17 March 2023
Salsa Digital

Over the last few months, most technophiles have watched in shock and awe at the rise of the Large Language ModelsExternal Link , like OpenAI’s ChatGPTExternal Link . Anyone who’s played with it can see that it’s a qualitatively different beast from anything we’ve seen before, able to generate complex and human-like responses to any topic under the sun, from silly poems, to code, to summaries of scientific papers.

There are numerous problems with much of what it produces, but the rate of progress has been startlingly rapid. In parallel with this, there’ve been huge advances in generating imagery and audio using AI. Not only is this going to automate a lot of UX, but enable more personalisation (and potentially more impersonation!). People are scrambling to make use of these technologies, and trying to figure out how to keep pace with these rapid developments.

As Yogi Berra once quipped “The future ain’t what it used to be” and, right now, we don’t really know what direction AI will take in the future. The predictions of an AI winterExternal Link less than a year ago, appear to have been completely off the mark. Bold prognostications about what AI will mean for the future of web development will likely be totally wrong. That being said, it’s great fun to try, and if the reader will indulge us, we can give you a glimpse of what the online experience might look like in 5 years from now, and what this implies for web development.

What’s wrong with the web now, and how might AI make it better?

At the moment, you use the web for a lot of different things such as:

  1. Ecommerce
  2. Finding things out (how to get from A to B, or what is the best Boba shop nearby)
  3. Entertainment (both naughty and nice)
  4. Social media (e.g. Facebook, Instagram, Twitter)
  5. Interfacing with the government (e.g. paying taxes)

These five common categories are all closely intertwined with overlapping drawbacks. Most of us will whinge at one time or another about the experience. Here we can contrast how things are now, with what our future AI servant (hopefully?) will be able to provide for us. Let's call it “ChattyLackey". It might be Bard IV or ChatGPT 10. These are scenarios that we boldly predict could be reality in 5 years.


Example of a problem:

You go online to check if a payment cleared with your bank a couple of months ago. There’s two-factor authentication, but you’ve turned off LastPass since it’s had too many hacks, your phone’s ringer is off because your partner is taking a nap, so it takes a couple of minutes to get into your account before you can start doing anything there at all.

Then there’s a list of popups that tell you about the latest offers (“remind me laters”) from your bank or new updates to the terms and conditions. Then the website has a dozen tabs, and you just don’t seem to be able to see which drop down to choose to find out if your cheque to the plumber has cleared. He says it didn’t. Seems to be an honest bloke, but who knows? How do you find the damned cheque? This is taking way too much time. It’s a very frustrating experience.

You finally see that it has cleared but somehow he hasn’t been paid. It’s probably in some limbo state. OMG. Now you have to try to talk to someone. You know you’ll be on hold for 45 minutes if you call, so you try their messaging system. Click Click Click: drop downs, more clicks. You need to do more authentication. Crikey! Will this ever end?


Hey ChattyLackey, the plumber’s cheque didn’t seem to go through. Can you investigate?
No worries Pat. Apparently the cheque is in a limbo state and I’ve gone through the message system and also called them (on hold at the moment). I’ll follow up and make sure it clears. If there's something that requires your further intervention, I’ll keep you apprised. If your bank has messed up, I’ll report this to the relevant agencies, and online reviews, as you’ve asked me to do in the past. All good?

So behind the scenes, ChattyLackey has done all of the donkey work you needed to do, navigating, filling out forms, including composing texts, dealing with two-factor authentication (a bit scary but very convenient). At this point, none of this is rocket science except presently it’s not totally reliable.

Note we still mentioned online reviews. It’s not clear how much weight they’ll carry. What will be much more powerful in the future is ChattyLackey’s assessment that the bank screwed up. It’ll use that extra data to change its recommendations in the future for the bank, not just for you, but everyone. The same will be true any time it interacts with any business.

Finding things out

Example of a problem:

Suppose you want to go to get Boba with your friends at someplace close by (yes, I’m talking about those sweet gelatinous balls, nothing more risque). You’re in a big city and most shops have somewhat of a web presence. But as we’ve all seen, it’s not always easy to get a hold of the menu beforehand; you might have to go through some third party site, or look through Google images.

You will, somewhat obsessively, go through reviews to figure out which shops are considered acceptable. But your experience with a lot of reviews is that they’re fake. And to add to this, your friends have different requirements. One insists it be Kosher, the other Halal, a third organic, a fourth vegan. What you thought would be a nice relaxed social event has turned into a complicated extra credit problem on a graduate computer science final exam.

Now the geeky part of you would love to create a spreadsheet of all these requirements, write some python code to read your csv file, and then apply Bayesian statistical maximum likelihood analysis to optimise the joint utility function of your little outing. But eyes start rolling when you do this (except for Irving who is even geekier than you). And so you all flip a coin, go to some random shop, and all get food poisoning. Lovely.


Hey ChattyLackey, got my mates Irving, Sandy, Bruce, Paul and Chip here. You know them right? Can you recommend a good boba place for us?
Sure Pat. I’d recommend a Taiwanese Boba place, and here are the directions. It's a nice day and close so I’m assuming you’re walking. It’s got all of your eclectic requirements, and here’s a brief video I’ve made for you and their extensive menu, that I think will please everyone. I’ve texted everyone in your group with a personal tour of the place if they’re interested.

All the geeky stuff you wanted to do? ChattyLackey did it for you. It also is great at filtering out fake reviews (if that’s still a thing in 5 years). And it personalised the videos the way that your mates like. Paul likes to see himself in the videos wearing hip hop clothes and rapping. He sees himself in this Boba shop, wearing a golden tracksuit with the employees in the back extolling the virtues of Oolong Lychee with notes of Tie Guan Yin. Paul’s getting thirsty already!


Example of a problem:

You want to choose a movie to watch with your sweetie. You feel like a comedy thriller or a musical. Being somewhat miserly, you want to use a service for which you already subscribe. Let’s see, what are they? Netflix, Hulu, Peacock, Prime, Disney+ (no, Disney? say it isn't so!).

Normally the evening deteriorates as you incessantly search through IMDB and these streaming services for hours, at which point you’re both too tired and just want to go to bed.


Hey ChattyLackey, Lenny and I are trying to figure out what to watch. We think we’d like a comedy thriller or a musical.
Sure Pat, you can watch some new Sweeney Todd Netflix musical that just appeared the other day. It got 7.3 on IMDB but a lot of the poorer reviews were from people who don’t like musicals. I think you’ll like it quite a lot more. But if I had to choose for you, I notice that Prime is now streaming a new season of the Marvelous Mrs. Maisel. Did you know that? Want me to cast it to your living room TV?

Social media

Example of a problem:

Being bogged down with work, you haven’t been able to keep up with all of your social media, like Facebook and Instagram. Your friends are very chatty and some of them go on and on. They get upset when you don’t seem to be aware of big life-changing events like weddings, dead pets, etc.


Hey ChattyLackey, can you go through my social media and tell me about important events for my friends.
Sure Pat, it seems like Irving’s goldfish died. He seems to be in a lot worse shape than Steve and Charlene who just broke up and are having a bit of an exchange on Twitter. I’m assuming you want me to text Steve and Charlene individually and avoid opining on who gets the Xbox controllers. I can respond nicely to Irving’s tragic loss.
Hey ChattyLackey, sounds good. Could you also make up some little silly responses to some of the less touchy posts just so I appear in the loop?
Sure, no problem. You know if you keep doing this, people will suspect. But I guess most of the posts are ChattyLackey generated in any case (not that I’m allowed to tell you who they are from).

Admittedly this isn’t a great solution, but chatbots have already infected social media and it will only get worse. What the ultimate solution to all the societal problems of social media is up in the air, and we won’t wade into that quagmire.

As we said above, aside from reliability, AI can readily do all of the tasks in the scenarios we just went through. Obviously the brainiacs are working really hard in making them reliable, and there’s a good chance, in our opinion, that they’ll have that solved over the next 5 years, and there'll be a lot of other improvements as well. This presents a much improved way of interacting with online systems, such as Google’s search. That’s why Google is freaking out.

Interfacing with the government

Example of a problem:

After coming up short Googling, and having no success trying to understand government sites, Bernie calls the U.S. Internal Revenue Service (IRS) with a question about his taxes. After 90 minutes on hold, he asks "Can I claim the Child Tax Credit and deduct $2000 for my 10-year-old son?", "No problem," says the IRS tax person. Fast forward a year, and Bernie gets a nastygram from the IRS saying that he couldn't deduct the $2K given the particulars of his situation, and owes that to them with interest. The “helpful” IRS tax guru should've probed the details, but didn't. Bernie and his wife were working two jobs trying to make ends meet and this unexpected tax payment tipped them over the edge. They ended up having to foreclose on their house and had no place to go. But, finally, his wonderful mum came to the rescue, letting them stay with her, helpfully reminding them (every five minutes) how fortunate they were and giving Bernie's wife frequent child rearing advice at 6AM.


Hey ChattyLackey, it's getting pretty close to the deadline for filing taxes. Just wondering how it's going?
I already have it all filled out and ready to submit. Just waiting till closer to the deadline so you can earn more interest. By the way, it doesn't look like you can claim a Child Tax Credit this year. I can go over the details of that if you want. But everything else looks quite straightforward. Want to know the damage?

In order to do your taxes effortlessly, ChattyLackey will go through your financial records, log into online portals, look through apps and emails for receipts, and then fill out your tax returns, making full use of the tax code to minimise the amount you have to pay. There won’t be a need for programs such as TurboTax or days trying to understand arcane rules.

The byzantine and voluminous legal system, combined with a lack of human-readable resources, make it a nightmare dealing with government agencies. Problems range from trying to deal with car registration to building codes to paying taxes. Fortunately, machine learning paves a way out of this mess. Not only does AI now have the ability to assimilate huge amounts of data, but there are initiatives such as Rules as Code using OpenFisca, for making the government more efficient and reliable. Although governments tend to be behind industry in making changes, we hope that products such as ChattyLackey, will make this aspect of everyone’s life a lot easier.

How does AI affect how content is presented?

In the above scenarios, we’ve predicted quite a different way that web content will be presented. This aligns with what others such as Dries Buytaert (founder of the popular CMS, Drupal) have recently said, but perhaps more extreme. You don’t want to be looking at content as presented to you by the merchant, the bank or search engine. ChattyLackey makes the whole experience way more pleasant.

But you also don’t want to be looking only at ChatGPT text boxes so we’re pretty confident that web content will become a lot more varied and richer.

How will web content look?

Well here’s one way it could play out... AI will likely evolve to integrate media and UX in a highly personalised way. It’ll take content from websites, throw away the spam, toxic pieces of industrial waste (err we mean all of that over-the-top marketing fluff), and give you nice succinct visuals, audio and text, appropriate to the task at hand, and what makes you happy. The graphics/videos can be generated by the AI itself to suit your mood and tastes, and you can be spared the endless dropdowns and text boxes currently necessary for conducting run-of-the-mill business. If you really want to wade through all of the marketing blather on a website, sure, it can show you it, but who wants to do that when dealing with a bank? How exciting are term deposit rates really?

And don’t be fooled by the above example involving just one site either. In general, it'll be taking data from zillions of websites, databases, videos and images, feeding it all into its trillion parameter brain and doing an awesome job performing the task you want. This is the future of the web. It will screen out marketing drivel, fake reviews and whatever bad things it can find, use its own experience with these businesses, and do a lot of intelligent interpretation to present you with a new and greatly improved experience.

Life will be so much better with all of that gone. We’ll feel like it’s 1990 again. The AI-generated birds will sing, the AI generated soothing music will play. And we’ll get less tummy upsets from rancid boba.

But this will mean that the UX of websites will be, sorry guys (it hurts we know!), a lot less important than it is today. AI will be doing the look-and-feel for the user who will be blissfully unaware of the awful drivel on many websites. Old timers will still be typing in http://GreedyBank.comExternal Link and the anklebiters will be asking “Hey Grandma? What’s a website?” So sure, some people will want to go to websites, but it’ll become less and less important.

But content will obviously still need to be put online (content will still be king!), just in some form that is digestible by the bots. The thing is, that the bots are like … really smart. They won’t need to have metadata or necessarily even html text, to suss out the content we put up. In many cases, they’ll be happiest with raw (read only!) access to databases. This will certainly save companies money as well, if no one needs to spend tons of time on frontends that wrap the raw data in layers of CSS and JavaScript. You’ll still want to have nice images that ChattyLackey will hopefully include in what it serves up to you.

Trying to game what gets pulled in by ChattyLackey will probably be unwise. You’re dealing with a highly intelligent and suspicious brainiac, so it’s best to just be upfront and direct. Nothing about how you provide the best Boba made from the finest Brazilian cassava root, when in fact it’s powdered cardboard mixed with insecticide. The bots will look at reviews, complaints, etc, and most importantly, its own previous experiences with all other users, and so you might as well be upfront about what your product actually is. Pretty pictures of the Amazon jungle with parrots flying overhead will be of no interest to ChattyLackey, or getting the CSS just so. 99% of the time, the user will not see any of that.

Web content will become more plain and vanilla with simple ways of accessing underlying data (obviously in a secure way). It will not be worth the time and cost for companies to obsess over details of how a site looks, if only a small fraction of potential customers ever look at it. They’ll put their efforts elsewhere.

Companies that now try to use ChatGPT to generate huge streams of content are likely in for a rude awakening. ChatGPT will know that it generated the rubbish that you’re serving up. If it wasn’t the source of the generated blather, there are detection methods currently under developmentExternal Link that are already quite good at detecting suspicious content, and will surely improve over time.

How will companies market in the future?

Before the internet, businesses sent out mass market mailings, and advertised on the radio and TV. Currently they try to get ranked high up on search engines and hope that people click to see their beautiful and expensive sites, with tons of images, excessive marketing and videos. That will mostly be gone because ChattyLackey will deliver its own content that’s far more directed to the task at hand and this is what the user would prefer to see. The companies can still provide images, videos and blurbs, but ChattyLackey will excise any obnoxious noise. This means that the videos, images and text might as well be direct and honest as we discussed above.

With the traditional avenue for marketing more or less obliterated, there will undoubtedly be a lot of gnashing of teeth and attempts to get a leg up on competition using any means possible. What the companies can try to do is pay (i.e. bribe) the ChattyLackey creators to give a more rosy assessment of their products. But as long as Big AI Bots don’t become a monopoly, customers will gravitate to the more honest bots. This also might become country-dependent with consumer protection enforced better say in the EU than in the US. It will be a bloody battle waged for years, until humans are completely replaced by entities not interested in drinking bobas or watching musical thrillers. :) Hopefully that won’t happen for some time to come.

Companies could continue to pay people to post fake reviews of their products, but that will also go the way of the dinosaur. With 100s of millions using ChattyLackey, it will be able to gather far more reliable assessments of a product than we do now by reading Amazon reviews. You could try to pay people full time to pretend to be real, but the numbers necessary to sway the chatbot's overall assessment would make this prohibitively expensive.

What does this mean for web developers?

Backend developers should still be in decent shape. Content still needs to be delivered to the AI engines through simple web pages or content APIs. Headless DrupalExternal Link , for example, would do a great job serving up this content, and people working in that area will still have an important place in the online ecosystem.

UI, UX and frontend developers will, in general, see quite a lot of changes. We will still need people doing these things, but there won’t be as much need for them to be working on company websites. They’ll still be providing users with a great experience, but through content coming from the likes of ChattyLacky. In other words, frontend people will be working more on backend content generated by ChattyLacky than the backend content from websites directly.

Will OpenAI do this all themselves? Start employing lots of frontend developers to show the content, or continue to concentrate on what they’re doing now? This is hard to say but we would guess the latter. In either case, it’s likely that there will be space for a lot of variation in how content is shown. And as with the web now, there are many interesting ways that ChattyLackey content can be shown. After all, this will be the new preferred mode of interacting with the web, with billions using it, and so the frontend will continue to be a huge deal. There will probably be a whole new frontend ecosystem that will sprout up but it is likely to be quite different than it is now.

So if we continue to take our predictions seriously, then in more concrete terms, they imply the following:

  1. There will be a whole set of jobs that will need to be filled related to being a “Bot Accessibility Expert”. And developers will need to create tools that allow bots to navigate to all of the content we want them to see. This is quite different from our current focus of keeping them out of restricted areas (like banking portals), or keeping non-search-engine bots from spidering public content. We have to figure out protocols for letting some of them in rather than keeping them all out, and the logic will have to be a bit more advanced than what we have now (e.g. robots.txt).

  2. There’ll be more specialised expertise required to allow “retrofitting”, that is figuring out how to let chatbots into older sites that use legacy software. Big organisations will want to minimise changes to their sites and just keep them chugging along. So they’ll have to add new technology to allow the chatbots in while still keeping the vast majority of the site’s infrastructure unchanged.

  3. Regarding Search Engine Optimisation (SEO), it's unlikely that people will be able to game the bots (i.e. "Black Hat SEO") and it's unlikely there would really be an equivalent of traditional ("White Hat") SEO if a bot can read and analyse everything and figure out all the metadata on its own. It is also likely to penalise websites for attempting to fool it. So the best AI SEO will be to simply provide great content that is accessible to the bots.

  4. Headless CMSs such as Drupal that provide content APIs to access content easily for any channel will be well-positioned. Not everyone will be comfortable giving AI bots direct access to their databases even if it’s read-only, so structured content feeds will be one important way the AI engines get their data. Think of it kind of like RSS for the bots, but for content that may not have an equivalent web page. Also, CMSs that are tuned for AI accessibility will have a leg up. These areas give backend web developers plenty to chew on.

  5. Frontends won’t completely die. Secure portals (say for banks) need to maintain functionality. Since humans in the future will rarely want to use these directly, they just need to be navigable by the AI engines. But they will start to transition to a more barebones look-and-feel. This is because websites will become much less important for direct human interaction and, as such, developers won’t be spending as much effort on the look-and-feel anymore. Open source developers won’t want to put a ton of time into maintaining a feature that most people won’t use. So a lot of features will start to be deprecated related to frontend development such as theming. It’s not worth spending the time making things look pretty if hardly anyone is using them.

  6. There will be a whole ecosystem of chatbot frontends. Companies like OpenAI will most likely continue to specialise in what they’re doing now, which is to provide the brains behind the chatbots and a simple API to get queries and responses from it. But this is not what most users want to see. You’ll still want to serve up the text and audiovisuals nicely. There will be a whole range of look-and-feels that will be developed to accommodate a range of tastes, and also to support accessibility. People will want to “skin” their chatbot, just like they do for a whole variety of other apps.


We’ve given you our predictions for what web development might look like in 5 years time. As we said at the beginning, these should be taken with a large pinch of salt. AI looks incredibly promising now, but it could still sputter as it might be really hard to make them completely reliable. That’s what happened with self-driving cars.

Still, we believe that AI is likely to cause major disruption particularly for frontend developers, but there’s no reason to think that this will be a Terminator scenario for them. Developers will continue to adapt and change their skillsets to keep up with the rapid pace of technological development. Thanks to the rise of the Large Language Models and the vastly improved user experience, employment for frontend web developers will shift and there will be quite a lot of technical changes for them, as well as new opportunities for specialisation. And the use of AI assistance through tools like CopilotExternal Link will continue to make web development way more efficient and fun! On a positive note, AI will give frontend developers extremely powerful new tools that will enable new kinds of user interactions and experiences. So this is something that we expect frontend developers to pivot towards.

And, no, this isn’t ChatGPT giving you the happy talk, we are genuine humans (so They would have us believe!)