The AI opportunity is actually in Intelligence Augmentation
If you’ve been paying attention to the world of AI and machine learning, it’s easy to believe that intelligent algorithms are on the verge of making your life infinitely better (or perhaps destroying us all in The Robot Wars).
In amongst this future-gazing there is often a critical misconception - that algorithms are even close to doing a better job than humans at most tasks.
Artificial Intelligence is doing a terrible job at outperforming humans in most tasks. And this presents an opportunity for anyone looking to build a business around Intelligence Augmentation (IA) instead of AI.
I first came across IA applied at scale through Apple Music. On signing up to the service, the playlists recommended for me were spot on - including a bizarre (for most) playlist which included Brian Eno, Sigur Ros, Bruce Springsteen, and Dire Straits. Anyone who knows me would recognise how creepily accurate that playlist is (the UX was far from spot on, but that’s another issue).
Impressed by what I thought were automatically generated playlists, I dug deeper. And discovered that Apple Music actually employed an army of real people to create playlists. The AI bit is then applied to the playlists to get the right ones in front of the right people. The AI couldn’t come up with the playlists themselves - that’s something only human intelligence can achieve. But augment the human intelligence with machine smarts, and you have an extremely powerful and effective product.
With Apple News we’ll see the same thing applied to news content. A task that many have tried to automate (Zite, Flipboard, Pulse, Feedly, and now even Pocket) is being tackled by humans. I’ve got no doubt it will be a brilliant experience (not to mention that Apple acknowledges the existence of a world outside Silicon Valley, and are hiring an Australian editor).
It’s also telling that in launching its messaging based virtual assistant ‘M’, Facebook is open about the fact that humans are actually the back end, guiding the algorithms.
The argument against IA is that AI is ’on the verge’ of being so much better. And the moment it is, AI will continue to march forward and become 2x, 10x, and 100x better.
But it’s a running joke that human-equivalent AI is ‘5 years off’. It was 5 years off in the 80’s, it was 5 years off in the 90’s, and it is still 5 years off today.
For proof at just how poor algorithmic curation is, you only need to look at the technology businesses that would benefit if it actually worked. Here’s just a sample of places where AI should be making our lives amazing, but just fails:
- Your Facebook timeline is currently generated by computers. And it still hasn’t worked out you don’t care about your friend’s baby photos.
- Your Twitter timeline is either ‘everything’, or ‘while you were away’ computer-generated timeline (which I’m yet to see actually inform me of the single most important thing that happened while I was away)
- Finally, if automatic curation based on individual signals was actually any good, advertising would be a whole lot better.
There’s a big opportunity in combining the scale of the internet with mobile devices and human intelligence right now. IA making diet recommendations based on my fitbit or Apple Health data. IA creating movie and TV recommendations to me (because Netflix isn’t great at this, and IMDB is terrible). IA to find me a restaurant based on where I am, who I’m with, and what I want to spend.
All these things are so valuable that I would happily pay for them. And yet so much energy is put in to trying to build human-less algorithms instead, ignoring the fact that the best computer on the planet is the one in our heads, and will be for a while.
Note: Curation is not the same as identification. The use of ML-based AI in areas such as reading medical imaging is amazing. Give me a choice of having my medical imaging read by a human or a computer with 99.99% accuracy, and I’ll take the machine.
- October 2015