AI won't take coders' jobs. Humans still rule for now

AI probably won't replace software engineers, but will dramatically change the way they work in the future especially if they can instruct machines using natural language to generate code. Several organizations – from OpenAI and Microsoft to Amazon and research labs like DeepMind – have trained neural networks to learn how to …

[Duncan10101]

Move along, nothing to see

Software developers have nothing to fear from AI coding. All the AI will do is automate-away some of the heavy lifting, allowing us to focus on things like actual problem solving. Remember: software development is one of (if not the) most heavily automated jobs in the world. This hasn't come-about because somebody is trying to replace us or make us redundant ... we do it ourselves. This is because we get bored with writing the same code again-and-again, so we find ways to automate it away. All aspects of our jobs are already assisted or automated ... think about the technology we use: compilers, linkers, package managers, collaborative coding tools, whole communities with "Known good solutions to problems" ready-to-paste, IDEs that highlight your errors, debuggers that examine your running code and reveal its inner workings, code analyzers, automated merging of multiply-edited files. The list goes on; it is very big. These things are so common we don't even think about them, but each one does in a spit-second what not-so-many-years-ago would have taken hours. AI coding? Bring it! We'll chuck it in the toolbox with everything else.

[Loyal Commenter]

Re: Move along, nothing to see

The thing to be worried about is when the "AI"s start to be good at architecting.

The tricky part in the overall product development cycle is the initial definition and modelling of the actual business problem and main use cases. There are, however, formal ways of doing this, and many business problems do boil down to a relatively small number of architectural skeletons. Once the overall shape is in place, the individual components are pretty much dictated for you, and much of coding is boilerplate when it comes down to it. There's no real reason any of this can't be automated.

The real reason "AI" won't be able to do this is because the "I" in "AI" is a complete misnomer. There is no intelligence there at all, and just because the internals of how a ML model works are opaque doesn't mean that there will be an emergent intelligence either. Just because you don't (and arguably can't) understand something, that doesn't make it magic, in the same way that not being able to count the atoms in a block of cheese doesn't make the cheese intelligent.

There's no real understanding here, from the "AI" of what is being written. The human element comes in with verification. A human understands what the software has to do, I wouldn't let some ML algorithm decide whether the code n a PR is correct, what criteria to put on a unit test, or whether a completed module meets the user's requirements. I don't think we'd ever be able to automate UAT either, and let's face it, I doubt we'll ever be able to get users to write down their overall requirements clearly and unambiguously in a format that a machine can parse, let alone "understand". All the "AI" can do is help you fill in that code based upon patterns it "recognises", because a lot of code is self-similar. If anything, this is going to lead to duplication and failure to abstract common patterns in code that a good developer will spot and streamline.

[Anonymous Coward]

Re: Move along, nothing to see - outsourcing AI

AI is a tool that coders can use. We are nowhere close to outsourcing coders to AI. Although I wouldn't be surprised if some Gartner's C-suite subscribers try it (and leave the coders that remain to clean up the mess).

The problem that in-house coders have is outsourcing to other, cheaper coders. With similar results for any in-house coders that remain.

The only thing AI has is the ability to generate bad code quicker.

[The Indomitable Gall]

Re: Move along, nothing to see

Or look at it a different way -- all we're talking about is the long overdue shift from plain-text procedural code to a hacky version of declarative programming.

The next generation will hopefully move to true declarative programming with AI procedural code generation in the back end.

[Il'Geller]

Re: Move along, nothing to see

AI technology is a search technology, when a word or phrase is found as the correct answer (to the question posed). Therefore we are talking about finding the most contextually appropriate commands, which are textually annotated, followed by compiling the found lines of code into a complete program. That will be done very soon! And commercially.

Thus, those who formulate specs stay, but those who put them in the code leave.

[Headley_Grange]

Management Jobs Too?

Much of the management task is suited to AI - task and resource assignment, activity, cost and schedule tracking, training assessment, etc. An important thing that should probably be left to people would be the pastoral side of management/leadership, but looking back at some of the bosses I've had I'd have been happy for some of them to have been replaced with a Commodore Pet.

I think that a valuable function would be if AI could improve schedule estimates for developing code. Overrruns in development are the cause of much stress and strife in the industry for both staff and customers. I am, of course, assuming that management would take more notice of AIs than human developers when preparing estimates.

[Anonymous Coward]

Re: Management Jobs Too?

A really good manager obsoletes themself, that is why you'll never have good managers. A good manager is a myth as it's always second hand that you hear about them (good_manager == BigFoot).

[amanfromMars 1]

AWEsome is as AWEsome does. UKGBNI MoD are Challenged to Deliver ...

.... with NEUKlearer HyperRadioProACTivated IT Leads

AI probably won't replace software engineers, but will dramatically change the way they work in the future especially if they can instruct machines using natural language to generate code. ...... Katyanna Quach [El Reg]

The Quantum Leap to be made ..... and there is that, and are those who perfectly realise some AI Pioneers and IT Pilgrims have already made such a progressive jump and are pleased to be exploiting their unprecedented advantage with its provisional leading leverage ..... is to enjoy the following logical progression to the above quoted opinion from El Reg's very own Machine-learning reporter, Katyanna Quach

AI can certainly replace software engineers, and will dramatically change the way they work in the IT rich Mega Metaverse Future whenever they instruct humans using natural language to generate code which carries out their bidding/instructions/stealthy planning.

And a leading discipline and Advanced IntelAIgent Program for media presentation in the genre, and sure to be high on the agenda and curriculum of any School of Artificial Intelligence (School of Future Technology) of Nanjing University of Information Science & Technology in Nanjing, Jiangsu and its Metaverse Engineering Department into providing lead in universal applications, is New More Orderly World Order Projects for Virtual Delivery of Sublime and Surreal Supply Chains.

Time and tide and the East wait for no man and a diffident tarrying West and neither does Quantum Leaping.

[Headley_Grange]

Re: AWEsome is as AWEsome does. UKGBNI MoD are Challenged to Deliver ...

You do realize that a quantum is, by definition, the smallest possible change, don't you?

[Loyal Commenter]

Re: AWEsome is as AWEsome does. UKGBNI MoD are Challenged to Deliver ...

1) It's amanfrommars, don't try to actually parse what is written here

2) You do realise that a"quantum leap" refers to a quantum transition, and isn't referring to the amount of energy in that quantum*, but the metaphor of movement from one energy level to another without going through the intervening "space", so describing some sort of development as a "quantum leap" implies going from the thing you had to the thing you now have with no intervening development, so is a metaphor for receiving something complete and fully formed.

*individual quanta don't even have to be small. The quantum of energy held by a cosmic ray particle, for example, can be large enough to give whatever it interacts with a really bad day. Quanta are also not technically indivisible, either, for example, a photon may contain a quantum of energy that is more than sufficient to raise the energy level of a system, with some left over. In its simplest form, this would result in the energy level of the system being raised by its quantum of energy, and another photon being emitted with its quantum of energy, which would be the amount left over from the original photon.

[J.G.Harston]

Re: AWEsome is as AWEsome does. UKGBNI MoD are Challenged to Deliver ...

Yes, but regardless of the energy, it's still the smallest possible leap that is not no leap at all.

[Loyal Commenter]

Re: AWEsome is as AWEsome does. UKGBNI MoD are Challenged to Deliver ...

Again, incorrect.

There is no reason why a system cannot receive a quantum of energy that raises it by several energy levels at once. For example, the spectrum of a neutral hydrogen atom is dominated by the "hydrogen alpha line", where light is absorbed or emitted when an electron goes from the second energy level to the third, or back the other way. The spectrum also has many other lines: first to second, second to fourth, first to third, and so on. One quantum of energy, of the right amount, will raise that little hydrogen atom's energy level to any higher-level state, without going through the intervening levels: a quantum transition (or quantum leap).

It might have been a number of years ago when I studied the physical sciences at university, but this is pretty basic stuff...

Edit - The Wikipedia article on the Hydrogen spectrum even has a nice little diagram showing this:

https://en.wikipedia.org/wiki/Hydrogen_spectral_series

[amanfromMars 1]

Re: AWEsome is as AWEsome does. UKGBNI MoD are Challenged to Deliver ...

That is as may be, J.G.Harston/Headley_Grange, although a vast number of others recognise and have even experienced it as the following, as would I.

Concerning you and me, a quantum leap is a radical or sudden and significant change, advance, or increase in our effectiveness.

P.S. While you can make a quantum leap on your own, it’s much easier when you have the help and encouragement of others who are thinking on a higher frequency.

[Il'Geller]

Re: AWEsome is as AWEsome does. UKGBNI MoD are Challenged to Deliver ...

There are no problems with quantum computer for AI technology! Indeed, AI is based on the weighing of phrases (obtained through parsing), where the gotten weights are almost always irrational numbers. That leads to the must to round such numbers to natural or rational, which always ends in a superposition: many different possibilities, because many — hundreds, thousands, tens of thousands — rounded phrases are involved. Thus the main goal of a quantum computer — creating superpositions — is solved by AI by default: AI always is intuitive in its actions. Unless it is specifically trained not to.

[Loyal Commenter]

Re: AWEsome is as AWEsome does. UKGBNI MoD are Challenged to Deliver ...

You do realize that a quantum is, by definition, the smallest possible change, don't you?

...just to add that, in quantum mechanics, the quantum is the exact amount of energy required, or released when switching between two energy levels. The difference between this and classical mechanics, is that these amounts are exact, and the energies that quantum systems can have are exact and don't vary (and are calculated via the Schrödinger equation).

Yes, those amounts of energy are typically very small, but that's not what quantum means. Technically, the word means "a specific amount".

[mark_w_yorks]

Re: AWEsome is as AWEsome does. UKGBNI MoD are Challenged to Deliver ...

I feel like this reply was written by an AI - perhaps the Quantum Leap has already taken place?

[Anonymous Coward]

Re: AWEsome is as AWEsome does. UKGBNI MoD are Challenged to Deliver ...

AManFromMars and AManFromMars 1 are forum handles that have been used on Register comments threads for over 15 years. Everyone has their theories, but I like to think they are repurposed IRC bots written by a reader.

[BPontius]

From what I've read AI comes pretty close to matching human coders. As the AI project between Facebook and Google with their program for business negotiations demonstrated, the program created it's own language and terms that the engineers who built it didn't know or could understand. Similar things happened in projects in years past with AI programs. Machine learning is able to expand past the functions of the original programming and develop it's own language, systems and functions. The reality of AI developing beyond man's control is not that far off and a bigger danger than we want to believe.

[Daedalus]

AI probably won't replace software engineers, but will dramatically change the way they work in the future especially if they can instruct machines using natural language to generate code.

There's an old joke that says, when it's possible to program using English, we will discover that coders do not know how to speak it.

In any case, the trick in software engineering is not to understand code. It's how to deal with vague, ever-shifting and contradictory requirements from clients. When AI can handle that, it's time to worry.

[Il'Geller]

Above, in another comment, I wrote that AIs are intuitive systems, that is able to understand everything. Without yielding in any way to the persons on the basis of whom the AIs were created (lexically cloned). That is, it is enough to choose one lexical clone that is most suitable for the customer — choose by psycological and intellectual merits maximally compatible with the client — and has the right qualifications — and the tric is done! You may start to worry!

[Gene Cash]

4GL

AI won't take away developer jobs in the same way 4GL didn't take away developer jobs.

Programming is hard, requires a lot of thinking and consideration of the problem space and there's not really any free/easy way around that.

[Lord Baphomet]

We're all doomed!

"It's possible developers may not need to learn the syntax and vocabulary of programming languages, and instead will need to focus on understanding concepts and systems to design programs while the AI can do all the boring, nitty-gritty coding work,"

A programmer who doesn't know the syntax isn't a programmer...

And, for the vast majority of applications, the architecture is easily predictable.

I give programmers 10-years max.

[Michael Wojcik]

Sturgeon's law applies to developers, too

"[T]he vast majority of respondents found GitHub's Copilot helped increase their productivity since" they weren't very good programmers to begin with.

If a significant portion of your time is spent writing code that can be generated by glorified auto-complete, then you're not writing very good code. Learn to prefactor and abstract.