What's the point of AI?

Hello!

Thanks for coming back to Speci. It's been a busy few weeks, where we've somewhat unintentionally spent a lot of time talking about how Meta are on an absolute roll acting like Apple in the early 00s, especially now that Apple is the incumbent innovator rather than the disruptor.

If you've missed it, here's a brief summary of the last month or so:

• Meta’s Orion Swagger was all about how brazen and confident Mark Zuckerberg has been lately. Particularly, he showed off smart glasses Meta will never ship at Connect. This was the right choice for them, and as noted by John Gruber on a recent episode of Dithering, is an effective Osbourning of the Apple Vision Pro.
• I bought a Boox Palma, an e-ink, phone shaped device that is incredibly limited in functionality. Instead of the iPhone (Reader… I still bought the iPhone 16 Pro a few weeks later. I’m sorry).
• I tried the Ray Ban Metas, and was more impressed than I expected to be. Turns out cameras on your face and Meta AI in your ears is cooler than you'd think – especially when they look just like real Wayfarers.
• And finally, the most popular post amongst the Speci community last month: How AI infiltrated the iPhone event. I argued that Apple were uncharacteristically confused in its messaging about Apple Intelligence (and it continues with TV spots and billboards about a feature that still doesn’t exist yet).

If you’re not subscribed yet, please consider it! I share links to my posts on socials, but the quickest way to get the latest is to subscribe. It’s easy, free and lands in your email inbox the moment a new post is published.

Anyway, today it's time for something a little different. Let's get into it.

The year is 1721. An enterprising man by the name of John Lombe has just opened what would become the world’s first factory: a water-powered silk mill in Derby on an island in the River Derwent.

Arguably, Lombe’s new venture is the beginning of a revolution that would slowly, then quickly, take over the world – a technology revolution.

The mill was built to replicate Italian silk throwing machines on a larger scale. Italian silk throwing machines were operated by hand: Lombe’s innovation was that his factory harnessed the river to power his machines. Though gruelling human effort was still necessary, Lombe had just invented a machine to increase the productivity of human effort.

This is the very definition of machinery, and technology by extension. Augmenting and improving human capacity.

By and large, this has the been the history of machinery and technology. As the industrial revolution flowered in the decades following Lombe’s innovation, machines like the steam engine, spinning jenny and electrical telegraph all sped up inherently human things.

The steam engine brought muscle and power, moving items from one place to another, across the factory floor and eventually across the country.

The spinning jenny revolutionised the textiles business by increasing a human’s output. The first spinning jenny let a human control 8 spools at once. The technology kept getting better too, peaking at a huge 120 spools, a massive increase in productivity.

The electrical telegraph dramatically increased the speed at which humans could communicate: no longer were letters only way to convey information over a distance. The telegraph turned signals into messages, and long haul communication was born.

These innovations share a history with computing. They are technology in the same way that a computer is technology. Their sole purpose is to augment human capacity: increasing and enhancing productivity.

Calculators, word processors, spreadsheets, databases and email all augment human capacity. Calculators and spreadsheets help humans do maths. Word processors and databases made it effortless to document information. Email shared this information at the speed of an internet connection. This meant humans could complete tasks quicker and do more as a result – no longer were humans taking 10 minutes to complete a complex maths problem when a calculator allowed them to do it in ten seconds. Word processing and the internet meant publishing was no longer dependent on paper and ink; information could be shared and gained instantaneously.

Smartphones, cellular networks and app developers made these human augmentations faster and put them in peoples’ pockets. Think about this astounding development for a second – even fifty years ago, to learn something new, a human would need to go to a library or ask a knowledgable friend. Now, they can pull out their phone and ask Google or watch a YouTube video.

Technology has always existed to augment human capacity.

Determinism

I hope you’ll forgive me for the history lesson in a newsletter about technology and marketing, but I think the role history plays in both is important. You’ve got a little more history left to come, but stay with me.

Of course, I left out a whole host of important innovations, but for the sake of brevity the history I charted out above is broadly correct. Humans, even before 1721, had and have a knack for making things that help them do things quicker. Tools. Technology.

All of these things are deterministic technologies. You give it an input and it is designed to give you a fairly reliable output of some kind. The silk mill, if set up correctly, would turn raw silk into a fine thread. The calculator is programmed to receive numbers and equations and turn them into answers. Smartphones are very complex calculators, programmed through incredibly complex software to give outputs based on inputs. As you scroll this article on your phone, it’s receiving the touch input of your thumb and moving the screen down as you do it as an output. This is remarkable stuff.

But it’s all the result of human beings having anticipated these actions and writing rules for it. That’s what computer code is, and it’s how the calculator and ultimately the silk mill also worked.

The probabilistic breakthrough

Now, let’s talk about AI. Technologists everywhere cannot stop talking about AI because it turns the history of human augmentation through technology on its head. Most AI applications, the most popular you’ve probably experienced being large language models, are probabilistic rather than deterministic.

Simply, this means that rather than being programmed to take inputs and give reliable outputs, AI is fundamentally different. To massively oversimplify, large language models do not take inputs to give programmed outputs, they take your instruction and use a huge array of information it has been trained on to guess what you want, then receives feedback on whether it is what you wanted or not. Then it learns whether it’s giving you the right thing based on the task you’re trying to carry out and whether it actually helped you.

This is a huge shift in how we approach technology. Because here we’re not programming anything, we’re just… letting the technology work for itself. That’s why we call it artificial intelligence, because, well… it feels intelligent.

This has got everyone incredibly over-excited. It’s a fundamentally different approach to technology. It’s such a fundamental shift that technology companies are bending over backwards to show how ahead they are in implementing the tech. As I noted a few week’s ago, it’s even knocked the best product marketers in the world off their game because they’re so desperate to show off their AI story.

We don’t know how to use AI yet

So here’s the wild thing about AI: it seems to have arrived earlier than anyone expected. Companies like DeepMind and OpenAI have been training AI models for years, expecting us to reach a magical moment in the future where AI does stuff for us. They didn't think it was coming yet.

Except in 2022, an experimental tool called ChatGPT went viral and changed the face of AI: it was the first large language model application that blew people’s minds and gained a critical mass of interest and usage. Completely by accident.

Since that moment companies big and small have been battling for the consumer AI crown, inspired by the phenomenal demand ChatGPT was able to create. Chat boxes have been shoved where they don’t belong, Google’s search engine is floating AI generated answers above their results and Apple is summarising your notifications (or, about to).

Yet none of these things feel like the AI application. They’re awkward – the Google Search AI summaries are good but not better than the websites they’re trying to replace. Apple Intelligence notifications can be strange and soulless. Brands that have no business offering AI experiences are shoehorning them in in an attempt to win the next wave. Let’s not even talk about the Rabbit R1 or Human Ai Pin.

Looking for a problem to solve

So we have an exciting new technology that has come before its time. We have a bunch of technology companies, from some of the biggest in the world to small upstarts, creating AI experiences that don’t feel right, either inside existing products or by creating new ones.

Then there’s the narrative that AI hasn’t had its ‘iPhone moment’ yet. A killer AI device that makes AI make sense. Sam Altman, Open AI’s CEO, has apparently paired up with Jony Ive, ex-longtime design chief at Apple, to try to create such a device.

Through all of this, there’s this belief from many that AI is going to solve one big problem for us. It’s a technology, that will become a product, like the iPhone or Google Search or TVs or the telegraph, that will fundamentally change something.

And because it’s a brand new technology paradigm, unlike the deterministic computers of today, it will replace humans. It will ape what humans do.

That may well be true, for some tasks in some situations. But here’s the thing: understanding history helps us understand the future. And history shows us that machines, like computers themselves and machines before that, are about augmenting human capability. Not replacing it. I don't think it matters whether it's probabilistic or deterministic. It's all... technology.

AI is impressive. Without question. I’m using it everyday. I have ChatGPT as a Spotlight-like set up on my Mac (it’s always there by pressing option-space, much like Spotlight is if you press command-space) and I have it hooked up to the Action Button on my iPhone. So I'm using it all the time. I’m fully in. I’m finding it incredibly useful.

And in some ways it is replacing humanity for me (replacing my knowledgable friend or the next person to write a book on something that’s already happened), but it’s mainly augmenting my own capacity. I’m using it to shorten the time it takes to do labour-intensive tasks. Like Lombe’s silk mill in 1721.

Settle down, everyone

We’re in a strange moment, at the moment. On the one hand, technology products are packed full new, slightly weird of generative AI experiences. On the other, companies are desperate to have the next iPhone-sized hit, but AI.

I don’t think AI will have an iPhone moment. And I think companies will stop forcing weird AI experiences on their customers.

This is also not me saying that AI will go away. It will not. There will be more of it than ever.

It’s just that people don’t buy technologies, they buy products. AI is not enough to be a killer product in itself. It has to be packaged in the right way with the right technologies.

So whatever Jony Ive and Sam Altman are designing together: it will need to serve a real human need ready to be augmented. It can’t be a general purpose AI device like so many are predicting.

And Google, Apple, Samsung, Notion and everyone else offering weird chat boxes or clunky summaries or writing tools or image generators will calm down too. They’ll see what customers are using, liking, loving. They'll remove the things that don't get used. They'll add more of what does.

AI, which is having a big in-your-face moment now will fade into the fabric of what we do every day. ChatGPT is arguably already doing a very good job of this, and that's because OpenAI, very smartly, realised that ChatGPT is a consumer product, so they started treating it like one. The result today is a tech-demo-turned-fully-fledged chatbot that millions around the world love to use.

But there won’t be an ‘AI product'. It’s not an iPhone, which is the packaging of multi-touch and cellular networking and software and cameras and processors.

AI will become one of these components, and it'll fade into the fabric of what already exists, like so many enabling technologies before it. AI is a technology, to augment human capacity.

AI is not a product. It's a technology.

And people buy products, not technologies.