Why Your Technical Expertise Might Not Be Your Greatest Asset
Why LLMs are pushing me from Quantum Theory to Coffee Chats
Technical prowess has always defined my value as a physicist (turned quant trader). Part of me wants it to stay that way because of how much time I've already spent on it. I've spent hours in the Cambridge library studying quantum field theory, high-performance computing, and doing arduous mathematical derivations. Last month, I spent a day grappling solved problems in materials physics, just to feel the thrill of getting something right. I don't want that experience to go to waste.
But now, LLMs solve technical problems with a level of accuracy and speed that makes my years of effort feel increasingly obsolete. Doing problems by hand feels like an artisanal activity, like water-colour in the age of the camera. For now, I still have to sense-check things the LLM says. But I'm guessing it'll be less than two years before I won't need to do that anymore.
I need to solve things to feel useful, make money, and stay interested. So LLMs threatening to take that away has triggered an important question:
In a world where most questions are easily answered, how will I continue to add value?
Even though I feel so so uncertain about this topic, I think there are at least two partial answers applicable right now. The first is learning what to ask LLMs, and the second is building judgements on whether answers are 'morally good'.
Learning what to ask LLMs involves figuring out what's important. And the first months of my biosecurity investigations have pushed me to find ways to do this better.
One killer technique has been reaching out for feedback on preliminary ideas. After doing 'technical work' in the morning to build minimal projects, I've been spending afternoons emailing, texting, or connecting with people asking for feedback on these minimal buds. I usually just ask them if anything sticks out as noteworthy in a polite email. I don't use LLMs to write these emails, and I try to ask people across different sets of communities. These last few months I've been reaching out to policy people, building-safety people, physicists, artists and friends.
Feedback recalibrates me to what people care about. I find myself almost always surprised by feedback - not in terms of whether an idea is good or bad - but rather in terms of what people find noteworthy. I've found this evident in my new habit of sending essays and lists to people before publishing; the sentences that I think are noteworthy often aren't, and passing points often are flagged as crucial.
I think feedback has a kill two-birds-with-one-stone mechanism. On one hand, I'm checking for significance, but simultaneously, I've built a nugget of importance by mechanically making more than one person involved. I'm lucky that when I reach out to people, they usually are supportive enough to be invested one way or another.
An example is when I sought to build physics-type content for people also interested in biosecurity. I got feedback that the idea of a curriculum was not too interesting in its current form, and that question sheets would be much more exciting as a medium. I then sent those problem sets to about ten physicists. In addition to getting their feedback, ten more physicists are now aware that there are cool technical problems in air ventilation physics!!
Wide feedback gathering is a skill that I've neglected as a quant trader and technician. Markets are win-or-lose, and so the feedback mechanism of trading is strong enough that you don't need to ask actual people.
But most non-profit type problems are wicked, undefined and difficult. Something like biosecurity has no clear answers, and non-obvious feedback metrics. So then the strategy of asking feedback from people feels like your only solid option.
LLM can't do this at all right now. Personally, I don't think that I would respond to an LLM emailing me and asking for feedback. Importance feels like a function of what the human 'community thinks', and being part of a community feels like a place where I still add value. So I'm doubling down
Ironically I used to feel guilty about not doing 'hard work' like math and coding, in favour of chatting ideas with my friends too much. But now it feels like chatting with my friends is all the value I'm able to bring.
The optimist in me thinks that we've come full circle here. There has been a recent meme were social media has made us more and more lonely. As we've industrialised, social critics has claimed that our sense of community has faded. But perhaps when technology becomes so good that our only value is 'deciding our values', perhaps we might see a resurgence in community once again.
Acknowledgements
Thank you to Amelia Gates, Alya Hatta, Shukri Shahizam, Mohlomi Taoana, and Ming Yao for feedback on drafts and ideas.