Hello and welcome to Eye on AI. In today’s edition…Rabbit launches Teach mode for the R1; Microsoft and HarperCollins make training data deal; Google Gives Gemini Memory; an AI pioneer cites the emerging model of OpenAI when calling for regulation; Stanford ranks countries based on how vibrant their AI ecosystems are.
Rabbit, maker of the portable orange AI-in-box device known as the R1, today released a new ability: Teach mode. This mode allows users to “teach” their devices how to perform specific actions by describing the process step-by-step on a computer using natural language and then syncing the lessons to their R1.
The new capability is a step towards Rabbit’s vision of an app store for model actions – and ultimately a whole new way of interacting with devices without the graphical user interfaces we’ve become accustomed to. It’s also the latest release from a company that has largely built its product in public (and Teach mode will be no exception). Critics at launch called the R1 clumsy, effectively useless and more of a prototype than a final product. Speaking to founder and CEO Jesse Lyu ahead of the Teach Mode announcement, he described how the company has used the criticism and user feedback to improve the product, pushing more than 20 updates, including a second generation of its LAM system (large action model). that makes it possible to interact with any site instead of just the four services available at launch. He described this back and forth as essential, claiming that the only way to build into the AI age is to bring a product into the world and go from there.
If you are one of the demonstration videos for Teach mode, you may think that this seems like the most tedious way to do something. Take an example where an early tester teaches the device how to compose a tweet. On his computer, he instructs the program to “click the icon to open the box and compose a tweet,” “click in the text box to start typing,” “type your text into the box,” and so on before the lesson is synchronized. to the R1.
No coding is required, but it’s still much more complicated than just composing a message on a smartphone or laptop, as we’ve always done. Rabbit’s idea, however, is that most users will eventually stop doing this tedious work or interact with Teach mode altogether. Instead of learning the R1 actions themselves, they can go to some sort of app store and purchase actions created by others. (Which is a bit ironic, considering Rabbit talks about deprecating apps.)
Story continues
However, the store isn’t launching today with Teach mode, and Lyu said there’s no timeline for it as they still need to figure out how to monetize it (both for the company and for the users creating the actions) to make it to earn money. . For now, users can access a limited number of lessons created by early Teach Mode testers, or create their own.
“You can imagine that, just like you train an agent, you should be able to choose whether you want to keep it public or private. And if you choose to publish it to the public community, and if a million others use your license, you should get paid. You should be able to benefit from it, right?” Lyu siad. “So I think for us this is the app store for iOS, but for agents.”
The press release announcing Teach Mode states that it is still “experimental” and that “output may be unpredictable.” This is becoming a common disclaimer with the launch of generative AI products. For example, Anthropic said the same thing last month when it released Computer Use, which allows the model to use computers like humans do.
Lyu believes this is inherent to the nature of how AI works (models are not pre-programmed, so you never know exactly what they will do) and a result of how quickly the technology is developing.
“You have to come across all the edge cases, adapt immediately and move on. That’s just the whole nature of developing with AI models,” he said. In the case of Rabbit in particular, he also pointed out that the startup doesn’t have a 10-year runway or the resources of a tech giant to allow it to take its time. “We have to make sure we take our chances and act quickly. This is the only way we can stay in the competition.”
This isn’t to say that Rabbit doesn’t test or fix issues before launch. The company worked with 20 testers to create more than 400 lessons, had them collaborate with the company’s engineers in a dedicated Slack group, and implemented improvements and security measures before launching Teach mode. Still, Lyu’s philosophy may sound unsettling to many who follow the tech industry’s “move fast and break things” mantra. Particularly in AI safety and accountability circles, there is increasing discussion about how companies should investigate each edge case internally before launch. And as my colleague Jeremy Kahn noted Tuesday’s newsletterSome believe it would be a good idea to slow down the entire train.
While Rabbit continues to work out his kinks, who uses it? And for what?
Teenagers, Lyu says, are currently the driving force behind the product. On the other hand, he said there are older users who simply find it easier to press the button than navigate through apps on a smartphone. Moreover, it finds application in specific sectors: doctors who use it for translations when speaking to patients, and truck drivers who cannot use a smartphone while driving, but can press one button on the R1, similar to how they use the radio in their use car. truck.
While it’s easy to still see the R1 as cumbersome and unnecessary, these examples of how it’s used show that it certainly has potential. I wouldn’t count the R1 (well, a future version of it) yet.
And with that, here’s more AI news.
Wise Lazzaro
sage.lazzaro@adamgale
sagelazzaro.com
This story originally ran Fortune.com
Leave a Reply