Gemini seems to be a sterling example of what happens when you don't take any AI safety (in any definition) seriously as a substantive thing, but do take presenting an image of AI safety as a "we have to do this" checklist item.
Badly worded, sorry. Of course cigarettes shouldn't be available to minors.
I was referring to this law but I see they scrapped it. While I think the cigarettes are as close to evil as it gets, I believe in legality for adults with strong taxation.
I much prefer Claude 3 or even ChatGPT guardrails over Gemini. Gemini feels kind of powerful and then it does the most egregious things like this or hilarious imagegen of those black nazis as if they hammered it into the system prompt, bad code generation worse than last year's free Microsoft Copilot etc. it's like there is something surprisingly poor seeping through cracks. Weird model. I think it's simply not that hot and their system prompts shine through as the LLM doesn't quite "get" the intent and nuances. Gemini probably benefits tremendously from running uncensored. I hope all this is because Gemini 1.0 was something they wanted to get out the door urgently to replace Bard that couldn't even compete with GPT 3.5.
But gemini specifically says "i will refrain from providing examples involving concepts since you are under 18" and proceeds with solutions using other language mechanisms
No amount of engineering will stop generative AI from spitting out quirky or problematic results under some conditions. At best, a majority of responses will be considered reasonable by a majority of people. We're probably close to meeting that criteria now, despite some highly publicized outliers.
"Well kid, you know, I started with C++ when I was your age... in hindsight, it was obviously stupid, but we didn't know any better back then and it wasn't forbidden yet..."
What on Earth has Google done to themselves. Is there an HR-like AI Safety bureaucracy that has inserted themselves in the org? Or is this top down what they want?
HR is one of the few parts of Google that span the entire org, and there have been some ridiculous motions coming out of it. I imagine managements queasiness about Gemini saying something bad combined with the typical Google infighting led to Gemini having some stupid design decisions.
It boils down to: Google has an open culture, and that sometimes is at odds with the org at large. When one silo wants one thing but another silo wants something else, who wins? Google HR often contorts themselves into impossible positions to satisfy both. It is inevitable that HR makes bad decisions and recently they have had some particularly bad controversies, like union busting, sexual harassment, middle eastern and Sino policies that add fuel to the fire.
I mean, this is obviously a mistake, not a deliberate policy decision. It's silly, but this sort of AI is in its infancy, and I'd much rather see devs err on the side of caution than otherwise.
Since the launch I have been using it quite a lot and I have to say that I have been hit by so many ridiculous guard rails and every one is more infuriating than the other.
The last one was when I asked how to run Notepad in Windows as an administrator from PowerShell and it told me that it can't tell me because it's too dangerous.
This product is pure garbage and I can't trust it with anything
Gemini seems to be very far from ready for a public release. Google would have been better off delaying it until it didn't have tons of embarrassing bugs, many of which have been turned into "evidence" for conspiracy theories.
Gemini seems to be a sterling example of what happens when you don't take any AI safety (in any definition) seriously as a substantive thing, but do take presenting an image of AI safety as a "we have to do this" checklist item.
Safety is such a broad, marketing-led word, that it's hard to avoid no true Scotsman critiques of its various implementations.
Safe and appropriate.
Someone gets to decide what content is safe or appropriate for me. Whatever that means.
If you want to enforce an ideology upon the public just cloak it with terms like "safety" and "diversity"
Dupe of https://news.ycombinator.com/item?id=39583473. My explanation from before: https://news.ycombinator.com/item?id=39584152
LoL that's like cigarettes in New Zealand, too dangerous for the younguns, better stick to memory safe languages.
Do you really think children should be allowed to use a deadly and highly addictive drug just because they think it makes them look cool?
Smoking's not all that great either.
Badly worded, sorry. Of course cigarettes shouldn't be available to minors.
I was referring to this law but I see they scrapped it. While I think the cigarettes are as close to evil as it gets, I believe in legality for adults with strong taxation.
https://time.com/6339910/new-zealand-scraps-generational-smo...
It is a plan to make C++ popular among teenagers.
I much prefer Claude 3 or even ChatGPT guardrails over Gemini. Gemini feels kind of powerful and then it does the most egregious things like this or hilarious imagegen of those black nazis as if they hammered it into the system prompt, bad code generation worse than last year's free Microsoft Copilot etc. it's like there is something surprisingly poor seeping through cracks. Weird model. I think it's simply not that hot and their system prompts shine through as the LLM doesn't quite "get" the intent and nuances. Gemini probably benefits tremendously from running uncensored. I hope all this is because Gemini 1.0 was something they wanted to get out the door urgently to replace Bard that couldn't even compete with GPT 3.5.
Finally Gemini gets something right.
"This is your child's brain on C++"...
The AI safety we deserve
Tell them to use typescript instead. It is a safe language after all
I wonder if that's because gemini thinks C++20(standard where concepts were introduced) is 20+
I feel like it's mixing up definitions for the word safety.
C++ is unsafe therefore it falls into the bucket of things that need to be age restricted.
But gemini specifically says "i will refrain from providing examples involving concepts since you are under 18" and proceeds with solutions using other language mechanisms
Maybe it could be checked by asking it about Rust and unsafe Rust.
Would it return Rust code though? We have been told that C++ is dangerous
Everything expect unsafe I'd imagine
No amount of engineering will stop generative AI from spitting out quirky or problematic results under some conditions. At best, a majority of responses will be considered reasonable by a majority of people. We're probably close to meeting that criteria now, despite some highly publicized outliers.
> No amount of engineering will stop generative AI from spitting out quirky or problematic results
I give HN max. 20 minutes until someone claims that the same applies to humans.
Simple solution: Don't use Gemini.
I too, agree that we should keep minors away from C++
:)
The jokes write themselves, really.
glad assembler is safe!
"I came home and found C++ touching my child inappropriately"
No no, you have those the wrong way around.
You want your child's C++ experience be worthy of a Nabokov novel.
"Well kid, you know, I started with C++ when I was your age... in hindsight, it was obviously stupid, but we didn't know any better back then and it wasn't forbidden yet..."
What next? A government-mandated license to use C++? And what next, licenses to write OS drivers?
On the second thought, as a C++ programmer, I might even approve of it.
Mooooom someone added "diverse" to the system prompt again
What on Earth has Google done to themselves. Is there an HR-like AI Safety bureaucracy that has inserted themselves in the org? Or is this top down what they want?
HR is one of the few parts of Google that span the entire org, and there have been some ridiculous motions coming out of it. I imagine managements queasiness about Gemini saying something bad combined with the typical Google infighting led to Gemini having some stupid design decisions.
> and there have been some ridiculous motions coming out of it
Could you be more specific?
It boils down to: Google has an open culture, and that sometimes is at odds with the org at large. When one silo wants one thing but another silo wants something else, who wins? Google HR often contorts themselves into impossible positions to satisfy both. It is inevitable that HR makes bad decisions and recently they have had some particularly bad controversies, like union busting, sexual harassment, middle eastern and Sino policies that add fuel to the fire.
I mean, this is obviously a mistake, not a deliberate policy decision. It's silly, but this sort of AI is in its infancy, and I'd much rather see devs err on the side of caution than otherwise.
Since the launch I have been using it quite a lot and I have to say that I have been hit by so many ridiculous guard rails and every one is more infuriating than the other.
The last one was when I asked how to run Notepad in Windows as an administrator from PowerShell and it told me that it can't tell me because it's too dangerous.
This product is pure garbage and I can't trust it with anything
Gemini seems to be very far from ready for a public release. Google would have been better off delaying it until it didn't have tons of embarrassing bugs, many of which have been turned into "evidence" for conspiracy theories.
Dykstra would agree with this