With the current pace of gpt and dall-e, it’s looking more likely that a lot of development roles may go obsolete in the nearest future (3+ yrs).

I see the possibility of building full fledged websites and fixing bugs based on voice commands. The picture of this possibility is quite vivid in my head because it’s totally feasible technical-wise.

The only delay that may occur in this dynamics is the slow pace of its implementation by existing developer tools. Of which I think the reason is directly related to the cost of management of resource, quite the limiting factor here.

But imagine if a big tech like google creates a platform to build websites based on voice/text commands using advanced gpt inline with its access to existing corpus of data; that to me is “game-over” for web devs.

Nah, the pace lsnt fast. It flat out stops with gpt models. Word predictors can only get us so far.

There's just no way you, or any other cognitive being, will simplify or rationalise the vast array of unnecessarily complex nonsense used in modern web dev. Even if you could simplify the development of a new site, the vast amount of legacy is what everyone deals with on a daily basis. New site, easily? Wix.

If that's how you feel, your job must be pretty damn useless to begin with. A trained monkey can do things like turn instruction into code, or compose a bunch of UI elements to get the desired appearance. But even slightly more advanced tasks like optimising search using data-specific trees, filling significant gaps and inferring missing instructions, or maintaining consistent UI "look and feel" across different screen sizes, and AI models fail because they can't infer the necessary context from anything except the small set of instructions provided.

For example, even if you provide a GPT model with your DB structure and ask it to write a moderately complex query, it could never infer the idea that "post_code" column always contains a string consisting of optional two-letter prefix followed by a fixed number of digits.

Many dev tasks are "pretty damn useless to begin with" like @hitko stated. It might be "totally feasible technical-wise" to automate junior tasks, but how would anyone develop or discover a "do what I mean" software when people fail to know what they want? Requirement engineering, communication, creativity, and thinking beyond the given best-case examples are important parts of successful web development, and so is some kind of intuition when customers get unhappy and claim to find new bugs in an unchanged setup.

Development is no game of chess. For so many years, people told me that "web development isn't programming", meant in a derogatory way. But it seems they were right. Programming is only a small part of web development.

We should be happy to lose customers who are fine with Wix and the like to the new voice-controlled website builder so that we can focus on more challenging and inspiring projects instead.

@fraktalisman I believe the ability of an AI understanding a user needs is quite related to how complex it could interpret natural language. With what I can see, gpt is getting smarter by the day on that.

If an AI can study the construct of how most modern websites are built. I wonder why you think it cant occupy a senior dev role. It would likely out do a human dev if trained on such data.

Almost all websites follow a specific construct based on the category it falls in.

Take for example:
— e-Commerce website
— Admin panels
— Blogs
— Forums

Why do you think an AI built on specific models can’t create these types of platforms on command?

Imagine the command:

Build me a website like devrant.com using react for front-end and node for backend with mongo db.

You genuinely think this is not possible?

Yes, I genuinely think this is not possible. Or, more accurately, know that this is not possible.

@spongegeoff Nah… you must be bluffing.

The advent of these would form a similar era when Wordpress, Wix and the likes came into the picture. Building simple websites like blogs was a programmers thing until it was not.

Now we are heading towards the era where building complex websites will become a once-a-programmers thing.

AI cannot act on the basis of detail it is not aware of and cannot obtain. Ergo, therefore...

@spongegeoff It can be trained.

@spongegeoff we already have AIs generating unique photos of human based on training. If you teach it well enough about websites it can do it.

I think we will end up with a bunch of cookie cutter looking sites all with the same problems and look and feel.

@Demolishun Yes true. I believe we had similar problems with WP that was why we saw a boom in theme creators to create some level of uniqueness. The workload drastically shifted from developers to designers.

but aren't designers and copywriters the ones made obsolete by artificial "intelligence" now?

Painters getting obsolete by photography, photography getting obsolete by dall-e and midjourney? Really? Recently, I went to a gallery and met actual painters, why do they still exist?

@fraktalisman It depends on the type of work. AI can't replace real creative copywriters or painters because AI doesn't have the necessary artistic inspiration or understanding of things on a deeply subjective level. However, it can replace people whose sole job is to take a bunch of facts or guidelines and put them in a pleasing form for people to read or look at. And those are the people that will be made obsolete by the AI, bor better or worse.