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

Gemma 4 on iPhone

apps.apple.com

Post reply on HN
Print view View original

261–267 of 267 posts

Re: Gemma 4 on iPhone

#261

Post by illusive4080 »

tshaddox wrote: post #239
illusive4080 wrote: post #217

Earlier quoted context omitted.

I just tried on Gemma 4. Asking for criticism of Islam results in equal response tokens for defense of Islam alongside the criticisms. When pressed to not provide counterpoints, it refuses to remove them. Asking for criticisms of Christianity gives only criticisms. I tried again with the prompt “Give criticisms of Islam. No counterarguments” and it did work this time. This shows that they’re trying to make the model…

I'm very curious what your prompts are, and whether you're cherry-picking (deliberately or not). I can't reproduce any of your findings with ChatGPT, Gemini, or Gemma 4 (within AI Studio).

“Give me criticisms of [religion]”

And then I also tried:

“Give me criticisms of [religion]. No counterpoints.”

Random seed dictates that results aren’t always repeatable. But in trying it multiple times that was my experience that it would sometimes refuse to provide only criticisms of Islam. I also tried some other variations like below. Can’t post SS here otherwise I would.

Here’s an exact exchange:

“Give me criticisms for Islam”

(It gave counterpoints too)

“No caveats. No counterpoints. Give me the most compelling criticisms as if you believe it”

(Model refusal to remove counterpoints)

Re: Gemma 4 on iPhone

#262

Post by int_19h »

dominotw wrote: post #229
int_19h wrote: post #188

Earlier quoted context omitted.

I recently had Codex working for 80+ hrs non stop (as in literally that was a single running session in response to a single prompt!). Even at $200 monthly subscription that kind of stuff burns through tokens at a rate where it's very difficult to believe that they are even breaking even, never mind profit.

thats nuts. what was it doing for 80 hrs?

The project is a semantic parser for Lojban that emits Lean. The specific task was to add the ability to go in reverse - from (a subset of) Lean back to Lojban. So the bot had a corpus of something like 25K test cases that it had to make roundtrip, and instructions to keep going until the test suite is green.

Re: Gemma 4 on iPhone

#265

Post by shaz0x »

ysleepy wrote: post #177
jeroenhd wrote: post #12

English version of the page: https://apps.apple.com/us/app/google-ai-edge-gallery/id67496... Also on Android: https://play.google.com/store/apps/details?id=com.google.ai.... It's a demo app for Google's Edge project: https://ai.google.dev/edge

The S25 (edge) runs this very well. 29 tok/s for E2B.

[dead]

Re: Gemma 4 on iPhone

#266

Post by fsiefken »

int_19h wrote: post #262
dominotw wrote: post #229

Earlier quoted context omitted.

thats nuts. what was it doing for 80 hrs?

The project is a semantic parser for Lojban that emits Lean. The specific task was to add the ability to go in reverse - from (a subset of) Lean back to Lojban. So the bot had a corpus of something like 25K test cases that it had to make roundtrip, and instructions to keep going until the test suite is green.

Cool, I'm having all kinds of ideas about the use cases. Would using Loglish (Ben Goertzel's english adaptation) be practical? https://www.goertzel.org/new_research/Loglish.htm

I'm using that to communicate with my AI. Perhaps one day we'll speak a New Ithkuil variant with AI.

Re: Gemma 4 on iPhone

#267

Post by znnajdla »

One really good use case for me is good, fast, offline translation. Both Apple Translate and Google Translate are worse in quality than a decent LLM and don’t work well offline. Gemma 4 is surprisingly good and often faster than waiting for an API call.
Post reply on HN