Over the past 12 months, I’ve been on a world tour and soul searching journey – flown over 12 times, visited over 50 cities, read 100s of books and essays, and made long lasting friends.

I’m forever grateful for this life opportunity, glad that I did it, and highly recommend friends to seek such experience in life while still young. Although reading elevates perspective and maturity; visiting Rome of your time and wandering around brings knowledge and thoughts closer to reality.

Earlier this year, I was working on a startup in fintech, trying to find a way to innovate on payment networks like Visa/MasterCard using GPUs – unfortunately (or fortunately) it reached a dead end and put the project on hold.

Reflecting on what are the most important problems to solve in the current decade, AGI and the ecosystem built around it will likely be what historians will discuss most with enthusiasm in a hundred years.

I recently visited the Space Museum in Ottawa with a friend who’s an Aerospace Engineer, which sparked really exciting conversations around technological advancement. The museum has a beautiful layout where you start off with the most simple plane, and adds a layer of complexity over each model as you move forward.

This experience really got me thinking about GPTs, if in the next century someone were to build an Artificial General Intelligence Museum – perhaps visitors will start looking at GPT2 as the first model and move on to more capable models as they move forward.

I believe that a general intelligent system akin to the Wright brothers first airplane has been achieved, and neural networks at the end of this century will look aesthetically beautiful and capable the same way a f-35 plane would appeal to someone in the 19th century.

Therefore, I’m launching a new research lab XResearch to focus on AGI research.

It remains unclear what the roadmap will look like, as building these systems requires substantial resources – hence I will be focusing on problems at the intersection of [ low resources, high impact ] with first initial work around improving reasoning capabilities.

Also, I’m going to be in San Francisco and Palo Alto in late June ;)