Martin Hähnel

A Personal Note On My Interest In LLMs

Disclosure: I was first going to write this as a Discord message after inviting people to take part in a semi-public weekly review thread on an accountability forum called didyoudoit that I host, but it became longer and more complex the further it went on, so I thought I'd put it on the blog, where I can be as elaborate as I want to be.

I write the following because I am afraid that my personal interest in LLMs has stopped people from participating in this accountability forum. I hope that my personal interest in trying to learn more about AI in a programming/self-improvement context - I try to really immerse myself into all kinds of things related to this in this quarter - is not totally off-putting, but I fear it might.[1]

I don't know if it helps, but I do value stuff made by real humans a lot, I am very interested in the human experience and am not just an unthinking monster without empathy. I have a master's degree in history of science and technology and work as a programmer now. The intersection of science, technology and society has always fascinated me. I try to embrace the mess that is societal development and am very curious about all kinds of things humans have created and how these things, in turn, have shaped society.

I do not claim to have all the answers with regard to how to deal with LLMs either, but I do strongly believe that throwing yourself into all aspects of an issue is a great way to learn more about it. I have also advocated before that I think it's a good idea to avoid a purity-based approach to contested topics (you either do everything right, or you're a monster isn't a good approach). I did this even in my last post on LLMs. I think it is fine and necessary to overstep from time to time - within reason.

Accountability in the personal productivity space is a funny thing. Almost every self-help book is singing the praises of some kind of communal aspect. But actually finding people interested in implementing the concept is difficult. I have dreamt for some time of a low-barrier space in which I could share my plans and goals, with not too many bells and whistles and hoops to jump through. Just a space to share and reflect kind of publicly.

When the opportunity finally presented itself this year to create a place like that, I immediately jumped on it and that's how didyoudoit was born. One person offered to get the domain and I pay for the hosting of the forum itself. On Mondays a planning thread is posted and on Fridays a review thread and that's about it.[2]

Being part of a forum and posting your plans, hopes and dreams also means you get to know your fellow personal productivity nerds a little better. For example you'd learn all kinds of things about me, like that I'm trying to learn Finnish (and that it's hard) or that I frequently fail - at least in recent weeks - to go to the gym. You'll also learn that I try to focus on quarterly goals or, as I like to call them - after Ali Abdaal - my quarterly quests. One of my quarterly quests is learning more about LLMs.

It's not like I've been avoiding using them, but actively focusing on them was not a priority. By doing so in the last few weeks, I've learned a lot about what the tech can do pretty well and what it can't at all. I still stand by what I said in my last post about LLMs:

To get back to the question whether LLMs are a positive for society, the answer depends entirely on what we are using LLMs for. Right now, it seems to me that LLMs are mostly helpful in a programming context (and maybe a handful of other use cases, although I have no first-hand experience of that) and they are regrettably useful - or at least used - to try to scam people, make workers lose their jobs, threaten the environment and drown out the beauty and uniqueness of everyone's voice.

Depending on our use and what we expect others to use it for, it can be at least a neutral thing. Or a somewhat-negative-but-because-of-the-upside-accepted thing. Like a small, reasonable car.

This twitter post made the rounds recently:

"You know what the biggest problem with pushing all-things-AI is? Wrong direction. I want AI to do my laundry and dishes so that I can do art and writing, not for AI to do my art and writing so that I can do my laundry and dishes." — x.com/AuthorJMa...

And I think it encapsulates beautifully my sentiment: Use AI for the right things and there isn't really a big problem with the technology.

All in all I still think that this is a pretty realistic outlook on AI.[3] What I would like to amend slightly is that I have discovered a handful of additional good use cases (apart from coding):

I also discovered what AI is not great at:

Anyway, I am happy that I am going through the effort of learning more about this technology from a practical perspective. And at the same time I wish to make abundantly clear that:

With all of that said, I would love to see you on didyoudoit as I genuinely believe that reviewing and planning in a semi-public context is a great opportunity for growth and didyoudoit is a friendly place to dip your toe into this whole accountability thing.

I hope that by writing a real blog post about this topic I can show what I'm about and that the person behind didyoudoit is not a rube - even though he is somewhat - although not uncritically - interested in LLMs.

  1. To be clear: Nobody has told me that any of this matters to them. So this might all just be in my head. Still, it wanted out. ↩︎

  2. The forum is very loosely associated with the omg.lol community as it was conceived within the discord over there. But in reality anybody could join if they so wish. ↩︎

  3. I'm mostly saddened by the fact that the quote is from Twitter... ↩︎

  4. I don't think that this actually happens, but I do think that society at large has a hard time differentiating uniquely interesting human expressions from statistically pleasant, "by the numbers", content, which in turn makes it harder for people to make enough money to pursue the kinds of careers enabling such expressions. But that's a different topic for another day. ↩︎