Rabbit co-founder and CEO Jesse Lyu isn't afraid of death…at least not of his company. He told TechCrunch that while the company is a startup whose fate can be dictated by the whims of multibillion-dollar rivals, that's no reason to give up and go home.
Appearing on stage at StrictlyVC LA, Lyu explained his rather philosophical approach to the threat of Google, Microsoft or Apple coming to crush them. (Quotes have been lightly edited for clarity.)
Rabbit's r1, the pocket AI assistant that garnered a lot of attention after its debut at CES, is certainly an original proposition. Half the size of a mobile phone, the device technically functions as a voice assistant, but it can not only remotely interact with apps and perform complex actions, but also answer questions and carry on conversations, much like ChatGPT. can also do. He described these two parts of Him as “intentions” and “actions.”
“I had this vision many years ago, 10 years ago in fact, but the technology wasn't ready. It's the first time in history that a device like this was actually possible. It’s my first time,” Liu said.
Given that he was interested in LLM's ability to understand language and intent, and given the apparent versatility of transformer-based systems, it was natural to try to get LLM to perform actions. I explained that there is.
“We immediately tried using super prompts to force this language model to do the work, and the results were pretty disastrous,” he recalls. “I have a demo from another company that I like, so please go to Mr. Beast's latest YouTube video using LLM and leave a comment. Yes, in theory you could do that with a language model. But… When you do that, you literally have to watch a screen to do it step by step. And the time it takes to complete one such task is about 2-3 minutes. That's how good it is. We don't think it translates into end-user experience.”
Their solution is a “large-scale action model” trained on “hours of real user interaction with popular apps such as Spotify, Uber, Expedia, and DoorDash.” Listing the top 800 most frequently used apps. Next, we set up this neural symbolic network and ask this AI, which we now call the large-scale action model, to review these clips frame by frame. Symbolically, the idea is that eventually the AI will be smart enough to extract all the buttons and all the elements and basically build the logic to automate them. ”
Currently using Usagi r1. Hand model: Chris Velazco of The Washington Post.
The language part is still run by third-party LLM services like Perplexity, which seems to be trying to capitalize on Rabbit's success and is offering a year of free service on top of what r1 offers. I suggested that API costs and other considerations could pose a risk to a startup's solvency.
“First of all, we are not losing money selling R1, which is a very, very, very important accomplishment, especially for a new first-generation startup. It's not going to go bankrupt either.'' Thanks to the great members of the hardware team who were able to essentially negotiate the parts and BOM. [bill of material] It’s costly,” he said. “Our orders are really approaching 100,000. He told the team two days ago in Keynote II that it would be really great if he could sell 500 units on the first day. I did.”
When it comes to subscriptions, Lyu doesn't think it will work, especially when the device claims to be cheap and simple. He said users will later be able to train and sell their own app-specific models, and Rabbit will receive a cut of that, but this is a long-term plan with no details yet finalized. I warned you.
Finally, when faced with the fact that the world's biggest and richest companies are spending billions of dollars to advance in AI, Lyu said that Google, Microsoft, or Apple (CEO Tim Cook has , who just said it would “break new ground” when it comes to AI).
Image credits: tech crunch
“It's not a delusion to think we're not a startup. We are a startup,” he said. “So the first lesson I learned from Y Combinator two years ago is that 99% of startups die. Yes, no matter what the big tech companies try, I can guarantee this will work…'' then you're delusional. That's not true. The truth is, a startup is a survival game, so it's better to spend your time focusing on yourself. ”
“They're going to do what they're going to do and I'm going to do what I'm going to do, right? I'm sure some founders quit when they heard Apple was doing Apple Cars, right? They just canceled. So? I think it's good to have this level of competition, but it only helps us grow faster or die faster, and this is why startups Nature. It's one or the other. We don't know yet. But I'm doing my best – like I said, it's a survival game.”
You can watch the entire panel below.