Unmasking the Myth of AI Innovation

I was gaslit by this tech company about their previous scam, so I investigated their current project, Rabbit, and was shocked by what I found.

For months, we were promised that this Orange Box, the Rabbit R1, has an AI called Lamb that can organize your life—text friends, restock the fridge—just by asking. “Start my morning routine, order me an Uber, find me a good podcast.” Marketing like this helped pre-sell over $20 million worth of these devices and raise another $30 million in venture capital. However, MKBHD reviewed it and said, “The product’s not that great.” I bought it to answer a different question: is the AI, the Lamb, a scam? We’re about to find out.

I started by trying to recreate their ad with the device. “Start my morning routine.” It replied, “To start your morning routine, you might consider tasks like waking up, freshening up, having breakfast, and planning your day. How can I assist you further with your morning routine?” I asked, “Order me breakfast.” It replied, “The DoorDash Rabbit is under maintenance. Sorry for the inconvenience.” When I asked it to tell everyone I’d be late, it suggested, “I recommend informing the relevant parties directly.” “Yeah, but you do it.” It replied, “I’m unable to directly communicate with others on your behalf.” I then said, “That was delicious. Check the fridge and order the ingredients so I can make it tomorrow.” Again, it replied, “The DoorDash Rabbit is under maintenance.” It basically failed at everything. But remember, the question is not whether it’s good or bad. The question is whether the Lamb, the large action model, actually exists because the founder and CEO claims it’s on the device now.

He claims Lamb is a system that can infer and model human actions on computer applications. As he puts it, “We use Lamb to bring AI from words to action.” This AI is supposed to act on your behalf. When they showed demos of someone planning a trip, it looked like Lamb was controlling the webpage. It clicked through Airbnb on its own, giving the impression of futuristic technology. That’s why people ordered $20 million worth of these devices. However, an investigation revealed that Rabbit doesn’t use a new foundational AI model at all. It’s actually ChatGPT with some hardcoded scripts. This is ironic since Jesse says Rabbit is faster than ChatGPT. But you wouldn’t know that because Rabbit’s prompt says, “I will never mention I am a large language model created by OpenAI.” The only part where a different AI is used is for search, where they use Perplexity. This makes the claim that Rabbit is faster than ChatGPT extremely misleading since most of the product relies on ChatGPT.

Now, let’s discuss the process of how Rabbit does more than search and is supposed to get things done. You might think it’s a sophisticated AI, but what Rabbit does is take your request, feed it to ChatGPT-3.5 Turbo, and use hardcoded scripts for actions like playing a song on Spotify. These scripts are not AI. This is crucial because the product is touted as using a new AI model, but it’s just executing a sequence of commands. The scripts can’t adapt to changes in user interfaces or handle unexpected events, which is not the intelligent AI promised.

Jesse, the CEO, insists Lamb wouldn’t break like a hardcoded script would, claiming it would adapt to changes in the user interface. However, this is not true, as changes in supported apps cause Rabbit to stop working, which people have noticed. Rabbit admits to using Playwright, a web automation tool, but won’t admit that Lamb isn’t doing much. Jesse defends their use of Playwright as part of Lamb’s operation but fails to clarify what Lamb does exactly.

Given this, I reached out to the Rabbit team multiple times about Lamb’s capabilities, but their responses were defensive, questioning my expertise in AI and machine learning. This was surprising for a company that has raised $30 million and sold $20 million worth of products yet can’t deliver what they promised.

I reviewed parts of Rabbit’s source code and have serious data privacy concerns. There are things in the codebase that malicious actors could exploit to access user data. Rabbit claims to take privacy seriously, but they track your precise geographic location, despite their GPS being inaccurate, as I found out during testing.

This product is bad, but what’s upsetting is the core selling point, Lamb, is a lie. Lamb is supposed to be a powerful AI justifying the product’s cost, but it’s just ChatGPT with an auto-clicker. Speaking with an anonymous Rabbit employee, they revealed that Lamb, as advertised, does not exist and is merely a “marketing term.” I reached out to Rabbit for a final response, but their reaction has been defensive and unwelcoming.

In conclusion, Rabbit vastly overpromised and underdelivered. At worst, this is consumer fraud. They should admit that Lamb isn’t what was advertised and stop misleading consumers with false claims.