Whoop Stress Monitor
I pitched this idea in my interview at Google, so I’m very excited to see something similar come to fruition. At a high level, this launch offers the user the ability to see 1) a real-time ‘stress’ score and 2) overview data about daily stress. What I like about it is the standard Whoop distillation of something complex down into a simple score. The tool offers you alerts about high stress, daily recaps, and Stanford-backed breathing exercises to bring you back to baseline. My initial impression is that the metrics are reasonably accurate, though it has a hard time distinguishing weightlifting from stress.
The business strategist in me thinks that this is both a feature worth building and one unlikely to change their business position. The root of their stress algorithms is the same previously available functionality: heart rate monitoring and heart rate variability tracking. This is a feature that Apple will add (Google’s watch already has this). Whoop will remain for “hardcore” types.
AI News
The AI front is moving incredibly fast. The most interesting area for me was Edge vs Cloud, though OpenAI pivoting into a consumer company changes the importance of this. Even two weeks ago my question was whether OpenAI’s models could outrun open source forever. Now it seems like a race to takeoff before Google distributes Bard to ~1 billion+ users.
Edge vs cloud
LLAMA – Open source work has been done to enable Facebook’s open source LLM to run on the device (aka the edge). For developers, this would enable substantial cost savings. It would complicate OpenAI’s plans for sure.
GPT4ALL – LLAMA is hampered by Facebook’s licensing. Other companies are putting credible models into more usable open source formats.
Stability AI – Stable Diffusion is a very interesting model for image generation. MidJourney seems to be the hottest company here though. The Stable Diffusion CEO has teased releasing an open source Large Language Model, which would further throw things into chaos.
Microsoft and Google peppering AI into all of their productivity suites reinforces what I said about Notion AI: this will generate a substantial consumer surplus, but leveraging OpenAI models doesn’t change the field of play.
On Google: I have no privileged information here, but the publicly available BARD continues to improve quite rapidly! My immediate take is that compared to ChatGPT BARD is less ‘creative’ and fun but increasingly at parity for information.
Lastly, Google seems to be responding in Search about how I’d expect. Notice a lot more of the below lately? I have :). So much for the critique that Google can’t give you an exact answer (far faster than the LLMs can).