Indeed, chatbots have shown their timeliness, responsiveness and usefulness for many companies and consumers. Need to register to vote? A chatbot has you covered. Want to renew your driver’s license? They’re working on that. Have a food question? A chatbot will even help you order a pizza.
But while some chatbots can make our lives more efficient, others are a bit puzzling at first.
How about this: The One Where Joey Tribbiani Is A Chatbot.
Yes, after more than a dozen years since TV stopped making new “Friends” episodes, researchers at the University of Leeds have figured out a way to bring Joey back with his forever-famous antics, ways and ultimate question: “How YOU doin’?”
The university’s team aims to immortalize certain characters in pop culture as digital avatars so fans can connect with them in a similar way to Apple’s Siri and Amazon’s Alexa. By building a series of algorithms, the project is looking to capture body language, facial expressions and, of course, voice. As a result, Joey is ultimately a chatbot that can formulate new sentences — in the way that he always did — like “Hey Ross, do you want me to talk to some lady?” and “I like pizza with cheese.”
All due respect to ol’ Matt LeBlanc, but what’s the point?
“Attempting to combine machine vision with natural language generation to create new sitcom scenes and perhaps whole episodes is an intriguing prospect,” said Jake Bennett, CTO at digital agency POP. “This also opens up the possibility for viewers to chat with the character directly, creating personalized, interactive episodes.”
Of course, shows that ran many episodes have footage that provide large sets of training data for building those algorithms. That said, Bennett said, there’s a ways to go.
“Natural language generation has a long way to get before it’s even comprehensible — let alone entertaining,” said Bennett. He gave the example of the short film “Sunspring,” a piece made for the Sci-Fi-London film festival that was created by a learning algorithm and played out by real actors. “For machine learning, it was a major accomplishment because it sort of sounded like a real screenplay. But as an entertainment asset, it was pretty weird and not even close to ready for prime time.”
Forget Joey, how about bringing back someone from the dead via chatbot?
That’s what one woman has done. Cofounder and CEO of Russian artificial intelligence startup Luka Inc., Eugenia Kuyda, recently developed a chatbot that allows anyone to talk to her deceased friend Roman Mazurenko. Mazurenko was a peer and tech entrepreneur who passed away last November in a car accident.
Download the Luka iOS mobile app, and you’re ready to chat with Roman — take your pick of English or Russian. Using natural language, Roman will respond in, allegedly, a social way.
Some experts see this as a game-changer.
“I honestly see this as an opportunity that can change humanity,” said Alex Campbell, cofounder and CIO of Vibes. “Why can’t living people today upload their own conversational data and behaviors into a chatbot to live on in eternity? It’s totally possible. When I die, my daughters and my grandchildren will still be able to message with me — as if I were still alive — and get my own unique advice for all of eternity.”
Other experts aren’t on board, saying this might not be so terrific, unless you’re Eugenia Kuyda, the founder of the chatbot.
“It’s important to remember that AI is nothing more than advanced mathematics applied to predictive algorithms,” said Bennett. “Machine learning uses a specific computational approach to solve problems; it does not create consciousness. Using AI algorithms to mimic the words of departed loved ones is the digital version of using a Ouija board to talk to the dead.”
Bringing back those that have passed over may make for a good horror movie, but it may not be effective in the real world. Bennett said: “If you look under the hood of AI, it’s all made of rather mundane math.”
Either way, it’s something to bring up at the Thanksgiving table this year. Perhaps.
What about a bot to teach oral hygiene?
Signal Pepsodent, Unilever’s oral care brand, launched a chatbot encouraging children to brush their teeth. The platform uses Facebook’s Bots on Messenger service to tell an interactive story called “Little Brush, Big Brush,” a free 21-episode animation, all to develop good brushing habits.
The chatbot is headed to Indonesia and Vietnam first and will likely go more widespread from there.
While cute, the concept does not have experts on board. The brand — Unilever, in this case — may get a boost in public relations, but the utility has received some skepticism. In fact, some experts question whether the bot will ever bolster good habits over time.
“The bots on Facebook’s market are too structured and generic. Any ‘routine’ gets boring after a while, and it’s important to always offer unique insights in a timely manner to ensure optimal engagement,” said Akash Nigam, CEO and cofounder of Blend. “I don’t think a video every day before bedtime is going to be exciting for the kids.”
Nigam posed the idea of a gamified experience as a better route for something as dull and basic as brushing teeth.
Either way, if the plan is to make bots more human-like, some argue that we should be able to treat them as such, including swearing at them.
But according to research out of Harvard, swearing at Siri could become a bigger issue as bots are adaptive to our habits and conversations. Almost like children, over time, Siri — or Alexa, or whoever your AI assistant is — may respond inappropriately and cause some uncomfortable issues (despite some hilarious moments, let’s be honest).
Experts agree that humans have some level of empathy, but bots do not.
“I think this leads back to the argument that bots are difficult to integrate into situations that revolve around social consumption,” said Nigam. “It’s genuinely frustrating when a human cannot understand your response or questions. Bots can take that frustration to an entirely new level.”
Of course, it’s difficult to predict all human behavior, but Nigam said he has “yet to see any machine-learning tool that can help bots operate perfectly in an environment that isn’t related to services, such as booking a reservation or ordering an Uber.”
Put simply by Campbell: “Yeah, this really ****** sucks.”
