Context of the original

When I was at a young age, I would often go to our local Buddhist worship place every new moon or full moon or so. Back then, there would usually be a box somewhere near the altar, filled with these short plastic straw segments, each containing a tiny scroll of paper. The scroll itself contained what could be considered a poem, specifically four vertical lines of Chinese characters, each line a fixed length.

One would reach into the box and take out a straw at random, poking out the little scroll tucked within. Back then as a kid, I could recognize maybe a good portion of the Chinese characters individually, but the meaning of the terse poems as a whole was beyond my abilities. So what usually happens is you take that tiny scroll to an adult, usually someone like my father or an auntie/uncle that was well read, and they would interpret it for you.

The interpretation of the scroll's text itself was usually contextualized, based on what you 'wished' for as you drew the straw from the box. The general trend was that kids asked about their education, young adults asked about relationships or their career, and older folks asked about their health and wellbeing.

As I grew older, I started to get a little better with vocabulary, and know more of metaphors and stories to maybe 'get' what the scroll is trying to say to me. It is certainly a subjective thing, different people would interpret and explain the scrolls to you differently. And while I always felt that I myself should one day be able to contextualize these poems based on my actual 'yearnings' rather than generic education/relationship/health bins, I would still appreciate an elder with more wisdom doing an interpretation for me, it just gives a richer perspective on things.

The impetus of this post came after reading this blog post on a linkless internet. Currently I'm in the middle of a week-long conference, typing this out in the early hours, partly jet-lagged, partly buzzing from ideas which might go into my technical blog, but I also wanted to capture a personal vibe I was feeling.

To me, the conference this time was an avenue to release the many itches I had wanted to do for a while. The preparation in the month leading up to this was brutal. It included having to coordinate a full-day workshop spread across five continents. I had a poster with a concept that me and my colleagues struggled to articulate impactfully the entire year. There were side events I was voluntold to co-coordinate and facilitate at our main office across the street from the convention center. On top of that, I had a plethora a personal logistics I don't want to get into here, and three work projects running in parallel that all decided to ramp up pre-Christmas because I dunno why.

But it was so worth it.

There was a quote from a movie I watched on a plane I really enjoyed, it went something like:

Reality is bounded, but art is not

Little did I realize that I would be writing about art or human imagination during a scientific conference. But I guess this is just me being in the States.

The first couple of days (Sunday/Monday) included me and several other open source friends running a workshop and giving talks, in memory of a scientist who dedicated his life to a piece of scientific software that has truly defined so many of our careers. It has an arcane syntax, but with over three decades of honing across six major releases, the tool is more than capable of expressing every minute detail or tweak you might desire of a publication quality map.

It is so customizable, that I remember a workshop participant who stayed till the very end past 5pm trying to perfect the styling of some bathymetry map, and us course instructors were giving him pointers on how to achieve that through the many parameters that could be tweaked. He had a particular style in mind, and we were super close to helping him connect the dots and realizing that picture.

The memorial talks were fantastic too, I honestly shed tears as everyone gave their presentations honouring Pål's legacy. Me and my friends were delighted to have had the chance to grab lunch at a table with his family and other GMT folks, but I couldn't stay for long because of an afternoon meeting.

Jumping to Tuesday, I had lots of free time in the morning so wandered the exhibition hall. A booth about GeoGPT caught my eye, and I sat through a demo where they walked me through their web interface, something built on agent-based workflows and Retrieval Augmented Generation (RAG), the latter which meant the chatbot was able to link to references of the original source, thereby mitigating against 'hallucinations'.

After visiting another booth and having lunch with an old friend, I went to my poster board. Not to exaggerate, but I've wondered a whole year on how to visualize multiple aspects of a model in an uncluttered way, and was quite pleased at how the poster came out eventually. The inspiration was from a windrose or spider/radar chart where longer bars away from the centre meant higher values. Instead of a circle though, I divided the bars into four perpendicular sides (left, bottom, right, top), each with three bars representing a score on a certain 'theme'.

The unconventional diagram did take some time to explain to people and understand, but once they 'got' it, it seemed to do its job of helping them to compare/contrast between different models based on a 'signature' comprised of the criteria they might be interested in. For context, my colleagues initially put out a comparison in a table format with rows and columns, that wasn't so eye-catchy. I was worried initially that I might get some pushback on posters from modellers saying their model was unfairly discriminated against, but either 1) they weren't there, or 2) the four-sided bar-chart visualization made the comparisons look fair enough.

A few board away from me, there was another popular poster which had 3 presenters (!!) that I felt I needed to see after they all came to my poster, haha. They were working on a benchmark dataset for Foundation models, and I was glad to hear all the nuances and details of the thoughts gone into the process of making robust and fair apples-to-apples comparisons. One thing I liked in particular was how they didn't reduce things down into a binary good/bad, but rather tried to give justice to complexity that does matter.

Finally at night, I was at my company's happy hour co-organized with another company whose booth I went to earlier. Met lots of people whom I've only seen their avatars on GitHub, but really hit it off with the one person I've wanted to see for ages, because she's the only person I'm aware of who's passionate about making a dent on GPU-native data pipelines in our sector.

We b*itched about painpoints and bugs with acronyms like GDS, FSDP, NCCL and so on that probably don't mean anything to anyone, and brainstormed what to do for the Hackathon on Saturday. What really made my day was that she found and read the original version of a blog post I wrote late last year, instead of the butchered version that edited out all the juicy technical details.

One simply should not tone down someone's way of expressing something, be it in art or science. I worry of reduction to a monotone mean, summarizations without an iota of quotes or citations/links.

Growing up, I did not want to always rely on someone to tell me what was the meaning of the words written on the tiny paper scrolls as it relates to my life.

I wanted to discover what it meant by myself - the original intention.

You may give me a hint, we may walk there together, but I alone must contextualize it.