I don't know about you, but for me Artificial Intelligence always was a vague term. It was a term far away. I did not think AI could be of any use to me. But thanks to South By South West all that has changed!
Why? Because Artificial Intelligence was the main topic of the festival in Austin Texas where tech, music and film professionals gather yearly. And because there were a lot of sessions, it gave me a chance to get to know the subject. And to think about use cases.
Artificial Intelligence is not new. In fact it's pretty old. It's a term that's been doing the rounds for decades. But it's relevant now because the power of computers has increased, data storage is cheap and cloud computing is everywhere. We can turn the power of computing into something truly useful.
It helps us for example when Spotify advises us with tracks we might like. And I think Spotify is doing a great job. And it helps us when Facebook selects relevant news for us. Well does it help? Let's say it's still early days in the world of AI ;-)
What is AI in a nutshell? In the past when we wanted to predict future stuff, we tried to make a formula ourselves and tested its quality. When the result was not the way it should be, we changed the formula and tested some more.
With AI there is no human making a formula. We let the computer do all that. We feed the machine with data and let him come up with a formula for an answer to a question. We let the computer do the hard work and we let the computer learn from its own mistakes.
So for example. Say if we would want to build a hitmachine. A predictor for succes in hitparades. We would feed the computer with tracks. And we would tell with each one if it had been a hit or not. And together with the music, we would upload as much as relevant data possible. For example: the amount of Spotify and YouTube plays of the artist, Twitter followers and Facebook likes, radio plays etcetera, etcetera. And we would follow those parameters over time. And we would let the computer learn from all that data. Would that help us in predicting hits? So would we then have a means to decide if we were going to put expensive marketing money in a track or not?
Ard Boer says yes. The Dutchman built just this with the HitWizard he presented at SXSW. It's still in the proof of concept phase, but at the moment in 99% of the cases HitWizard is right when it predicts a flop. And in 56% of the cases it's right in predicting a hit. We have to add that Ard and his HitWizard only test tracks that are already on the radio, not new songs that have just been released or recorded, but it's impressive anyway.
Ard Boer: "It’s AI, AI, AI, AI, AI, AI and a little bit VR. We used a neural network. A neural network learns by feeding it data. We used 20 parameters for the HitWizard. For example: where is it played, how many times, what time of the day, etcetera. And we used the Spotify API that tells us stuff about the key of a song, the tempo, loudness, liveness, danceability, etcetera."
Is this AI stuff only for companies with large IT departments? That used to be the case. But all the big players (for example Amazon, IBM, Intel) make AI platforms based on a pay as you go model available to the world. Of course it's still a little bit techie. But that will change. It's not long before we have AI For All!
Now you know about HitWizard, an example of AI. Now it's time for all of us to start thinking. What are the questions we have? What are the answers we know? What are the data we can use? And if we don't have that data organized, how can we start doing that now?
Start thinking AI now! It can help you.
(Photo: Jochem Koole)