Reid Orsten

I initially planned to start this blog with a “Hello World” post, introducing myself and what I intend to do with the site. It seemed necessary because there is a split in my audience: coming from instagram, where my account has primarily shown my art; and coming from linkedin, where my focus is on my professional life. Beyond the split, I hope to write about topics that aren’t directly relevant to either. It’s a personal site, not a persona site.

I confess that I started lazily:

Hi chatgpt, I’m having trouble coming up with ideas for the image to accompany a “hello world” post for my new personal website. I’m a software/infrastructure engineer, an artist and an avid philosophy reader; I plan to express all of these on my site. Can you give me some suggestions?

It turns out that the answer is basically no:

That sounds like an exciting website!¹ Here are a few ideas for the image to accompany your “hello world” post:

1. Abstract Tech-Art Fusion: An abstract design that incorporates elements of code, digital infrastructure (like network nodes or server racks), and artistic splashes of color and shapes.
2. Philosophical Coding Scene: An image of a person (perhaps you) sitting at a desk with a computer, surrounded by books on philosophy and sketchbooks. The background could feature code snippets, philosophical quotes, and artistic doodles.
3. Digital Canvas: A digital canvas with a blend of binary code transforming into artistic brush strokes, symbolizing the merge of technology and art.

It gave me more, but you see the theme: montages of cliches. And you know what? This is also exactly what I would expect from an untrained human in a business environment, trying to cram everything I said into one image without any process or rigour.

As I’m writing this, there is a mounting backlash against AI hype, which I believe is unfair. The people who are claiming that all of this LLM stuff will come to nothing are missing some very important points:

1. Before LLMs, processing text involved breaking down strings of letters into clauses and words, trying to use grammar rules to build a structure to represent the meaning of the sentence, and trying to reason about that structure. It sucked! Sentences ending in prepositions might not parse, it was clueless about idiomatic expressions, and was beyond hopeless at slang, misspellings, ambiguous grammar, missing punctuation: e.g. most people’s writing. Being able to work coherently with written text at this level is an advancement we are only starting to comprehend.
2. LLMs represent one kind of intelligence, and our current mode of just adding parameters is unlikely to generalize well. Google recently introduced AlphaGeometry, a LLM that can solve math problems at an honours high school level. Importantly, it has two minds: a neural language model and a rule-based deduction engine. Expecting LLMs to solve everything is like expecting an English major to perform surgery while singing an opera. The human brain has dozens of specialized centres, and expert humans have trained them to work together. We’re only starting to dig into what actual intelligence means.
3. The real problem that businesses are facing building AI-driven applications is “executive dysfunction.”² Solving complex problems for both humans and organizations involves defining the problem well, identifying what needs to be known and how to measure it, how to reason with the measurements, what actions to take with that reasoning, how to evaluate the results, how to recover from mistakes. Throwing LLMs at half-baked problems with bad data is the organizational equivalent of buying the most expensive camera to become a better photographer.

I expect to have a lot more to say about all of this in future posts.

---

There was one of ChatGPT’s suggestions that triggered a thought, though:

Modern Thinker: A modern twist on Rodin’s “The Thinker” statue, where the thinker is pondering over a laptop, surrounded by symbols of technology and art.

I already painted that, back in 2013.

At the time, I had just finished a research paper talking about accreted history and culture in chess and go, and the research was not easy. Google found relevant results for go or weiqi, but not much scholarship. It was useless for chess: SEO made it nearly impossible to get anything but tutorial sites. Academic search engines just returned every article with the word “go”, and there are several academics named Wei Qi. With libraries all over the world searchable in an instant, I longed for a card catalogue, where I could more easily search by subject.

And thus this painting: the muscular male figure, the classic humanist symbol of strength and rationality³, hunched to fit into the frame, in darkness except for the light of the laptop awkwardly balanced on his knee, struggling to make sense of what’s on the screen. I still like how it captures a tension between power and restriction that has become so familiar to us in the digital age.

1. What an ass kisser! ↩︎
2. Companies can have ADHD, it turns out. There’s no medication for it. ↩︎
3. This also references William Blake’s “Newton”, which has its own critique of the symbol. ↩︎