Live data from Hacker News

Gemini 3

blog.google

251–260 of 1001 posts

Re: Gemini 3

#251

It still failed my image identification test ([a photoshopped picture of a dog with 5 legs]...please count the legs) that so far every other model has failed agonizingly, even failing when I tell them they are failing, and they tend to fight back at me. Gemini 3 however, while still failing, at least recognized the 5th leg, but thought the dog was...well endowed. The 5th leg however is clearly a leg, despite being wh…

ah interesting. I wonder if this is a "safety guardrails blindspot" due to the placement.

Op is right.

https://imgcdn.stablediffusionweb.com/2024/4/19/8e54925a-004...

For the above pic I asked "What is wrong with the image?"

Result: - It totally missed the most obvious one - six fingers.

Instead it said this:

Based on a close inspection of the image, here are the things that are "wrong" or anomalous, indicating that it is likely an AI-generated image:

- The Thumb Anatomy: This is the most obvious error. The digit in the thumb's position (far left) looks exactly like a long index finger. A natural human thumb has two segments (phalanges) and is shorter and thicker, opposing the other fingers. This digit appears to have three segments and is roughly the same length and shape as the other fingers, giving the hand the appearance of having five fingers and no thumb.

- Unnatural Skin Texture: The skin is unnaturally smooth, waxy, and free of normal human imperfections like distinct pores, fine wrinkles, or hair. It has a "plastic" or "airbrushed" look typical of generative AI.

- Palm Lines: The creases on the palm are vague and generic. They don't follow the distinct pattern of major flexion creases (like the heart, head, and life lines) found on a real human hand.

- Wrist Connection: The connection between the hand and the wrist looks slightly distorted, with tendons that appear too prominent and smooth for the pose.

[Edit: 3.0 is same as 2.5 - both answered almost identically]

Re: Gemini 3

#252
post #82

Understanding precisely why Gemini 3 isn't front of the pack on SWE Bench is really what I was hoping to understand here. Especially for a blog post targeted at software developers...

Yeah, they mention a benchmark I'm seeing the first time (Terminal-Bench 2.0) and are supposedly leading in, while for some reason SWE Bench is down from Sonnet 4.5. Curious to see some third-party testing of this model. Currently it seems to primarily improve of "general non-coding and visual reasoning" primarily, based on the benchmarks.

They are not even leading in Terminal-Bench... GPT 5.1-codex is better than Gemini 3 Pro

Re: Gemini 3

#253
post #212

I wish I could just pay for the model and self-host on local/rented hardware. I'm incredibly suspicious of companies totally trying to capture us with these tools.

Technically you can! I haven't seen it in the box yet, and pricing is unknown https://cloud.google.com/blog/products/ai-machine-learning/r...

That's interesting. While I suspect the pricing will lean heavily into enterprise sales rather than personal licenses, I personally like the idea buying models that I then own and control. Any steps from companies that make that more possible is great.

Re: Gemini 3

#254
post #138

I truly do not understand what plan to use so I can use this model for longer than ~2 minutes. Using Anthropic or OpenAI's models are incredibly straightforward -- pay us per month, here's the button you press, great. Where do I go for this for these Google models?

Yeah, it truly is an outstandingly bad UX. To use Gemini CLI as a business user like I would Codex or Claude Code, how much and how do I pay?

Re: Gemini 3

#255

I asked it to summarize an article about the Zizians which mentions Yudkowsky SEVEN times. Gemini-3 did not mention him once. Tried it ten times and got zero mention of Yudkowsky, despite him being a central figure in the story. https://xcancel.com/xundecidability/status/19908286970881311... Also, can you guess which pelican SVG was gemini 3 vs 2.5? https://xcancel.com/xundecidability/status/19908113191723213...

Interesting, yeah! Just tried "summarize this story and list the important figures from it" with Gemini 2.5 Pro and 3 and they both listed 10 names each, but without including Yudkowsky.

Asking the follow up "what are ALL the individuals mentioned in the story" results in both models listing ~40 names and both of those lists include Yudkowsky.

Re: Gemini 3

#256

I asked it to summarize an article about the Zizians which mentions Yudkowsky SEVEN times. Gemini-3 did not mention him once. Tried it ten times and got zero mention of Yudkowsky, despite him being a central figure in the story. https://xcancel.com/xundecidability/status/19908286970881311... Also, can you guess which pelican SVG was gemini 3 vs 2.5? https://xcancel.com/xundecidability/status/19908113191723213...

Maybe it has guard rails against such things? That would be my main guess on the Zizian one.

Re: Gemini 3

#257
post #87

I've been so happy to see Google wake up. Many can point to a long history of killed products and soured opinions but you can't deny theyve been the great balancing force (often for good) in the industry. - Gmail vs Outlook - Drive vs Word - Android vs iOS - Worklife balance and high pay vs the low salary grind of before. Theyve done heaps for the industry. Im glad to see signs of life. Particularly in their P/E whic…

For what it's worth, most of those examples are acquisitions. That's not a hit against Google in particular. That's the way all big tech co's grow. But it's not necessarily representative of "innovation."

Re: Gemini 3

#259
post #93

Earlier quoted context omitted.

Ironically, OpenAI was conceived as a way to balance Google's dominance in AI.

I thought it was a workaround to Google's complete disinterest in productizing the AI research it was doing and publishing, rather than a way to balance their dominance in a market which didn't meaningfully exist.

That’s how it turned out, but IIRC at the time of OpenAI’s founding, “AI” was search and RL which Google and deep mind were dominating, and self driving, which Waymo was leading. And OpenAI was conceptualized as a research org to compete. A lot has changed and OpenAI has been good at seeing around those corners.

Re: Gemini 3

#260

I asked it to summarize an article about the Zizians which mentions Yudkowsky SEVEN times. Gemini-3 did not mention him once. Tried it ten times and got zero mention of Yudkowsky, despite him being a central figure in the story. https://xcancel.com/xundecidability/status/19908286970881311... Also, can you guess which pelican SVG was gemini 3 vs 2.5? https://xcancel.com/xundecidability/status/19908113191723213...

He's not a central figure in the narrative, he's a background character. Things he created (MIRI, CFAR, LessWrong) are important to the narrative, the founder isn't. If I had to condense the article, I'd probably cut him out too. Summarization is inherently lossy.
Post reply on HN