Recently I've been telling people that there is maybe 2 more years left in this field before the door starts to close. Some types of work will continue, but the overall era of throwing rooms full of coders at software will have ended. Surprised it took this long; much of my work for the last 10 years has been a process of cut/paste from my earlier work, or just Goolge a question, follow a link to Stackoverflow, read a few posts for 10 minutes, and only then copy/paste. This process has almost never failed, certainly it works well enough to hand 95% of what I "do" to a machine that can then "do" the same thing except at 1000x.
=^..^= all your rodent are belong to us
The hard part isn't the coding, it's figuring out the boundaries of the problem. I have no doubt it can write it's slack alternative in 50 lines of code or a "social network" in 150 lines, but they'll be trivial examples that don't take into account the realities of complex software development.
This whole concept of ultra-high level was tried before with Visual Basic and Java and Scratch and other languages that promised to make programming accessible to everyone, the problem is that the *problems* are complex to understand. Most programmers that are capable of understanding edge cases and corner cases don't struggle so much with the language. On the other hand, people who start off programming and are struggling with the languages are likely going to have trouble with articulating the non-trivial edge cases. So for them, making the programming language faster and easier doesn't really solve the underlying problem.
It's nice to think about functions and objects as being "plug-and-play", of having square and circle and triangle connectors that only let you attach them certain ways. The reality of software development is that the glue interactions between modules are much more complex than that. You can do some tutorials and say 'oh that makes sense' or even build some trivial apps for yourself and gain false confidence, but as soon as you try to build a real world application you'll realize that hiding complexity doesn't remove the need for understanding complexity.
I want to cure world hunger, I have a hand-wavy idea about giving food to hungry people.
Maybe we should save news articles until people actually accomplish something instead of wanting to?
Democratize is a buzzword meaning let unskilled people in. I imagine with democratized AI web apps you'll get a fairly similar results with the same look & feel - but the cream developers will continue to rise to the top, and still develop fresh and interesting user interfaces instead of the garbage poor-people webapps made by AI. That'll be the new baseline, and it'll open up a new frontier to be democratized / homogenised.
The fired journalists being told to #learntocode need a chance at software development too, by having AI do all the work.
What's the point of software development anyhow? presumably to solve problems not to develop software.
Douglas Adams proposed the interface of the future would be a desk you work at trying to solve a problem. The computer would observe what you were doing, then write an algorithm to do it for you. At the time he meant this as a joke. But this is infact exactly the sort of problem that so-called Artifical Intelligence is getting good at. It's getting good at recognixing a start on something then completing it. For example deepFakes fills in a face into a removed face. Adobe's eraser removes defects and fills them back in. And combinatorial material ascience is having success in taking in some basic physics and examples of compounds that exhibit desired properties and then suggesting new molecules that might have similar properties.
AI is really crappy at figuring out what to do. It's really good at observing what you think is important then extrapolating that. Thus Douglas Adams desk interface is no long a joke concept.
How hard would it be to have a computer write a sorting algorithm just by watching someone sort numbers? It's plausible.
Some drink at the fountain of knowledge. Others just gargle.
People have been touting variations of this concept for decades, and it NEVER pans out.
You know what happens when you let "AI" do the "heavy lifting" of writing code?
You wind up with crap like Dreamweaver's garbage HTML code...
This MAY work for trivial, formulaic crap like CRUD coding, but for the 50% (minimum) of programming that requires coming up with something novel to address a unique situation? It's going to produce nothing but non-performant, fragile, unmaintainable garbage.
prepare for 40k lines of code for something that takes some input text, truncates it into 10 character blocks of ASCII and outputs as a text file.
Processing/RAM/HDD advances are used to cover up for crap/inefficient code: this has been growing for decades, so the natural end result is 40 thousand lines of code to do fuck all. Ironically the AI that hipster BS'ers bandwagon on will do their hipster "coder" (aka cluserfuck of inefficient junk) brethren out of a job into the bargain!
Thankfully that methodology can only get you basic scripted shite - anything that needs the actual full power of current processors will still need intelligent humans to ensure that said full power is harnessed.
They make it sound like that's the easy part.
Also noted the common failing that you get what you ask for not what you want, which AI has made even more relevant than it was before
Some drink at the fountain of knowledge. Others just gargle.
Dry.io, a developer playground that helps you write web apps using just a few lines of code
Just what the world needs. More crappy web apps....
So, other than for ideological reasons, why would you event want to democratize coding? Why not have people who are good coders code? I mean, is it important that I, as a coder, can take out an appendix? Should medicine be democratized?
And in terms of ideology, I'm all in favor of fairness. I think all should be given equal opportunities, and when social conditions are such that that can't happen, fixes should be sought. But it's kind of brain-dead to think that its a good idea to enforce some kind of equality of outcome, where not only do all have the opportunity to code, anyone who wants the job can have it, regardless of skill.
The dream of businesses since the dawn of programming ... yes, as long as you can fully specify all the branches of logic, the machine can write the code for you!
Of course, it would help if we could devise some sort of symbolic written language to represent the logic, since human languages tend to be imprecise ...
Then the computer could tell you if you got the syntax wrong with the symbolic language or something.
It should only take Marketing a few years to get up to speed with using this. (In the meantime they will stop actually marketing, is that a problem?)
Hurrah, no more pricey programmers!
You mean they want people to train their AI system for them? And I'm sure they'll compensate you for that, right?
Whatever, it's a startup, with a .io domain ... which means they're garbage and I don't care what they're doing because it has nothing to do with me.
Me, I never give any startup anything for free, I'm not propping up the business model of some asshole of a tech bro.
are 'learning to code'? Are they out of a job already?
99.9% of HUMAN developers suck at writing software. Somehow though machines will be that much better?
Crap.
I, for one, welcome our new AI coding overlords.
I'm a big proponent of AI myself, but I think it simply shift what work gets done - that is to say, there will still be a lot of programmer jobs, but they will be more about directly higher level concepts than lower level programming we are used to...
But even with that, two years sound really optimistic for taking over programming, because there is such. large mash-mash of things it could possibly help with.
I think we'll have real honest-to-goodness self driving cars running around the world long before we have a significant number of programming jobs taken over by AI helpers. I'm thinking maybe 8-10 years before we see significant strides in this space.
"There is more worth loving than we have strength to love." - Brian Jay Stanley
It needs people with insight, skill and experience. Doing it by committee routinely produces the worst possible outcomes.
Most ACs are not even worth the keystrokes to insult them. Be generically insulted by this and ignored otherwise.
Make 'merka Great Again.
This is probably more of an attempt to let the software pick the underlying software for you with pre-packaged open source software ready to configure. Rather than going to WordPress for a blog, Shopify for a store, or whatever, you just go to one provider and their software picks the package that suits your needs.
Developers will always have a job because the skilled ones already have the toolbox for setting up the baseline software or working with the existing software. They are paid good money to fill the gap between what the software does and what the company needs.
The work that the AI is doing is an afternoon for a skilled developer. Any skilled developer already has written code that writes code for them based on patterns they use that need repeating. The most common being an ORM. The skilled developer designs the database and then the code generates the code so that the developer does not need to hand code a bunch of repeating patterns.
It's funny that they call it DRY.io because my own PHP framework is QuickDRY. It has written millions of lines of code for me so I can focus on business logic and not grunt work.
https://github.com/BenKucenski...
Work Safe Porn
I heard the same thing with a little programming language came to market called Visual Basic ... and Access would end the need for DBA's ... and in a way they did ... both lowered the bar for entry into creating applications. But... The end result was a torrent of sub-standard programs (and programmers) and a lot more work and opportunities for experienced/skilled programmers. To be sure things like AI and Dry.io are going to solve lots of problems but they will created a whole new set of problems (read opportunities).
I don't understand why data-dictionaries are not used more. Automatic code generation speeds up initial development, but is still difficult to maintain, as you have to sift through auto-generated verbosity to change things. Why can't I describe the field "employee number" in one place, with its database name, title, min size, max size, validation type etc.?
You are advocating auto-replication instead of true DRY. To me, that's auto-bloat.
One problem with data dictionaries is how to override the defaults for specific problems: the "delta problem". With auto-bloat, you just change the local copy for variations.
I've found ways to deal with the delta problem of data dictionaries using PHP arrays and some "helper" functions, but this approach isn't mainstream. The rest of the shops are addicted to auto-bloat instead. I don't get it. Habit?
Table-ized A.I.
FOAM: https://www.youtube.com/watch?... Naked Objects: https://en.wikipedia.org/wiki/...
I'm probably missing something, but did they just re-invent fourth-generation programming languages? Or maybe they're just recycling the hype, because now "AI" which no-one can even define in the first place.
I've been hearing that promise for 35-plus years. Programming is more than specifying an algorithm, than melding an algorithm with data, than maximizing the serial throughput of a Turing machine: It's dealing with the fiddly edge cases that appear in mathematics, business and life. It's why software needs maintenance, because the rest of the universe changes, not only because mistakes made by software developers must be fixed.
The writer has no idea about programming. They randomly decided to explain pair programming in the middle of the article.
The company has 4 people and no external funding. They're making the same thing as Wix.com, except with the idea that "maybe **AI** will generate the whole app for you." Meanwhile, Microsoft Research has been trying similar stuff for years, and they can't even really make something that learns regular expressions from examples. The company will just end up making a whole bunch of website templates and charging a monthly fee to users.
If "heavy lifting" means handling almost all error conditions with actually useful error messages and actually human-readable code which is kept separate from the business logic, then hell yes.
Well, one can dream :)
the bit at the bottom saying 'what are you planning on writing' is apparantly a 'name' field so can only take alphanum, underscore, dash, fullstop as input
so rejected me putting an amphersand in.
thats a bad start if it can't understand that its some kind of free text field or the expression of its definiton somehow isn't intelligent enough to understand its a free text field.
i'm out. for the moment.
Most any project worth having is a system at heart.
Sometimes a quick chunk of code will get you a feature. But wherever true systems roam, actually software developers will be found to formulate the interactions and clean up the mess.
For years now, the hardest part of creating any new software was...getting the requirements for what it should do.
At my company, the development bottleneck is not the programmers, it's the business analysts, trying to figure out what the company wants to build. Once they decide, our team is able to build it quite quickly with existing tools.
Houses today can be build pre-fabricated in a factory, and thrown together in a few weeks. You get what you pay for. If you want a quality house, you still have to build it from the ground up, on site, using more traditional methods.
Software isn't much different. If you use tools that use "AI" (i.e., pre-fabricated parts), you'll get what pre-fab can do. Crap.
All the tools that came before it were good at building some very specific kind of software. The most common use case is data entry screens. Yeah, that's easy. It's also easy with traditional tools. If you want to do something sophisticated, "AI" isn't going to get you any farther than the old 4GL or RAD tools did.