Will AI, or artificial intelligence, take over the job of event planners? Who knows? What we can tell you is how AI can help you organize better events, today.
Hi Deevid, welcome to our studio.
Hi there, Kevin.
Artificial intelligence. We all know it from the movies. It's the future.
How far are we with the technology, today?
Well, the thing you should know about artificial intelligence, is that the technology has been a long time coming. We've been working on artificial intelligence for the past seventy years, or so. Almost as long as computers themselves have been around. I think the term was coined somewhere in 1956.
And it's funny that you mention that you know it from the movies.
The Spielberg movie. How was it called?
Oh, it was AI.
It's AI, right, right.
Yes, that's what people typically think about, when they think about artificial intelligence.
But is it that, or...
No, no, no.
Oh, that's disappointing.
It's both much more and much less than that.
It's sort of weird.
So, artificial intelligence, like I said, making computers do what they do in the movies, is all about expanding the possibilities of what computers can do, compared to humans.
So, in the early ages of the computer history, humans were mostly responsible for transforming the input in a manner that's understandable for computers. So, they needed to program it, right. They needed to provide it with specific instructions.
And as time went on, we sort of got better and better at bringing computers closer to humans, right. We even invented graphical user interfaces. Now everything is very shiny and clickable. And you've got user experience experts, that all focus on bringing that computer closer to humans.
Artificial intelligence is all about expanding the possibilities of, expanding the toolboxes of computers.
So, for instance: where is this technology today?
Well, artificial intelligence today, for instance, is being used in fields like computer vision. Helping computers see our world. In natural language processing. Helping computers understand our own language. And by expanding these tools we are, once again, bringing computers closer to us.
So, in the future we won't be talking about graphical user interfaces. We won't be talking about point-and-click interfaces. We'll be talking about, well, talking with computers.
We already do that, with CDN and Alexa and...
And I myself was quite recently surprised by this, because I've been a smartphone adopter for quite a long time. And I remember using those voice interfaces during the early 2000’s and remembering that they were absolute rubbish. You pretty much had to speak the perfect Queen's English, in order for it to be transcribed correctly.
But I don't know if you've tried these Siri's and Android systems recently, it's getting amazing. You can just talk in natural language.
I dictate a lot of e-mails, for example, in my car. Okay, sometimes it's totally off, but, in most cases it's surprisingly correct.
Yes, and that's all thanks to a new, well, a new old technology in the field of artificial intelligence, called machine learning.
So, speech to text, so, when you speak, and a computer converts it to text, is an excellent example of this. So, we in Belgium, we have a very famous history, with speech to text systems, of course.
Lernout & Hauspie, yes.
Lernout & Hauspie. The famous Flemish company that went sort of bankrupt in a very shady way.
But, it's an interesting example, because, what Lernout & Hauspie built, was a very potent artificial intelligence system, but it was still developed by humans. It was a very smart algorithm that they built.
To summarize: they listened for phonemes, which are sounds, instead of individual words, which, sort of, made it easier to recognize sentences.
But, after Lernout & Hauspie, things, sort of, came to a standstill, right. And that's not just because they went bankrupt. I think Google took a good deal of their technology. But it was, sort of, because we came to the limits of human understanding. We, humans, couldn't think up better algorithms for our speech to text systems.
But machine learning is different. So, machine learning learns from examples, right. So, all we had to do now, to create better speech to text systems, is to give it examples, right. Give it examples of speech and then show the system what the output, so, what the text version, what the expected text version should be. And so, by feeding it loads of data, we are now able to create speech to text systems, that are dramatically improving, simply by feeding it more data. Not by requiring more thinking power.
So, those systems are watching all the movies with subtitles on there.
Is that also the reason why we get...
If we go online and we have to fill in a form, then there's a Google captcha popping up, showing us some images.
Exactly, so, if there's one technology that Google has, that I am most jealous of, it's that one: captcha.
So, they manage to use this tool, to gather enormous amounts of data. So, it used to be that you had to put in weirdly shaped numbers, when you're solving captchas. That was to improve the street number recognition in Google Streetview. So that they could, more accurately, bring you to a certain house. I myself certainly profited from that, when I came to your studio.
But, these days... So, it always seems sort of incognito, what they were doing. After weird numbers, it was weird words, right. That was when they were digitizing books.
Yes, their library project, yes.
With the Google Books project. And now, people are starting to catch on, on what Google's doing.
Because right now, I have to put in traffic signs. And that's to train their self-driving car systems.
So, they use this enormous amount of data to train their algorithms. And that's why you will hear people saying something like: data is the new gold. Companies like Google...
The thing that makes, for instance, an AI algorithm offered by Google, different from the same algorithm, offered by a company like IBM or Microsoft or something, is simply the amount of data.
The training of their system.
Yes, the amount of data that they have to train their system. That's really what is making the difference, right now.
There is a test: the Turing test. To determine whether a machine's thinking is equal to the humans'. Are there already algorithms that passed that test?
Yes, these algorithms have been around for quite some time, actually. And I think one of the very first chatbots, called the Elisa, was built during the... also during the 1950's. Somewhere during the 1950's. It already passed the Turing test for some people. It's all about...
What we are all doing right now, is to make passing the Turing test more robust.
I think there was one famous example of a system that beat the Turing test, during, I think it was fifteen-minute conversations. So, blind conversations. And it achieved this by imitating, I believe it was an eleven-year-old, Pakistani boy.
But that's cheating.
It's cheating, right.
Because, it, sort of, wanted to steer away the difficult questions, by pretending that it was just an innocent street boy. While steering people to talking about the child's experiences. Of course, they programmed all this information into that. So, yes, we are already creating systems that are sort of beating the Turing test. But, it's an ongoing process. We're getting better, and better, and better at it. And the coming years will...
Well, during the next five to ten years, you will see at least as many news articles, claiming that there's a new system that beat the Turing test, once and forever.
It's a really fascinating technology. But, what I'm wondering is: today, if we look at the event industry, what could we do with that technology?
But now, not in the future.
Yes, of course every industry is going to be using artificial intelligence sometime soon.
But I think, in the event industry, there are multiple, possible avenues.
So, on one hand, for example, you have demographics, right? You, as an event planner, want to know who is going to visit your event. Who you should be targeting. Who your main demographics are. Where you should be spending your marketing dollars, or euros in this case.
This is something that artificial intelligence can help in.
There's a famous book, that's about using artificial intelligence to get medical insights. It's called: Everything wrong with you, in particular. A statistical approach. So, by analyzing large amounts of data, artificial intelligence systems can better predict the behavior of one single person.
So, event planners could use this, for instance, if there's one person that attended, I don't know, three different concerts, by certain rock bands, they can use the information of that one single person, combined with the information of millions of other people, that also attended concerts. And then it can, sort of, look for correlations, to see if there are other groups like this one person, that behave similarly. And it can use this information to then make recommendations to this one single person, by looking at the vast amount of information that's available.
So, demographics, getting to know who is attending your events, is one major use case, I think.
Another one is: crowd control. Using computer vision to analyze large amounts of people, is becoming more and more common. Especially due to security concerns. If you have a, I don't know, ten thousand people crowd, it's almost impossible to, you know, find one suspicious individual. You need hundreds of cameras and at least as many police people, to actually look at these images, actively, to actually spot someone.
Artificial intelligence can just do this for tens of thousands of people at a time, without any issue. And keep as much attention on every single one of the people in that camera image. As a normal human could only look at one person at a time.
So, that's really the strength of artificial intelligence.
And what about more artistic features? For example, if you look at a concert, there's a light show. And a light show, it takes sometimes days and days to program a light show for a concert. Or for a festival. What if you could do that by just pressing on one button and let the computer generate a light show.
Yes, technologies like this already exist. A very good example, actually. We even see artificial intelligence algorithms today, being used to generate actual music.
So, yes, that would be an amazing example.
Yes, but there are also some paintings and if you look at them, well, yes, it's called art, but I don't find it beautiful.
Is it even possible to make a show, where you now have an artistic director, who sets moods on the stage and...
Would that be possible with artificial intelligence?
Yes, definitely. So, as you said, it's still...
Okay, a lot of you will lose their jobs then.
So, AI-generated art, let's say, still has a bit of an experimental feel to it. That's definitely...
But, you do need to keep in mind: ten years ago, this wasn't possible.
Ask any computer scientist, ten years ago and he would say, like: computers making music, that's... No, we're not going to do that.
So, it's an ongoing process and it's moving very fast.
But, once again, yes, like you said, a light show, is a perfect example. All you need is examples, right. Examples of good light shows, perhaps. A combination of, on one hand, the music being played. And on the other hand, the programmation of the different light sources. So, this is, once again: examples. Inputs, outputs, that you could feed to a machine learning algorithm. And if you gathered enough data, you would have systems that would perfectly replicate a lights-person.
Moreover, you could even combine this with camera feeds, to actually track how excited your crowd is during the event and use that as a way to dynamically improve the effects of your lights, simply by looking at the mood of the crowd.
That's exactly what light jockey does. In a club.
An AI could do it better. So, an AI...
Maybe an AI system could look at the total camera feeds of an entire club and look at one single person, who's not having a good time, somewhere in the back.
And put a spot on him.
And put a spotlight on his face. In order to get him dancing as well. So, that's what artificial intelligence does best: scale.
Us humans, we work pretty well, but we have our shortcomings, we have our limits. There's only so many things that we can keep track of, at a time. There's only so many things that we can do at a time.
AI has no such problems. It can simply keep track of ten thousand people at a time, no problem. It could...
You could write one AI that does light shows for clubs all around the world. That's the major difference.
I was maybe joking about it earlier. About people losing their jobs. But, if I hear this, it's getting real. Why would you hire a disc jockey and a light jockey and a lot of other people in the event industry? If the computer can do all of that.
Yes, that's sort of a big concern, right now. So, of course, there's always the human element, right. That's typically what you will hear: people will not pay money to see a computer, they will pay money to see another human, which they can identify with.
Agreed. But, the thing is cost, right. A human costs a lot of money. If you, one day, create a computer system that could generate, I don't know, excellent techno music. Well, an event planner could save a lot of money with this. Instead of having to hire a DJ, to make a nice mix for your party, you could have an AI system that is constantly creating music mixes, even adapting to the crowd moods as well. And this for as many parties as you would want.
So, this is a very real concern. Of course, not only in the event planning sector.
In all the industries.
Look at transportation, right. Self-driving cars. What is the effect going to be of self-driving cars, on the transportation industry? And the answer is: well...
A lot of people will say: nobody knows. But I think it's rather a matter of: nobody wants to think of it yet.
Yes, or nobody wants to say it out loud.
Nobody wants to think out loud what the possible ramifications could be. And that's definitely something that we, as humans, will have to, sort of, agree amongst each other.
There will be applications that will not, you know, be preferable to have done by humans. I'm talking about, I don't know, care. Elderly care, for example. We don't want to put our elderly people in boxes, that are completely maintained by AI-systems, right. This is something that we want, we would want to do ourselves.
And at the same time, there will still be human deejays, right. Human deejays will still have fans. Human singers will still have fans, due to their persona. But it's... Yes, it's going to be less.
Yes, but even there. If you look at Hollywood. They're thinking about making AI-versions of their actors, so they can use them forever.
Right, right. It's actually been...
I've heard a lot of examples of actors, putting it in their will. That they are not to be digitized and that their likeliness is not allowed to be used by Hollywood.
Exactly. We are getting better and better at that.
There's a... I think it's notarealface.com. It's a new website, released by NVIDIA. So, you can go to their website and it will generate a face for you. Not a real face. But one that looks, convincingly, like a real person.
Once again, this technology is going to be used to make stuff like voice acting more simple. Simply generating faces. Simply generating voices out of nowhere.
It could be that, in the next ten to twenty years, you will see your first live action movie, completely generated by a computer system. Generated from screen play.
Like I said before, it's a very fascinating subject and we can talk about it for hours, but our time is up for today.
David, thank you very much for coming over to the studio.
Thanks for having me.
And you at home, thank you for watching our show. I hope to see you next time.