Cheered by the news that OpenAI, the company behind ChatGPT, had released a free iPhone app for the language model, I went to the Apple app store to download it, only to find that it was nowhere to be found. This is because – as I belatedly discovered – it’s currently only available via the US app store and will be rolled out to other jurisdictions in due course. Despite that, though, the UK store was positively groaning with “ChatGPT” apps – of which I counted 25 before losing the will to live.
For example, there’s AI Chat – Chatbot AI Assistant (“Experience the power of AI! Create Essays, Emails, Resumes or Any Text!”). Or how about ChatGPro (“The Best AI Chat of 2023”)? Or Chat AI – Ask Open Chatbot (“The ultimate AI chat app that can assist you with anything and everything you need”)? Anything and everything, eh? And so on, ad infinitum. Interestingly, while the official OpenAI app is free, all of these cheery parasites, though free to download, require in-app purchases once you are enmeshed in their clutches. Same thing over in the Android universe, where something similar is under way, except that you find even more AI wannabes there than on the rarefied heights of the Apple store.
If anyone was looking for an example of a feeding frenzy, this is it. We’re beginning to see why, in 2016, Google’s boss Sundar Pichai was talking about “AI everywhere”. He may not have envisaged the hucksterish chaos currently reigning on the app stores, but at least he gets full marks for prescience. (Although it does make you wonder why his company was so taken aback when OpenAI – backed by Microsoft – launched ChatGPT last November.)
But then maybe even OpenAI was taken by surprise by what happened. After all, ChatGPT went from zero to 1 million users in five days and by January was up to 100 million, making it the fastest-growing app ever. At first it looked as if the company’s business model would be to provide access to their model as a service, charging premium users $20 a month. (Those who were using it for free probably didn’t realise that they were – in a roundabout way – doing OpenAI a favour, because their interactions with the model were helping to fine-tune it.)
But then in March, OpenAI decided that it needn’t just be a service provider – it could also be a platform on which other companies could build businesses. So it published a set of application programming interfaces (API) that would allow developers to add a version of ChatGPT to their services (for a fee, of course) without having to sink shedloads of money into building and training their own language models. This was the step that more or less guaranteed that ChatGPT would, in due course, be everywhere.
A good analogy is what happened with Google Maps. Way back in the mists of time, the Google co-founders decided that they would map the entire planet. It was a stupendously expensive and ambitious project, made possible only by the fact that their company had money to burn. But they did it, in the process creating one of the networked world’s most useful resources.
And now? Try booking a hotel, a restaurant, finding a garage, a sports venue, or almost anything else that has a geographical location, and under “location” on its website you find the relevant segment of a Google map, which is incorporated into the site using the company’s Maps Embed API.
Something analogous is already beginning to happen with ChatGPT: in time, whenever you encounter a text box on a website or in an app, you’ll find yourself dealing with ChatGPT (or one of its digital peers) courtesy of an API. In this way, Pichai’s idea of “AI everywhere” will be realised, even if the AI in question isn’t particularly intelligent.
If this sounds a bit deterministic, that’s because it is. But not because of any grand design. It’s just that, in billions of mundane cases, making use of the technology will seem the obvious thing to do. It’ll be like using a spreadsheet or a satnav. That’s why many of the current discourses about AI (née machine learning) are, as Ethan Mollick puts it, “too distantly apocalyptic”, too focused on big questions about whether AI is something we should – or should not – build. These are decisions that lie above all our pay grades, but because it looks like being a general purpose technology – ie one that can affect an entire economy – in the end, every organisation, company, school, university, charity etc is going to have to make decisions about whether and how to make use of it. Which is why it might make sense to download the OpenAI app when it finally arrives in the UK. Sometimes experience can be the mother of invention.
What I’ve been reading
Performance anxiety
Just calm down about GPT-4 and stop confusing performance with competence. So runs the headline on a marvellous IEEE Spectrum interview with Rodney Brooks.
Money talks
The rise of pluto-populism – and its consequences. Read Jonathan Kirshner’s sobering review of Martin Wolf’s new book on capitalism and democracy for the Los Angeles Review of Books.
Perfect storm
The One Best Way Is a Trap is a thoughtful essay on his Substack by LM Sacasas on the inhumanity implicit in technology’s relentless quest for optimisation.
