Of Centaurs and Creative Directors
Seenapse + GPT-3 = more and better ideas, now even faster
(Versión en Español)
During one of my classes at centro, in which we were looking at emerging technology that could help them as creative professionals, we started to talk about machine learning, and its present and future impact in all of the creative industries.
I remember that all my students were quite welcoming of any tool that would help them to be more productive, like removing the background from a photograph or making people disappear from a video with one click.
But when we began to talk about software capable of writing original stories, designing interesting clothes, or proposing innovative architecture, their faces began to show concern, which was finally voiced by one of them:
“If machines can create stuff now, and they’re constantly improving, what am I going to do for a living, then? Does it even make sense to be studying this major nowadays?”
In Clive Thompson’s book Smarter Than You Think: How Technology is Changing Our Minds for the Better, he recounts how the term centaur was originally adopted by the chess community to describe a human and a computer playing as a team, “a hybrid beast endowed with the strengths of each.”
This idea of playing in tandem with a computer came from none other than grandmaster and world champion Garry Kasparov — the person who had been defeated by the strongest combination of hardware and chess-playing software available in 1997, IBM’s Deep Blue supercomputer — once he recovered from the blow¹.
Meditating on Deep Blue’s capabilities gave Kasparov an audacious idea, writes Thompson. “What would happen if, instead of competing with one another, humans and computers collaborated? (…) That way, he theorized, each might benefit from the other’s peculiar powers. The computer would bring the lightning-fast —if uncreative— ability to analyze zillions of moves, while the human would bring intuition and insight, the ability to read opponents and psych them out.”
Thus advanced chess was born, and with it the centaur concept, which is different from human augmentation, because with centaurs² the emphasis is in the collaboration between human and computer.
I like very much Kasparov’s approach, and I believe it’s a great way to think about our relationship with machines, particularly for creative endeavors. Some of them are definitely more than tools now: They’re able to suggest and propose things to us, in some ways much like a human partner would.
Back in 2014, we noticed that the way information is organized on the web, and is made accesible by search engines, is not helpful for creative purposes — so we started building an alternative, with a completely different organization principle: the mental association.
Mental associations are the raw material of ideas. Creativity is about connecting the dots, as Steve Jobs said. And so having a space in which people around the world (the more diverse, the better) share their mental associations is not only theoretically good; it has proven to be of value to Seenapse users.
However, we know that for many people it’s confusing to see mental associations as search results. A typical reaction is “what does this have to do with the project I’m working on, with the problem I’m trying to solve?”
Paradoxically, thinking of the answer to that question is what makes Seenapse work³ —it’s what triggers ideas in your head— but most people expect to see actual ideas in their search results, something less abstract to work with.
Doing that in an effective, scalable way, was very difficult, though. But then GPT-3 came along.
GPT-3 is a language model capable of generating text that mimics human-produced text pretty much perfectly. Even when it produces nonsensical sentences, they still feel human⁴. Since its beta release a year ago it has spawned a plethora of applications. And one of them, of course, is generating ideas.
Since GPT-3 by itself can generate ideas, is there need for something like Seenapse anymore?
Well, I’m happy to report there is.
Seenapse’s motto is “Inspired. By Humans.” We’ve always felt that the capacity we have to inspire each other is quite remarkable. To be able to build on each other’s ideas, and take them somewhere else. And it turns out that software like GPT-3 can also be inspired by Seenapse users and their mental associations, and produce more interesting ideas than those it generates on its own.
Let’s look at an example:
When we ask GPT-3 to create a summer advertisement for Heineken 0.0, a “great tasting, zero-alcohol beer”, these are some of its outputs:
A refreshing summer drink for all those who don’t want to be tied down with alcohol. The perfect beer for the upcoming months.
Heineken 0.0: Great taste, zero-alcohol beer.
Heineken 0.0 is a delicious taste, with no alcohol.
Summer is a time for good times with friends. The cool taste of Heineken 0.0 and the refreshing feeling will have you celebrating all day long!
When asked the same thing, using Seenapse’s mental associations as inspiration, GPT-3 outputs things like these:
Pour your favorite part of the day with zero worries.
Summer is the perfect time for memories. It’s time to reminisce about late night swims under the moonlight with your friends. All while enjoying a refreshing Heineken 0.0
We want your perfect summer to be nothing short of that.
Heineken 0.0: The Perfect Morning After
Comparing the two, we think that when Seenapse is added to the equation the results are richer, more evocative —but you be the judge. We’re currently integrating GPT-3 to Seenapse and will be beta-testing it with users in a month or so. If you’re interested, make sure to create an account, and we’ll notify you when we roll it out.
Going back to my student’s question, does it make sense to pursue a career in the creative industries in this day and age? Of course it does, and probably more than ever.
Yes, machines will be creating ideas, and at a much faster pace than we ever could⁵. Brainstorming as we know it will cease to exist. The job won’t be about having ideas —it will be about identifying the good ones, and to tweak and recombine and polish them, until we have something that we feel is worth producing and bringing into the world.
This particular collaboration between humans and machines, this centaur, is what we already know as a creative director. And we’ll be needing more and more of them.
Notes
¹ A bigger blow happened more recently with the game of Go, and I wrote about it at the time. See The Divine Move and The Human Move.
² That goes to show one should choose their metaphors carefully, because a half-human, half-horse creature seems like a better choice for symbolizing human augmentation than a team, but anyways.
³ An adept Seenapse user can easily generate several divergent ideas in minutes —and more importantly, ideas they couldn’t have come up with on their own, because they were exposed to other people’s heads by looking at their mental associations.
⁴ The model is very impressive, but there’s no magic involved, of course. The reason it feels human is because all of its input is human-created text available on the web, in repositories such as Common Crawl.
⁵ With Seenapse you will soon be able to generate hundreds of divergent ideas in less than an hour, for example.
