May 9, 2024

I am an MIT Senior Fellow, 5x-founder & VC investing in AI In the brand-new world of AI, we’re slowly learning that there's a big difference between a small routine transaction like buying a hamburger, and something much more complex and high-stakes. As people wonder about ‘what are the limits of AI and what can it do?’, we have some interesting examples cropping up, including one that came up in our Imagination in Action event in the ‘startup alley’ portion, with some representatives of a company called MultiOn. “What we are trying to do is, we are trying to build the next generation of AI agents that can take actions for you,” said Div Garg, who is actually the co-founder and CEO. “And how our agents work is: they can control your computer … they can control a browser, and they can interact on your behalf. So you can talk to the AI, and the AI will go do things for you.” Runner cyborg girl is running triathlon in the future. Who's the best in the future? Will the ... [+] difference between men and women disappear?.. getty For reference, here’s more on the company’s mission statement: “The company’s mission is to unlock and elevate human agency for everyone, freeing us up from the drudgery of daily boring tasks to focus on doing what we love. The vision is not of machines replacing humans but augmenting them. Through AI Agents, the team aims to amplify the human potential, allowing individuals to truly focus on what they love. With their AI products, MultiOn aspires to unlock parallelization for humanity: Imagine a world where tasks are conducted concurrently, no longer bound by the linear constraints of time of a single person.” Kinda sounds like something that could have been written by AI. Remember FUBU? (For You, By You) – FAIBAI. Anyway, MultiOn can do some interesting things. Garg first talked about the application performing some routine transactions with DoorDash, or booking airplane travel. Those are sort of mundane uses, but then he transitioned into the idea that the application interface can actually go out on the user's behalf, and negotiate and buy a vehicle, in this case, a cybertruck. Screen video showed the AI working overtime to communicate with the dealership and get an actual cybertruck bought and delivered to a remote location. The team called someone named “Brian,” who, after some technical difficulties, confirmed that the cybertruck in question was on site. future car getty Garg explained more about the interface: “We actually have this user mode where it stores your personal information,” he said. “It has access to my credit card info – they're filling out my purchase details using a one time password, and then I can actually say ‘okay, this looks good - I can actually place this order.” In other words, there does seem to be a human in the loop, although that person only has to sort of sign on at the end, maybe, perhaps, something roughly the equivalent to granting another person power of attorney. And you've got to be careful doing that! As they try out this powerful technology proxy, you can hear the MultiOn folks giggling at the prospect that they've just ordered a vehicle in a matter of seconds! It's an eye-opener for a lot of people – the idea that AI is not just prepared to act as an intermediary for transactions that we might do every day. For all of the history of automobiles, something like a car requires people to sit down with people, talk about insurance, haggle over prices – fill out forms… But the extreme streamlining of digital activity with AI and the Internet is making all of this effectively go away – Garg explains that the application is working through the cloud and with its permission levels, there's really not much that the human has to do. With the curtain down on that particular demo, let's turn to the other side of this coin – if you Google the process of AI buying vehicles, you're likely to come up with the cautionary tale of a GM dealership that sold a Chevy Tahoe for a dollar. Well, they didn't actually have to sell it for a dollar because of some kind of legal fancywork – but the liability exists, and it is real. Experts are warning companies that if users are able to negotiate a ridiculous low price with an AI, that may become legally binding, or the company might have to pay a settlement. So a lot of the risk is on the corporate side, which tends to make people more gleeful about these use cases. Still, we should be paying attention to how applications test these transactional limits, because fairly soon, it may seem like there are no limits at all! Follow me on  LinkedIn . Check out my  website .