One of my favorite movies, Real Genius, has a joke that seems to apply directly to many of the current discussions about AI. (It's an 80s movie, so it's not a scene, it's a montage set to “I'm Falling” by The Comsat Angels.)
In the film, the protagonist, Mitch, attends an ordinary math lecture, but as the montage progresses, most of the lesson is replaced with tape recorders of various sizes. In the final shot, Mitch walks into the lecture hall and notices that in place of the professor there is a large reel-to-reel tape recorder, which is just a single tape recording being played to all the other tape recorders.
One of the announced features of Apple Intelligence, Smart Reply, offers an easy way to respond to direct inquiries in email, asking a simple question (“Do you like me? Answer yes or no”) and drafting a response.
Apple isn't the first company to suggest that in the future your phone will write emails for you: Gmail's Smart Compose has implemented the feature for several years, and Apple has offered its own version of multi-word auto-completion for almost a year now.
But with this latest round of AI announcements, I once again heard many people joking that soon your AI will be emailing my AI, with no need for humans to be involved. This is usually couched as an absurdity, but I think there may be more to it than that.
Let's imagine AIs endlessly emailing each other, having meaningless conversations, and living their own minds. That might make an interesting science fiction story, but I'm not sure it would really matter to us as humans. Think about it this way: email is just a channel of communication. It was made for humans to talk to each other, yet we've been receiving automated emails, newsletters, spam, and more for years now.
If you're into technology, you've probably heard of APIs (Application Programming Interfaces). At the most abstract level, an API is an agreed-upon way for software to use or communicate with other software. They exist everywhere: in the cloud, on the web, on devices. But why not also in email messages?
I know it's silly to think that free-form email messages are better than programmatic APIs, but email has flexibility that other APIs don't have — you can use it for literally anything — and often APIs are underused because the people who use them are lazy, busy, uninterested, or just didn't know they existed.
Let's say you need to find a common meeting time for you and 4 other people. Is there an Internet Calendar API for this? Yes! Are there any calendar apps with built-in support for this kind of scheduling? Yes! Are there any web apps that will do this for you? Yes! (I personally use StrawPoll). Still, I think most people would just send out an email asking everyone if they can meet at a certain time and try again and again until they get it right. Not efficient, but useful.
Now imagine the same scenario, but with an AI system that reads everyone's emails and has access to each user's calendar. The end result might be the same as if you used an existing API or web app, but instead, email messages between the AIs are organized. Some AIs might know exactly when personnel are available, while others might need to check. But instead of the onus being placed on the user to interface with other systems and integrate everything, the AI does most of the work and the user simply steps in when needed.
I don't think that's an unreasonable scenario. (And certainly, if the AI is particularly smart, it might solve the problem up front by using existing calendar services.) This is the same as each person having a dedicated human assistant to set up meetings, except that probably no one has the budget to hire a personal assistant.
In fact, the real problem for AI assistants is do not have When you talk to other AIs, you talk to an AI, but when you talk to a human, you talk to an AI. Remember when Google showed off a service where they pretended to be a human and called real people to check Google Maps data or make reservations? What I really fear is being bombarded with emails, texts and calls from AIs working for people and organisations who want my attention but won’t give it to me.
I don't care if they have email communication, as long as it saves me, the human, having to read a ton of email communication between AIs. The protocol doesn't really matter, I don't care if they use iMessage or RCS, as long as it gets the job done and I don't have to clean up after myself. Don't involve me except to answer questions or make requests of my own.
Email and text messages may be a silly way to build a web of interconnected AI software systems, but history has shown time and again that the simplest solution is often the one available, not the most elegant one.
If you like articles like this, please support us by becoming a Six Colors subscriber, where you'll get access to exclusive podcasts, members-only articles, and an exclusive community.
