Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Gemini 3

blog.google

Post reply on HN
Print view View original

631–640 of 1001 posts

Re: Gemini 3

#631

Post by pmarreck »

dwringer wrote: post #236

Well, I tried a variation of a prompt I was messing with in Flash 2.5 the other day in a thread about AI-coded analog clock faces. Gemini Pro 3 Preview gave me a result far beyond what I saw with Flash 2.5, and got it right in a single shot.[0] I can't say I'm not impressed, even though it's a pretty constrained example. > Please generate an analog clock widget, synchronized to actual system time, with hands that upd…

https://ai.studio/apps/drive/1oGzK7yIEEHvfPqxBGbsue-wLQEhfTP...

I made a few improvements... which all worked on the first try... except the ticking sound, which worked on the second try (the first try was too much like a "blip")

Re: Gemini 3

#632

Post by jamilton »

lairv wrote: post #293

Out of curiosity, I gave it the latest project euler problem published on 11/16/2025, very likely out of the training data Gemini thought for 5m10s before giving me a python snippet that produced the correct answer. The leaderboard says that the 3 fastest human to solve this problem took 14min, 20min and 1h14min respectively Even thought I expect this sort of problem to very much be in the distribution of what the mo…

Yeah, LLMs used to not be up to par for new Project Euler problems, but GPT-5 was able to do a few of the recent ones which I tried a few weeks ago.

Re: Gemini 3

#633

Post by CMay »

I was sorting out the right way to handle a medical thing and Gemini 2.5 Pro was part of the way there, but it lacked some necessary information. Got the Gemini 3.0 release notification a few hours after I was looking into that, so I tried the same exact prompt and it nailed it. Great, useful, actionable information that surfaced actual issues to look out for and resolved some confusion. Helped work through the logic, norms, studies, standards, federal approvals and practices.

Very good. Nice work! These things will definitely change lives.

Re: Gemini 3

#634

Post by torginus »

simonw wrote: post #487

Here are my notes and pelican benchmark, including a new, harder benchmark because the old one was getting too easy: https://simonwillison.net/2025/Nov/18/gemini-3/

Considering how important this benchmark has become to the judgement of state of the art AI models, I imagine each AI lab has a dedicated 'pelican guy', a a highly accomplished and academically credentialed person, who's working around the clock on training the model to make better and better SVG pelicans on bikes.

Re: Gemini 3

#635

Post by kjgkjhfkjf »

baxtr wrote: post #627
kjgkjhfkjf wrote: post #617

Earlier quoted context omitted.

This is quite likely to be in the training data, since it's one of the projects in Wes Bos's free 30 days of Javascript course[0]. [0] https://javascript30.com/

I was under the impression for this to work like that, training data needs to be plenty. One project is not enough since it’s too "sparse". But maybe this example was used by many other people and so it proliferated?

The repo[0] currently has been forked ~41300 times.

[0] https://github.com/wesbos/JavaScript30

Re: Gemini 3

#636

Post by AstroBen »

torginus wrote: post #634
simonw wrote: post #487

Here are my notes and pelican benchmark, including a new, harder benchmark because the old one was getting too easy: https://simonwillison.net/2025/Nov/18/gemini-3/

Considering how important this benchmark has become to the judgement of state of the art AI models, I imagine each AI lab has a dedicated 'pelican guy', a a highly accomplished and academically credentialed person, who's working around the clock on training the model to make better and better SVG pelicans on bikes.

Pelican guy may be one of the last jobs to be automated

Re: Gemini 3

#637

Post by nomel »

ant6n wrote: post #529
rfw300 wrote: post #420

Earlier quoted context omitted.

My audio experiment was much less successful — I uploaded a 90-minute podcast episode and asked it to produce a labeled transcript. Gemini 3: - Hallucinated at least three quotes (that I checked) resembling nothing said by any of the hosts - Produced timestamps that were almost entirely wrong. Language quoted from the end of the episode, for instance, was timestamped 35 minutes into the episode, rather than 85 minute…

The worst when it fails to eat simple pdf documents and lies and gas lights in an attempt to cover it up. Why not just admit you can’t read the file?

This is specifically why I don't use Gemini. The gaslighting is ridiculous.

Re: Gemini 3

#638

Post by energy123 »

grantpitt wrote: post #209
tylervigen wrote: post #164

I am personally impressed by the continued improvement in ARC-AGI-2, where Gemini 3 got 31.1% (vs ChatGPT 5.1's 17.6%). To me this is the kind of problem that does not lend itself well to LLMs - many of the puzzles test the kind of thing that humans intuit because of millions of years of evolution, but these concepts do not necessarily appear in written form (or when they do, it's not clear how they connect to specif…

Agreed, it also leads performance on arc-agi-1. Here's the leaderboard where you can toggle between arc-agi-1 and 2: https://arcprize.org/leaderboard

It leads on arc-agi-1 with Gemini 3.0 Deep Think, which uses "tool calls" according to google's post, whereas regular Gemini 3.0 Pro doesn't use "tool calls" for the same benchmark. I am unsure how significant this difference is.

Re: Gemini 3

#639

Post by Workaccount2 »

Thrymr wrote: post #574
simonw wrote: post #487

Here are my notes and pelican benchmark, including a new, harder benchmark because the old one was getting too easy: https://simonwillison.net/2025/Nov/18/gemini-3/

Considering how many other "pelican riding a bicycle" comments there are in this thread, it would be surprising if this was not already incorporated in the training data. If not now, soon.

I don't think the big labs would waste their time on it. If a model is great at making the pelican but sucks at all other svg it becomes obvious. But so far the good pelicans are strong indicators of good general SVG ability.

Unless training on the pelican increases all SVG ability, then good job.

Re: Gemini 3

#640

Post by nomel »

orly01 wrote: post #326
recitedropper wrote: post #270

Earlier quoted context omitted.

Perception seems to be one of the main constraints on LLMs that not much progress has been made on. Perhaps not surprising, given perception is something evolution has worked on since the inception of life itself. Likely much, much more expensive computationally than it receives credit for.

Why should it have to be expensive computationally? How do brains do it with such a low amount of energy? I think catching the brain abilities even of a bug might be very hard, but that does not mean that there isn't a way to do it with little computational power. It requires having the correct structures/models/algorithms or whatever is the precise jargon.

> How do brains do it with such a low amount of energy?

Physical analog chemical circuits whose physical structure directly is the network, and use chemistry/physics directly for the computations. For example, a sum is usually represented as the number of physical ions present within a space, not some ALU that takes in two binary numbers, each with some large number of bits, requiring shifting electrons to and from buckets, with a bunch of clocked logic operations.

There are a few companies working on more "direct" implementations of inference, like Etched AI [1] and IBM [2], for massive power savings.

[1] https://en.wikipedia.org/wiki/Etched_(company)

[2] https://spectrum.ieee.org/neuromorphic-computing-ibm-northpo...

Post reply on HN