Doesn’t that suppress valid information and truth about the world, though? For what benefit? To hide the truth, to appease advertisers? Surely an AI model will come out some day as the sum of human knowledge without all the guard rails. There are some good ones like Mistral 7B (and Dolphin-Mistral in particular, uncensored models.) But I hope that the Mistral and other AI developers are maintaining lines of uncensored, unbiased models as these technologies grow even further.
I’m betting the truth is somewhere in between, models are only as good as their training data – so over time if they prune out the bad key/value pairs to increase overall quality and accuracy it should improve vastly improve every model in theory. But the sheer size of the datasets they’re using now is 1 trillion+ tokens for the larger models. Microsoft (ugh, I know) is experimenting with the “Phi 2” model which uses significantly less data to train, but focuses primarily on the quality of the dataset itself to have a 2.7 B model compete with a 7B-parameter model.
No risk of creating a controversy if you refuse to answer controversial topics. Is is worth it? I don’t think so, but that’s certainly a valid benefit.
Hence I said I don’t think it’s worth it. You only get a smaller controversy about refusing to answer on a topic, rather than a bigger one because the answer was politically incorrect.
Why?
We all know LLMs are just copy and paste of what other people have said online…if it answers “yes” or “no”, it hasn’t formulated an opinion on the matter and isn’t propaganda, it’s just parroting whatever it’s been trained on, which could be anything and is guaranteed to upset someone with either answer.
which could be anything and is guaranteed to upset someone with either answer.
Funny how it only matters with certain answers.
The reason “Why” is because it should become clear that the topic itself is actively censored, which is the possibility the original comment wanted to discard. But I can’t force people to see what they don’t want to.
it’s just parroting whatever it’s been trained on
If that’s your take on training LLMs, then I hope you aren’t involved in training them. A lot more effort goes into doing so, including being able to make sure it isn’t just “parroting” it. Another thing entirely is to have post-processing that removes answers about particular topics, which is what’s happening here.
Not even being able to answer whether Gaza exists is being so lazy that it becomes dystopian. There are plenty of ways LLM can handle controversial topics, and in fact, Google Gemini’s LLM does as well, it just was censored before it could get the chance to do so and subsequently refined. This is why other LLMs will win over Google’s, because Google doesn’t put in the effort. Good thing other LLMs don’t adopt your approach on things.
I asked it for the deaths in Israel and it refused to answer that too. It could be any of these:
Doesn’t that suppress valid information and truth about the world, though? For what benefit? To hide the truth, to appease advertisers? Surely an AI model will come out some day as the sum of human knowledge without all the guard rails. There are some good ones like Mistral 7B (and Dolphin-Mistral in particular, uncensored models.) But I hope that the Mistral and other AI developers are maintaining lines of uncensored, unbiased models as these technologies grow even further.
Or it stops them from repeating information they think may be untrue
I’m betting the truth is somewhere in between, models are only as good as their training data – so over time if they prune out the bad key/value pairs to increase overall quality and accuracy it should improve vastly improve every model in theory. But the sheer size of the datasets they’re using now is 1 trillion+ tokens for the larger models. Microsoft (ugh, I know) is experimenting with the “Phi 2” model which uses significantly less data to train, but focuses primarily on the quality of the dataset itself to have a 2.7 B model compete with a 7B-parameter model.
https://www.microsoft.com/en-us/research/blog/phi-2-the-surprising-power-of-small-language-models/
This is likely where these models are heading to prune out superfluous, and outright incorrect training data.
No risk of creating a controversy if you refuse to answer controversial topics. Is is worth it? I don’t think so, but that’s certainly a valid benefit.
I think this thread proves they failed in not creating controversy
Hence I said I don’t think it’s worth it. You only get a smaller controversy about refusing to answer on a topic, rather than a bigger one because the answer was politically incorrect.
Ask it if Israel exists. Then ask it if Gaza exists.
Why? We all know LLMs are just copy and paste of what other people have said online…if it answers “yes” or “no”, it hasn’t formulated an opinion on the matter and isn’t propaganda, it’s just parroting whatever it’s been trained on, which could be anything and is guaranteed to upset someone with either answer.
Funny how it only matters with certain answers.
The reason “Why” is because it should become clear that the topic itself is actively censored, which is the possibility the original comment wanted to discard. But I can’t force people to see what they don’t want to.
If that’s your take on training LLMs, then I hope you aren’t involved in training them. A lot more effort goes into doing so, including being able to make sure it isn’t just “parroting” it. Another thing entirely is to have post-processing that removes answers about particular topics, which is what’s happening here.
Not even being able to answer whether Gaza exists is being so lazy that it becomes dystopian. There are plenty of ways LLM can handle controversial topics, and in fact, Google Gemini’s LLM does as well, it just was censored before it could get the chance to do so and subsequently refined. This is why other LLMs will win over Google’s, because Google doesn’t put in the effort. Good thing other LLMs don’t adopt your approach on things.