Quotes
Collection of quotes from others I found inspiring. I find it interesting how I have so many quotes on work and so little on life. I should work on life more!
Life
If you don’t respect your own boundaries, nobody will respect it.
Lee Yin
In work, you are replaceable. But in life, you are the only son/husband/father… you are irreplaceable.
Colleague
Happy wife, happy kids, happy life
Colleague. I found this funny as it was his reply when I asked him how his vacation went.
Before speaking, ask myself: is it true? is it necessary? is it kind?
Unknown
Work
A very common pattern for successful machine learning engineers ,strong job candidates, is to develop a T-shaped knowledge base. Meaning to have a broad understanding of many different topics in AI and very deep understanding in at least one area.
Andrew Ng. I’ll add my own flavour by saying that the horizontal base should be ancillary to your vertical. For example, if my vertical is AI LLM products, I should be great on deployment, multimodal, OCR etc.
Done is better than perfect
Yang Shun
A good mentor teaches you how to think, not what to think.
Yang Shun. Important to me as I tend to suggest what to work on specifically.
The biggest career lesson for me in the past few months was to learn from Yi, Ed, and Quoc about strategic awareness. What I mean by that is to know your own goals and incentives within the context of the broader Google organization. For instance, how does your project align with long-term goals that leadership has? What is the history and context for why a particular decision was made? I learned about the importance of considering these factors when choosing a project to work on that will not only have big impact but also garner the support of leadership. (And actually, Ellie told me the importance of garnering support for your work a while ago, but only now do I actually understand what she meant!)
Jason Wei
“In research, you cannot control the output; you can only control the input.”
Wojciech Zaremba. Great advice as I was initially focused on good research outcomes, which was stressful as we don’t have full control of the outcomes.
“When you WIN TI, it’s dangerous to think that you are a TI winner. That great should follow automatically. Instead, you should focus on the things you did that led you to being a TI winner.”
Sébastien “Ceb” Debs
“When you focus on the past, that’s your ego: ‘I did this. We were able to beat this team 4-0. I did this in the past. I won that in the past.’ When I focus on the future, it’s my pride: ‘Yeah, next game, Game 5, I do this and this and this. I’m going to dominate.’ That’s your pride talking. It doesn’t happen. You’re right here.
“I kind of try to focus on the moment, in the present. That’s humility. That’s being humble. That’s not setting no expectation. That’s going out there, enjoying the game, competing at a high level. I think I’ve had people throughout my life that helped me with that. But that is a skill that I’ve tried to, like, kind of – how do you say it, perfect it, master it. And it’s been working so far. So I’m not going to stop.”
Giannis Antetokounmpo
Don’t be trying to do someone else’s role. You can learn about it, but don’t be trying to do it especially if you don’t have access to the information and resources required to perform it. Know your role and know what is required. To be a rockstar, get stuff done. Not talked about, not analyzed, not discussed, but done.
Suzanna Sia
Don’t hold anyone to any expectations other than those that they have set for themselves. Try to understand what is the expectation they have set for themselves, and if you can help them to raise this. Rather than judging them based on your own expectations of them
Suzanna Sia
In the graduation-speech approach, you decide where you want to be in twenty years, and then ask: what should I do now to get there? I propose instead that you don’t commit to anything in the future, but just look at the options available now, and choose those that will give you the most promising range of options afterward. It’s not so important what you work on, so long as you’re not wasting your time. Work on things that interest you and increase your options, and worry later about which you’ll take. Flying a glider is a good metaphor here. Because a glider doesn’t have an engine, you can’t fly into the wind without losing a lot of altitude. If you let yourself get far downwind of good places to land, your options narrow uncomfortably. As a rule you want to stay upwind. So I propose that as a replacement for “don’t give up on your dreams.” Stay upwind.
Paul Graham. Good advice on how to make career decisions. Relating to AI, Jason Wei gave a similar advice on how one should get into an AI job despite being unsure if AI is for you: be a really strong SWE, get into a frontier lab as a SWE, then transition to AI.
How Jason Wei got into AI research: https://docs.google.com/document/d/1QREmdzLwJ0CR3kdFeenJbBowT1IFFREd46y10tW6pog/edit
Jason Wei. Treasure trove of good advice. Documentation really scales!
It seems to be not a coincidence that some of the strongest leaders in AI who manage large teams frequently do very low-level technical work.
Jeff Dean doing weekly IC (individual contributor) work while managing 3k+ people at Google Research is the canonical example, but I’ve also noticed more recently: (1) Andrej Karpathy, who managed autopilot at Tesla, made videos explaining backprop in detail. (2) Noam Shazeer being both CEO of character and writing lots of code. (3) At least 3 managers at OpenAI with big teams either wrote individual evals or gave low-level feedback on evals that I had written.
Some guesses why (though it probably differs from person to person): 1) Doing IC work gives technical insights that help them drive research or product direction. 2) Doing IC work earns respect and trust from the people they manage / set a culture. 3) They are so good at IC work that their direct impact from IC work can be higher than spending more time managing (e.g., probably Noam Shazeer). 4) They enjoy IC work.
Jason Wei. As a junior, I often feel there’s too much talking, too little doing. I aspire to always be doing IC work even as I progress.