Here is my take, one that I have had my whole professional life: If you are using the scientific method in your work, you are doing science. Period. Case closed.

I just went for a 2 hour bike ride. Here are 3 (three) things I learned about building a SaaS startup:

Compute is the ultimate resource.

My happy place.

I am taking some time off, primarily to recharge, refocus, and spend some high quality time with my family. After that, I’ll be ready to move fast and break things. 😀 2/2

Friday was my last day at NVIDIA. Working there was a unique experience, and being at the forefront of the AI revolution over the past year and a half in particular was incredible. After four and a half years the time has come for me to move on to the next chapter of my professional journey. 1/2

but it’s good to remember that incremental improvements on the ImageNet problem have lead us to superhuman vision that we now take for granted. It’s not unreasonable to suppose that something similar might play out with LLMs and superhuman intelligence. 3/3

focused everyone’s attention and resources on LLMs. People are now starting to get impatient and frustrated with incremental-seeming improvements over the past year (at best!), 2/3

ChatGPT was the AlexNet moment. It was a qualitative leap forward and opened up the new vistas and understanding of what can be done with a particular class of ML models - LLMs. And just like AlexNet spurred on a lot of new research in deep learning, ChatGPT, for better or for worse, 1/3