If you want a unique, somewhat unsettling, and downright hilarious theatrical experience, you could do to check out Deep Learning With Merrill Grambell. When I saw it at the Improv Asylum in Chelsea (in the former Upright Citizens Brigade theater), I didn’t quite know what to make of it. The first live show hosted by a bot named Merrill Grambell, Deep Learning is not exactly a seamless combination of artificial intelligence and live theater, but it’s a show you’ll probably remember.
The setup of the show is that the bot (who is essentially an animated disembodied male head who is inexplicably inside a coffee cup and sometimes turns to a boat, a flower, etc.) interviews a panel with a set of predetermined questions. The panelists are different every time — the show at Improv Asylum had video game writer Dan Ryckert, animator Tala Schlossberg, and gaming personality Gerard Williams aka HipHopGamer. The panelists seemed just as amused to be part of this show as the audience was watching it.
Now the idea is that the bot would interact with each panelist by responding to them, but his answers mostly ended up being unspecific and vague. There is also some plot having to do with finding a blue key, and then a random game show is introduced mid-way through. At one point there is a screening of one of Schlossberg’s animations. There really is no rhyme or reason to any of it, and some of it makes no sense at all, but it’s pretty fun to watch.
Deep Learning is the bizarre brainchild of Will Brierly, who owns video production company Snowrunner Productions. I had the pleasure of speaking with Brierly about his robot creation and what he has coming up next.
RTWS: What a unique concept Deep Learning with Merrill Grambell is! Where did the idea for the character and this type of show come from?
WB: The idea has been in development for a few years now, the first version of it was a bot I made named Krenshaw. He moderated a panel I put together at PAX East a few years ago. I’m very interested in communication between people and also technology and I thought it would be funny to see how well a bot could moderate a panel. His first performance (here’s a clip) was okay (the panelists were all fantastic though) but in the end I feel it needed more work to be a full show. At one point during the show Krenshaw accidentally ended the show in front of a ton of people with about 15 minutes left. If I was to restart him he would have done this whole long intro that couldn’t be skipped so instead I hopped under the desk and did a Q&A in his voice pretending to be him. It was a terrifying and fun experience! So I kept developing it and it’s come a long way since then!
RTWS: Is it always the same character, Merrill Grambell, or did you develop others like him?
WB: I’ve actually made a bunch of AI based characters, some of them have been in shows and some have performed on their own performances (here is a procedurally generated stand up comic I made named GARF getting booed off stage) and here he is performing at the Museum of Fine Arts in Boston (some people love him and others do not). I’ve made a whole bunch of characters where you can pretty much just press play and they will perform. With the Deep Learning show I want to showcase these and new characters where Merrill sometimes is the host leading the guests through the experience while other times it will be other characters. Because the show is based on AI we never really know what will happen during it too.
RTWS: Where does his name come from? It sounds almost as random as some of his answers!
WB: When I was initially building him I thought he kind of looked like Mr. Belvedere from the old TV show. That character’s first name was Lindsay, so Merrill’s original name was Lindsay Grambell, but then we thought that sounded too much like the politician Lindsey Graham so we changed it to Merrill and it seemed to fit.
RTWS: Not to ask you to give away your secrets or anything, or to delve too much into the technical aspects, but how did you program the bot? Is he intentionally designed to give hilariously vague and somewhat random answers, or do you have plans to refine it to give more detailed, personalized answers? Admittedly, I think the randomness of his speech makes it really funny, and I think it allows the audience to kind of let go and not try to make too much meaning out of the words. It’s sort of an experience unto itself.
WB: It’s a mix, most of what you saw at the show was somewhat vague and random answers, I actually used to perform for years as a mentalist/magician so a lot of techniques Merrill and the other bots use is actually some of the same things that make mentalism work. There are also times and things that can be triggered where the bots can understand quite a bit of what the guests are saying and sometimes that can totally change the direction that the show goes in based on what they say.
RTWS: At the show I saw, Merrill Grambell interviewed a mixture of animators and gamers. At one of your more recent shows, he interviewed John O’Hurley who played J. Peterman on Seinfeld. What other types of people has the bot interviewed?
WB: So far he’s only had three full live performances, but in those, he’s interviewed game developers, comedians, actors, writers, animators, TV producers, podcasters, reality TV stars and more. He also made a little appearance on Adult Swim’s streaming show Desperate Losers.
RTWS: How do you choose the panelists for each show?
WB: I like the idea of having sort of a theme to a show, but that could mean a bunch of people all from the same industry or not. I mostly just want to have people I find interesting that I like that I think would play well with everyone else on stage.
RTWS: Do you program Merrill differently for each show, depending on the panelists?
WB: Yup! He is trained on every guest on the show so he has different things to talk about what they might be interested in.
RTWS: I really love the concept of using AI and other technologies within the realm of theater and live performance. How do you see this type of AI playing out in the future of theater? Do you even view it as theater, or what would you call it? AI Improv? Talk-show robot?
WB: It’s a mix, I think it’s neat to use AI and other tech to make a show but in the end the goal is to make a show that people will enjoy and hopefully they are not thinking about how it works but just enjoying the show. I like to think of the AI serving a bit as improv prompts that hopefully bring out the best in the guests. It’s also nice with extra visual elements that would normally be either extremely expensive or difficult to do if they were physical sets, but if it’s made by a computer program that I made that can potentially generate infinite content that gives a lot to play with. Also, communication generated by a computer can sometimes be random, awkward, and silly and I just personally find it funny to see what it comes up with. But more than that I love to see how the guests respond and react to it. It’s as if these 4-5 strangers are going on a journey together (they have no idea what will happen during the show) and their AI bot is their guide in the adventure as the audience watches.
RTWS: Where can people see your show? Are you planning on returning to Improv Asylum in NY, or another NYC venue?
WB: We have more dates to be announced at Improv Asylum NYC coming soon (will be announced on the new website still being built at the moment MerrillGrambell.com soon), also some big comedy festivals this fall, and he will be moderating more panels at various conferences across the country. We are also in talks with some networks interested in turning it into a show based on the live performances. This is still in the very early stages but that could be very fun if that happens too!