Artificial intelligence is having a “holy shit” moment.
Dag Kittlaus would know. He co-founded Siri, Apple’s ubiquitous voice assistant. His reaction to Siri’s descendants—specifically large language models like ChatGPT—was simple: “Holy shit.”
“It’s an absolutely incredible leap forward,” Kittlaus said of the AI chatbot, a skillful mimic of human speech that has set the tech world on fire.
Kittlaus was speaking during the opening keynote of the Project Voice conference in Chattanooga, Tenn., a remarkable gathering of stakeholders in the world of conversational AI. The audience included suppliers, investors and a number of restaurants interested in how they might use things like voicebots to automate order-taking.
The opening panel, which included Kittlaus’ fellow Siri founders Adam Cheyer and Tom Gruber, was a fitting way to kick off the three-day event. Because while all three seemed awe-struck by the progress AI has made, they also agreed there’s a whole lot we don’t know yet.
“The [large language model] revolution is as important as the web,” Cheyer said. “The next decade will be figuring out how to apply it.”
That process is already raising some pretty deep questions, both for restaurants and society at large. Here are five thoughts for operators to ruminate on as we enter the uncharted waters of AI.
Will AI replace human workers?
Yes. That’s the idea, at least.
While restaurants dabbling in automation have mostly tiptoed around this point, vendors hawking their wares at Project Voice made it clear that their goal is to one day replace humans with AI in places like restaurants, hotels and more.
“There will be a time when it will be replacing [staff],” said Dan Vorderstrasse, senior sales manager for AI voice provider SoundHound. But for now, he said, it’s more about filling gaps in the workforce left by the pandemic.
So … what happens to those workers?
Many have suggested that restaurant employees replaced by bots can simply get a new job developing and maintaining said bots. But that might not be realistic.
“Some [workers] are not retrainable,” said Barney Stacher, head of sales for SapientX, a maker of conversational AI avatars. “Some cannot elevate.”
Stacher suggested that as more humans are replaced by AI, there will be a need for an “AI ROI tax” to expand the social safety net. For instance, he said, if AI is saving a restaurant $40,000 a year in labor costs, a portion of that would go toward supporting the people the AI put out of work.
“I think there needs to be a really deep discussion of this,” he said.
Of course, a tax on AI would no doubt have a chilling effect on adoption.
One of SapientX's AI assistants. / Photo by Joe Guszkowski
Can AI tell if you’re hungry?
One of the most fascinating topics discussed at the conference was voice biometrics—the ability for AI to tell how you’re feeling based on the sound of your voice.
A company called Canary Speech said its software can now tell by someone’s voice if they are tired, stressed or depressed. And it’s developing models that can identify health conditions including Alzheimer’s and Parkinson’s.
Besides the obvious healthcare applications, this could have all kinds of uses in customer service and HR. Imagine if restaurants could tell if a customer is angry or screen employees for stress. Or if a restaurant could tell if a customer was sick.
“Do I want KFC to have my health information so it can tell me not to have so much KFC?” Stacher ventured.
That scenario is purely hypothetical and a long way off. But at the very least, AI could soon be used to identify restaurant guests via their voice or face and automatically load their preferences, past orders and loyalty points for a personalized experience. Panera Bread is already testing a version of this using palm-reading devices from Amazon.
Is AI safe?
If the answer to the last question sounds creepy to you, you’re not wrong. AI’s ability to collect highly personal information could become a massive liability in the wrong hands.
Multiple attendees framed the coming years as nothing short of a battle between good and evil, making it clear there will be plenty of bad guys looking to exploit AI and users’ personal data.
It’s probably the biggest ethical issue facing the technology today, and the industry is already taking steps to prevent it with efforts like the Open Voice Network’s Trustmark initiative. The program recognizes companies that are committed to trustworthy conversational AI.
There’s also the fact that AI can get things wrong. Large language models like ChatGPT, for instance, are essentially playing solitaire with words instead of cards, guessing the next word in a sentence based on the words around it, said Siri co-founder Tom Gruber.
That means the response won’t always be accurate. And accuracy is important in restaurants.
“How do you train it whether to know if it’s right or not is a whole other problem,” Gruber said.
Will customers want to use it?
Anyone who has ever dialed a company’s help line has dealt with a voicebot. It’s not always a pleasant experience. This could be a challenge for restaurants that outsource some duties to AI.
Jay Ruparel, CEO of drive-thru voice company Voice Plug, said when customers are first given the option to speak to a human or a bot, they almost always choose a human. But as they get more familiar with the technology, “there’s been a huge shift in the preference to now speak to an AI,” he said.
He added that some people are unaware that they’re speaking to a robot. Indeed, suppliers are working to make AI more human-like. Aadu Pirn, director of engineering for Voicify, said a good bot will have a 90% success rate and a response time of well under two seconds. Otherwise, the conversation will feel, well, robotic.
They’re also trying to make the bots fun. With AI avatars, SapientX wants to give restaurants the ability to use mascots, brand ambassadors or animated characters to take orders.
“Wouldn’t it be nice to speak to the Colonel?” Sacher said.