> but I found that LLMs do a pretty good job of generating Typst code.
Interestingly, I've had the opposite experience. ChatGPT and Claude repeatedly gave me errors, apologized profusely, and then said, "ah, I had the wrong keyword. It's actually <blahblah>"--and that would simply give me another error and a subsequent apology.
At least Gemini had the good taste of telling me that it didn't know how to do what I wanted with typst.
It's certainly possible that I was trying to do something a little too unusual (who knows), but I chalked it up to the LLMs not having a large enough corpus of training text.
On the bright side, the typst documentation is quite good and it was just a matter of adjusting example code that got me on track.
Well, that just goes to show that these tools are wildly unpredictable. I've had bad experiences generating Go, whereas I've read many experiences of the opposite.
> I chalked it up to the LLMs not having a large enough corpus of training text.
I'm inclined to believe the opposite, actually. It's not so much about the size of the training data, but the quality of it. Garbage in, garbage out. Typst is still very young, and there's not much bad code in the wild.
And the language itself plays a large role. A simple language with less esoteric syntax and features should be easier to train on and generate than something more complex. This is why I think LLMs are notoriously bad at generating Rust code. There's plenty of it to train on, but Rust is a deep pit of complexity and unusual syntax. Though it helps when the language is strict and statically typed, so that the compiler can catch issues early. I would dread relying on generated Python code, despite of how popular and simple it is on the surface.
Specifically, the Far Side panel plays on the idea of the fact that cows would have a cow-centric view of the world and would likely develop tools that were alien to us. The other part of the 'joke' is that cows don't build tools (afaik). edit: I think the "3rd" part of the joke is that the tools look like shit, which is what you'd expect from even the most talented cows.
The humor of the space alien joke is similar in that it's pointing out the difficulty that everyone has in understanding how others (other people, other species, etc.) view and describe the world.
Human trials may eventually be allowed, but I suspect that the results are preliminary and much work still needs to be done to assess the drug's safety before it makes it to humans.
Let me tell you a story about a biotech startup that I worked at many moons ago.
- We were trying to develop new antibiotics to treat certain bacterial infections.
- We had created a new antibiotic that was chemically similar to an existing, commercially successful antibiotic.
- But our drug was ~2-4× more potent at killing certain pathogenic bacteria than the existing med.
- Sounds good for us, right?
- Well, the commercially successful antibiotic was also toxic to humans if administered for "long" periods (it causes severe anemia if administered for 20+ days [I may not remember the precise details here]).
- Therefore, we were concerned that if our drug was 2-4× more potent at killing the microbes, it might also be 2-4× more toxic to people.
- To obtain approval for human tests, we had to run toxicity tests of our drug using several non-human species. Those results were mixed (toxic in some species, non-toxic in others), but we did eventually get approval.
- Unfortunately, the original concerns were correct: our drug caused severe anemia within ~3 days (again, specifics may be wrong), which means that particular candidate died in Phase 1 (initial human trial assessing drug safety).
Thankfully, the severe anemia was reversible in our test subjects (stop taking the drug, and the anemia went away)
You know... my wife and I have repeatedly asked my mother-in-law (who lives with us) to not purchase junk food because--even though she has no problem eating it in moderation--we have a hard time doing so if it's just lying around the house.
The author does make a compelling argument for using a telephoto to compress planes--the shot of the people on the bench with the mountains in the back gives a good example (even though the bench and rock-wall are tilted :-( ).
Things like: "Hey I'm organizing a trip to Vegas. $1000 / head. Great hotel, meals paid, etc. etc."
Then the organizer has the great misfortune of being "robbed" of all the money he collected by a masked assailant.
Maybe the higher level guys were were brighter, but I kind of doubt it.
reply