Live data from Hacker News

Gemini 3 Deep Think drew me a good SVG of a pelican riding a bicycle

simonwillison.net

51–60 of 63 posts

Re: Gemini 3 Deep Think drew me a good SVG of a pelican riding a bicycle

#51
post #31
post #27

Earlier quoted context omitted.

> a great bicycle It's not. Sorry. Go look at some real bicycles for reference.

A better comparison would be the monstrosities generated by older models.

Those were shit.

This one is somewhat OK. Not good, merely passing for a bicycle-shaped object.

Re: Gemini 3 Deep Think drew me a good SVG of a pelican riding a bicycle

#52
post #27

Earlier quoted context omitted.

> a great bicycle It's not. Sorry. Go look at some real bicycles for reference.

What part is missing? It appears to have all the core parts of a bicycle to me?

> What part is missing? It appears to have all the core parts of a bicycle to me?

* I am not an expert in pelican anatomy, but it appears that the bicycle's saddle, if present, would be very deeply inside this pelican (try drawing it onto the frame, and see what you get). Let's be kind, and assume that the seat is simply missing.

* The crown on the fork[1] is missing. The fork is not separate from the frame; so you can't turn the front wheel.

* Neither of pelican's feet are on the pedals (the left foot appears to be missing altogether)

* The wheels are way too small for that frame.

Note: Brakes and gears are nowhere to be seen, but OK, let's just assume it's a fixie (or a single-speed with a coaster brake). Let's also assume the red thing on the basket is there for decoration.

Also, the spokes on a typical bicycle wheel follow a cross-lacing pattern [2], at an angle to the radial direction, and you can see them crossing each other. But OK, let's let that slide.

My point is: sure, this thing at least passes as a bicycle. But it's not a good bicycle drawing.

[1] https://en.wikipedia.org/wiki/Bicycle_fork

[2] https://spokecalc.io/how-to-lace-a-wheel.php

Re: Gemini 3 Deep Think drew me a good SVG of a pelican riding a bicycle

#53
post #27

Earlier quoted context omitted.

> a great bicycle It's not. Sorry. Go look at some real bicycles for reference.

This is a very reasonable drawing of a bicycle. It has a solid rear triangle, and forward swept front fork, which is an important detail for actually being able to steer the bike. The drivetrain is single speed, but that's fine, and the wheels are radially laced, which is also fine: both of those simplified details are things which occur in real bicycles.

>This is a very reasonable drawing of a bicycle

Reasonable? Sure. Great? Absolutely not.

Let's not move the goal posts here.

> and forward swept front fork, which is an important detail for actually being able to steer the bike.

...but no crown or any sort of indication that the fork is separate from the frame, making it impossible to steer. Yay.

> the drivetrain is single speed, but that's fine, and the wheels are radially laced, which is also fine: both of those simplified details are things which occur in real bicycles.

Sure! Now do the bicycle saddle.

Re: Gemini 3 Deep Think drew me a good SVG of a pelican riding a bicycle

#54
post #52

Earlier quoted context omitted.

What part is missing? It appears to have all the core parts of a bicycle to me?

> What part is missing? It appears to have all the core parts of a bicycle to me? * I am not an expert in pelican anatomy, but it appears that the bicycle's saddle, if present, would be very deeply inside this pelican (try drawing it onto the frame, and see what you get). Let's be kind, and assume that the seat is simply missing. * The crown on the fork[1] is missing. The fork is not separate from the frame; so you c…

It's fun when a subject matter expert shows up and blows your mind with their level of knowledge on normally unnoticed intricate details that go into everyday things.

Re: Gemini 3 Deep Think drew me a good SVG of a pelican riding a bicycle

#55
post #29
post #9

Simon notes this benchmark is win-win, since he loves pictures of pelicans riding bicycles — if they spend time benchmaxxing it’s like free pelicans for him. He originally promised to generate a bunch more animals when we got a “good” pelican. This is not a good pelican. This is an OUTSTANDING pelican, a great bicycle, and it even has a little sun ray over the ocean marked out. I’d like to see more animals please Sim…

Agreed, good is quite an understatement. Every item is drawn superbly, and the basket with the fish is just great. Feels like a big jump over the other models (though granted, this is such a known "benchmark" by now, it's likely gamed to some extent).

If we took a random sample of people from around the world, I wonder how long it would take them to be trained to create a better SVG.

Re: Gemini 3 Deep Think drew me a good SVG of a pelican riding a bicycle

#56

SVG generation is a surprisingly good benchmark for spatial reasoning because it forces the model to work in a coordinate system with no visual feedback loop. You have to hold a mental model of what the output looks like while emitting raw path data and transforms. It's closer to how a blind sculptor works than how an image diffusion model works. What I find interesting is that Deep Think's chain-of-thought approach…

Yeah, spatial reasoning has been a weak spot for LLMs. I’m actually building a new code exercise for my company right now where the candidate is allowed to use any AI they want, but it involves spatial reasoning. I ran Opus 4.6 and Codex 5.3 (xhigh) on it and both came back with passable answers, but I was able to double the score doing it by hand.

It’ll be interesting to see what happens if a candidate ever shows up and wants to use Deep Think. Might blow right through my exercise.

Re: Gemini 3 Deep Think drew me a good SVG of a pelican riding a bicycle

#58
post #52

Earlier quoted context omitted.

What part is missing? It appears to have all the core parts of a bicycle to me?

> What part is missing? It appears to have all the core parts of a bicycle to me? * I am not an expert in pelican anatomy, but it appears that the bicycle's saddle, if present, would be very deeply inside this pelican (try drawing it onto the frame, and see what you get). Let's be kind, and assume that the seat is simply missing. * The crown on the fork[1] is missing. The fork is not separate from the frame; so you c…

I just think this is insanely pedantic. Most art is not photorealistic, and in fact imo having a crown on the fork may very well make this image worse by taking away from the aesthetic.

Re: Gemini 3 Deep Think drew me a good SVG of a pelican riding a bicycle

#59
post #6

I don't think this is a good "benchmark" anymore. It's probably on everyone's training set by now.

The point of this benchmark is that making decent SVG art is actually useful. Simon has private image prompts he uses, since he didnt say gemini failed at those it is reasonable to assume those were also successful.

Re: Gemini 3 Deep Think drew me a good SVG of a pelican riding a bicycle

#60
post #52

Earlier quoted context omitted.

> What part is missing? It appears to have all the core parts of a bicycle to me? * I am not an expert in pelican anatomy, but it appears that the bicycle's saddle, if present, would be very deeply inside this pelican (try drawing it onto the frame, and see what you get). Let's be kind, and assume that the seat is simply missing. * The crown on the fork[1] is missing. The fork is not separate from the frame; so you c…

I just think this is insanely pedantic. Most art is not photorealistic, and in fact imo having a crown on the fork may very well make this image worse by taking away from the aesthetic.

Sure, if the "uncanny valley" aesthetic is what you're going for, then by all means, tell the machine to shrink the wheels further.

It's just not what makes a great bicycle picture.

Also, that bicycle's seatpost must be outright penetrating that pelican's cloaca.

You can say I'm being anal about it, and you'll have a point. But it's not that pedantic to notice it.

Post reply on HN