Googler Reacts To MKBHD's 'Rabbit R1: Barely Reviewable'. ChatGPT App Beats The Rabbit.

Googler reacts to MKBHD's Rabbit R1 Review, discusses the benchmark that shows humans can accomplish 72% of the tasks, while the best models can only achieve 12.24%. The hosts emphasize the importance of the AI agent consistently delivering results and express disappointment in the Rabbit R1's performance. They also discuss the challenges in AI agent research and the limitations of current large language models.

Show hosts Joe (Eng VP) and Jordan (M&A Deal Lead) worked at various companies from Google, Apple, Facebook, Microsoft, Salesforce, Slack, Carta, Splunk, Wealthfront, Adobe, and more.

Chapters
00:00 Introduction and Overview of Rabid R1
07:01 The Limitations of Current AI Agent Research
14:05 Exploring Alternative Options for Task Automation

#mkhb #rabbit #ai #aiagents

[avi]

I am also a former googler. I have not used it in years.

[matten_zero]

Anything you can do with Rabbit, you can do with the an application. A stand alone device has to compete with your phone.

[AxisSage]

Large action model? That sounds so dumb, its still a llm but some of the function calls interface with external apps?

[lancejones]

"MKBHD" -- not "MKBHB" 🤣

[thereal_JMT_]

first to feel the AGI unfortunately I have to admit that the R1 is borderline false advertising because the agent function isn't really there. It's a neat gadget if you like the design but that's pretty much it at the moment, the whole LAM thing has to be trained first by the customers. That being said I think most of the customers are ok with that. We will see if this approach is going to be successful.

[cagnazzo]

Let's be perfectly honest... Large Action Model = Alexa, Siri, Google Home, Bixby

They simply need an LLM update and we are already there.

Rabbit is trying to borrow aspects of a phone and aspects of an AI hub, but we already have those and integrated in deep ecosystems. It's going to be an uphill battle trying to cut into all that...

... and with a new device to carry no less.