Three predictions: probably wrong; possibly useful

Why make predictions about technology at all? I am writing this from a train which was predicted to be at its destination half an hour ago, but is still stuck between stations. If we can’t make accurate predictions about a well-known system with years worth of data, how could we possibly make predictions about the ever-changing field of technology?

Yet the process of attempting to make predictions tells us something, even if individual predictions are wrong. This particular train might not arrive as predicted, but I expect that there is a model somewhere which predicts the overall number of trains that will be delayed - and this delay may be consistent with that prediction. Somebody had to develop that model, and that process will have found something interesting about the factors that affect the reliability of trains.

Similarly, while it’s hard to make predictions about the future of technology, the process of attempting to figure out what’s coming next can be useful. It forces us to use our imaginations, to think about the consequences of choices we are making today, and to recognise the limitations of our own knowledge.

In that spirit, then, as promised last week, I’m going to attempt three predictions about the next five years. I confidently expect them to be wrong. If you’re reading them in five years’ time, feel free to judge them on their accuracy - but please also consider whether any of the thinking was useful.

The AI boom will feel like the 2000s dot com era

I’m not going to attempt to make any predictions about whether the current generation of AI models, particularly LLMs, will continue to get more powerful. I’m also not going to make any guesses about the fortunes of the companies building and selling the technology: plenty of other people will do that for you.

But I will make one guess: that the adoption of AI over the next five years will feel similar to the growth of the Internet in the dot com era. There are two reasons for this.

Firstly, the dot com era was a powerful illustration of the idea that we tend to overestimate the impact of new technology in the short term, and underestimate it in the long term. If you were around in the early 2000s, you may remember that we believed that every business would be online in a matter of months, at a time when most people were still too nervous to put their credit card details into a website. It took time for organisations to become competent in presenting their services digitally, and it took time for people to gain familiarity and extend trust to those services. The use of AI may require less time, but it will require time.

Secondly, the Internet of the dot com era was very much v1.0 (or even 0.1) of the Internet we experience today. Remember hand built home pages, with flashing text, garish graphics and background music? Remember when you had to buy your own physical infrastructure to serve an unknown user base, and when capacity-on-demand meant your boss demanding that you buy more capacity? The fundamental protocols and standard of the Internet stayed stable for many years, but the technologies we needed to make use of them took time to develop: I’m expecting the same with AI.

Major failures will accelerate cloud adoption

Perhaps the most certain prediction is that some organisation, somewhere, will suffer a major security breach or technology failure, resulting in harm to their end users, damage to their reputation, and a response from whoever regulates or oversees their industry. I doubt we’ll need five years for this to come true: I’d be surprised if it didn’t happen at least once in the next twelve months.

The slightly more speculative prediction concerns the consequences of such a failure. The most common response to such failures in the past, whether at an industry level or an organisational level, has been to tighten the screws on existing control mechanisms: to find the defences that failed and the processes that went wrong, and to make sure that they never fail and go wrong again.

Yet I see cases where people are starting to realise that it doesn’t matter how much they fix their processes and enhance their defences: they will never be as good as if they moved to a fully software defined platform - to apply the power of computing to the business of computing.

As I described last week, early cases for moving to the cloud were cost cases - but experience eventually showed that the agility case was stronger. Similarly, early moves to the cloud were made despite concerns about security and reliability, largely driven by unfamiliarity. I believe that, over the next five years, catalysed and accelerated by high profile failures, people will make moves to the cloud because of concerns about security and reliability in their own architectures.

AR will finally become a thing, but we might not notice

We’ve all got predictions that we’ve been hoping will come true for some time, but the technology never seems to quite get there. Augmented Reality is mine: I’ve been wearing glasses since the age of about 11, and have been waiting ever since for the day when they don’t just do the boring job of helping me see, but present me with all sorts of interesting information about the world.

It seems as if we’ve been on the brink of useful, pervasive AR for years, with Google Glass, Microsoft Hololens, and now Apple Vision Pro. Perhaps Apple’s attempt will be the one that succeeds, but, so far, these products have struggled to break barriers of cost, convenience, comfort and privacy.

My prediction is that AR will start to show up in our lives in less obvious ways. In fact, it shows up in our lives already: for example, some cars have heads up displays projected on the screen. But even cars with more traditional navigation systems offer a form of AR: it’s just that the overlay is a bit clunky, as we have to move our eyes from the road to the screen. Similarly, when we walk down a street guided by a Maps app on our phone, we are experiencing a form of AR, facilitated by the familiar head bob to compare our surroundings with the image on the phone. (Is this street that street? And am I facing the right direction?)

Furthermore, many of us spend an increased amount of time on video calls, especially since the pandemic. How many of us can say that we haven’t exchanged private chat messages in the background, or looked up the background of an attendee on LinkedIn? Our reality is at least in part screen based these days, and we have augmented that reality with the tools we have to hand.

My prediction, therefore, is that, while the advent of AR may involve obvious devices such as headsets and glasses, it may also involve many smaller changes which add up to a big change.

Whether these predictions come true or not, they represent different ways in which we experience technology change: big shifts which come with attention and fanfare, reactions and responses to things which went wrong, and incremental changes that add up to something bigger. Whatever their accuracy, the one sure prediction is that we will have the joy and challenge of working in a field where the landscape is different every day, and surprise is part of the job.

(Views in this article are my own.)