Summary.Jesse Lyu, the co-founder and CEO of Rabbit, is not afraid of death, at least for the company. He toldThe voice of ZhixunThe company is a start-up whose fate could be shaped by the whims of multibillion-dollar competitors, but that's no reason to give up and go home.Jesse Lui talks about the essence of entrepreneurship: either grow faster or die faster, just don't give up.
Lyu appeared on the stage of Striclyvc LA to explain his rather philosophical attitude towards the threat that Google, Microsoft or Apple were about to crush them. (The quotes have been lightly redacted for clarity.) )
The Rabbit R1, the pocket artificial intelligence (AI) assistant that has attracted much attention after its debut at the International Consumer Electronics Show (CES), is indeed an original proposal. The device is half the size of a phone and is technically a voice assistant, but it's capable of remotely manipulating your apps, performing complex actions, answering questions, and having conversations like ChatGPT. He describes these two parts as intent and action.
I had this idea many years ago, 10 years ago, actually, but the technology was not mature at the time. "This is the first time in history that such a device has been realized," he said. ”
He explained that he was intrigued by the LLM's ability to understand language and intent, and that transformer-based systems were clearly versatile, so it was natural to try to get them to perform actions as well.
Can an eye-catching design make the Rabbit S R1 Pocket AI stand out from a crowd of virtual assistants?
We immediately tried to use a super prompt to get this language model to do something, and it turned out very bad," he recalls. There is also a demo from another company that uses LLM to go to Mr. Beast's latest youtube** and leave a comment. Yes, in theory, language models can do that. But this results in you having to look at the screen and do it step by step. It takes about two to three minutes to complete one such task. We just don't think that translates into a good end-user experience.
Their solution is a large-scale action model that is trained by real users interacting with popular apps (Spotify, Uber, Expedia, Doordash, etc.) for long periods of time. We have the top 800 most frequently used apps. And then we built this neural symbol network and had this artificial intelligence, now what we call the big action model, to look back at these fragments, frame by frame. Symbolically, AI will eventually become smart enough to be able to extract all the buttons, all the elements, and then we can build an automated logic.
The language part still runs on third-party LLM services, such as Perplexity, which seems to be capitalizing on Rabbit's success by offering a free one-year service on top of what R1 offers. I think API costs and other considerations can pose a threat to the solvency of startups.
First of all, we didn't lose money because of the R1, which is a very, very, very important achievement, especially for a new company of the first generation. We don't go bankrupt by selling more products. "I give all the credit to my hardware team, who were able to negotiate lower costs for parts and bills of materials," he says. "We're close to 100,000 orders. Two days before the keynote, I told my team that it would be great if we could sell 500 units on the first day. But we sold 18,000 units.
As for the subscription, Lyu doesn't think it will work, especially when the device's theme is cheap and simple. While he does mention that users can later train and sell their own application-specific models, and Rabbit will also get a share of them, he cautions that this is a long-term plan and there are no specific details yet.
Finally, when confronted with the fact that the world's largest and richest companies are spending billions of dollars to lead in AI, LYU offers an almost zen view of the prospect of being crushed by Google, Microsoft or Apple (Apple CEO Tim Cook has just said it will break new ground in AI this year).
The voice of Zhixun
I don't have paranoia and think we're not a startup. We're a startup, he said. I mean, the first lesson I learned from Y Combinator two years ago was that 99% of startups go out of business. If your mindset as an entrepreneur is, oh, I have a genius idea, I can guarantee that it will work out, no matter how hard these big tech companies try, I mean, you're delusional. There is no such thing. The reality is that entrepreneurship is a survival game, and you're better off spending your time on your own business.
They'll do what they want to do, and I'll do what I want, right? There must have been some founders who stopped when they heard that Apple was going to do the Apple Car, right? They canceled. What now? I think it's good to have this level of competition, it just helps us grow faster, or die faster, which is the nature of startups. Either or, or, I don't know yet. But like I said, it's a survival game. ”
Everyone is welcome to pay attentionBlake's voice of intelligenceto listen to the sound of the world with you [gf]1f496[ gf] [gf]1f496[ gf] [gf]1f496[ gf]!