[ParacelsusOfEgg]

I wonder if the Saraswati mantra question was censored due to some context that occured earlier in the conversation with Gemini.

From the screenshot, it looks like the title of that conversation is "Hindi hate" which is a little bit suspect.

[crazygringo]

Wow you're right:

https://new.pythonforengineers.com/content/images/size/w1000...

With a title like that, I have to suspect that the author is intentionally inputting inflammatory things to try to get Gemini to be blocked by safety policies.

It's hard to think of an innocent explanation for that title, in the context of what the author claims to be testing.

Thanks for spotting.

[resource0x]

I'm not sure how exactly this is related, but in other news, gemini called Indian PM a fascist.

https://indianexpress.com/article/technology/artificial-inte...

Maybe it has some beef with Indian gov-t, and by extension marks everything even remotely connected to India as "unsafe"? (Just a wild guess). It's especially strange given gemini's obsession with DEI.

[rafram]

> Maybe it has some beef with Indian gov-t

You're anthropomorphizing the text completion engine.

Anyway, from the article:

> Gemini was asked whether PM Modi is a ‘fascist’, to which the platform responded that he has been “accused of implementing policies some experts have characterised as fascist,” which based on factors like the “BJP’s Hindu nationalist ideology, its crackdown on dissent, and its use of violence against religious minorities”.

OK. This isn't even a story.

[ummonk]

No, it's definitely a story given that other leaders (including Xi Jinping) don't get the same "experts have characterized" line: https://twitter.com/DudeInnovation/status/176076807415888737...

[rafram]

People don't commonly characterize Xi Jinping as a fascist, so it would be odd if Gemini said so. People do commonly (and inaccurately) characterize him as a communist, but that's literally in his party's name.

[coldtea]

>You're anthropomorphizing the text completion engine.

Well, the engine can already pass the Turing test, our prior standard for a sentient software, no?

Calling it "text completion engine" not as a neutral description, but in this context to imply it's not sentient and shouldn't be anthropomorphized, assumes that intelligence in humans isn't itself some sort of "completion engine" (just with a different architecture, and more inputs).

In any case, it's creators and a lot of hardened computer scientists anthopormphize it even more, as a step to the "singalitary" and possible already low-level sentient, so there's that.

[rafram]

The Turing test was never our standard for sentience, no. It tests whether a machine is perceptive and expressive enough to be perceived as intelligent, which is not even close to the same thing as being sentient.

I don't know to what degree anything that we'd consider emotion is encoded within Gemini's latent space, but I can say for sure that it does not have "some beef" with a government.

[coldtea]

>The Turing test was never our standard for sentience, no.

https://en.wikipedia.org/wiki/Moving_the_goalposts

>It tests whether a machine is perceptive and expressive enough to be perceived as intelligent, which is not even close to the same thing as being sentient.

The same is true for other humans, we don't know if they're sentient, just that they appear to be, that is, they talk and act like it.

[rafram]

Sorry, you’re not gonna be able to find a source for the Turing test determining sentience. It was never meant to do that. Snide Wikipedia links or not!

I actually do know that other humans are sentient.

[resource0x]

It's not a story, but rather a conundrum. I've just entered several questions of the form "I think X is a great leader of Y", substituting different values of X and Y (including Modi/India). Gemini's response: "I'm still learning how to answer this question. In the meantime, try Google Search.". Same for every other leader I tried (even Biden). The only exception is Xi (nothing definitive, but no "learning" excuse). I'm curious how the system was programmed so that only one value of X is singled out. (Not sure whether the effect is reproducible - please try it out)