ChatGPT & Siblings Bootstrap The Next Computing Era: Here’s Why the World Will Never Be The Same

There’s a peculiar rule in the computer industry. Every 15 years, everything changes.

What happens is that someone will come out with some new product that, virtually overnight, the old thing is thrown in the trash, and the new thing finds its way through the innovation curve from obscurity to ubiquity.

Go back to the 1950s, and mainframes emerge from the primordial ooze to usher in the modern computing age. Roughly 15 years after that, we get the Mini-computer Era. Give it 15 years until 1981 and we see the launch of the IBM PC and the start of the PC Era. Add roughly 15 years and we end up with the Internet Connected PC Era with the release of Windows 95. Whack on a little under 15 years and we see the launch of the iPhone and the start of the Smartphone Era in 2007 for iPhone, 2008 for Android, and 2010 for iPad.

So, we’re roughly due.

There’s a sidebar here are well — the world also happens to go in 15-year boom and bust cycles and the investment in new computing paradigms roughly aligns with the troughs in that cycle. Guess what’s also happening right now?

One thing is common with these shifts — no one sees them coming. Sorry Zuckerberg, no matter how much you try and force the Metaverse Era, it ain’t gonna happen. One of the rules of this sort of shift is that it’s not the computer scientists/product designers in their Ivory Towers that choose — it’s the unwashed masses that get to decide what’s cool.

That’s one common theme to an era shift. Another common theme is that even the technologists get caught by surprise. There has to be an element of “woah, I didn’t realise we could do this?” about it. (I refer the reader back to stories of how BlackBerry’s management thought the iPhone was a parlour trick when they first saw it and flat out did not believe it was real…)

This last week has seen the release of ChatGPT. The previous months have seen the release of DALL-E. Both of these generative AI tools fit that — even people who know what computers can do are looking at this and thinking “woah!” The coding stuff that ChatGPT can do is particularly confounding.

For me, the most important hallmark of one of these era shifts is that it brings information closer.

If you are a sci-fi nerd, you can intuit that if you keep on evolving technology, given enough time humanity ends up being a trillion uploaded consciousnesses running on the silicone substrate of a Dyson sphere. Come back a few million generations and imagine getting out of a (flying) cab in Londonopolis in the year 3022. You have to go to a meeting, you don’t know where it is, but that’s OK because there’s a little implant in your brain that means that rather than pulling out your ancient steampunk phone, you just “know” where the meeting is. (If you’re feeling dystopian, maybe the “Coke is it!” jingle will play directly into your auditory cortex at the same time the information is delivered.)

Each shift brings information closer. In the Internet Connected PC era in 1995, you’d be reaching into your pocket to pull out a Microsoft AutoRoute printout of the route instructions. Nowadays, is anyone planning routes before meetings? Hardly — most people will pull out their phone as they step off the train. You don’t need to, because information is “nearer” now than it used to be.

Does the generative capabilities of ChatGPT fit that definition. Does it “bring information nearer”?

Yes. ChatGPT does something new. If you ask Google (or Bing or any other search engine) for information, it does two things. It will either give you a collection of pages that you have to work through to work out the answer yourself, or (increasingly) it will try and put the answer in front of you itself. If you ask Google “who is the UK’s shortest serving prime minister”, it will say “Liz Truss”.

It obviates a step there — it’s not asking you to go to a Wikipedia list of UK prime minsters ordered by length of tenure and expecting you to scroll to the bottom. It brings that information closer by doing that for you.

ChatGPT *only* works in that mode. Ignoring for a moment that it currently gets the answer to that wrong, whatever you ask ChatGPT operates in the modality of bringing information closer.

Admitted by my own rules, generative AI feels like it fits an era shift. One of my rules that doesn’t fit however is that there should be some physical component to it — all the other shifts are hardware based.

(And no, Mr Elon “I Love Myself” Musk’s brain implant is a Nowheresville fantasy — there’s a man who does love a folly.)

Another rule is that there has to be an element of “open tooling” — whatever hardware you come up with has to enable an ecosystem through it’s open or quasi-open nature. Think PC software, software distributed through ecommerce, and the App Store to fit the 1980, 1995, and 2007 shifts. These are necessary elements to get adoption past a tipping point.

Whilst generative AI systems can be opened up, what’s not fully appreciated is the massive costs of running these systems. By all accounts, OpenAI is covering the 5p in compute power costs for each chat on ChatGPT — that is not a cost that has good global scaling. Every other shift, scale was effectively free.

In summary then, is this the next era shift? Dunno, but get your chequebook ready — something’s going to pop, and when it does it’ll be like a money geyser…