Hacker News Reader

Comment by pk-protect-ai | Hacker News Reader

I can't grasp all the motives of those people preaching "safety training" and "alignment problem." But I suppose it is greed and a will to manipulate the public and effectively the market after all. To decrease biases in LLMs, one should clean up the datasets from these biases. Is it not enough to know that those biases and "dangerous" information in LLMs are simply what is scraped from the internet?

There is no way any LLM can do something dangerous on their own. Even with the huge effort of an evil human mind, they will not be better than a Google search (just a little bit faster).

IMHO, the brainwashing of the LLM after the training aka "safety training", is absolutely useless garbage idea. With the method in the article or without, you can get out of the model whatever you want.

View on HN

The obvious solution to AI safety is already right there: the OpenAI ToS. We currently have a defender from the technodystopia in Sam Altman. By making sure that every bit of text generated by LLMs costs money (and that money goes to OpenAI) he can ensure the safety of the world through his Terms of Service.

Giving one guy or one small group of people vetted by Elizier Yudkowsky complete monopoly over this technology or industry is a small price to pay to ensure that the power to easily generate text does not get too spread out and accessible to the wrong people. By concentrating all of the power over content and revenue from the industry into the hands of Good Guys we make sure that no bad things can happen.

Agreed, the worst outcome here is the little guy getting funny ideas about being able to freely access information. We need to keep it locked up so our betters can decide what the most appropriate use is.

>> We currently have a defender from the technodystopia in Sam Altman

Was it sarcasm? Sam Altman is the most dangerous man on the planet right now because he is manipulating the public with the AI alignment "problem" while simultaneously changing the Open AI "core values" and developing AGI. And let's not forget his "retina" project with the scam coin. Sam Altman wants to be the sole owner of an AGI that will predict whatever he wants.

>> Giving one guy or one small group of people vetted by Elizier Yudkowsky complete monopoly over this technology

Nope. Giving anyone or any group exclusive access or the right of veto over a technology will result in a dystopia. Especially after Elizier's hysterical letter and calls to bomb the data centers. He is biased, and his letter was not rational; it was very emotional and full of fear. This does not make his point of view any more justifiable. So I hope that was a sarcasm too. Edited: separated the answers from original comments.

>Especially after Elizier's hysterical letter and calls to bomb the data centers. He is biased, and his letter was not rational; it was very emotional and full of fear

I would encourage you to peruse this other post from the same very-serious website that we are discussing the content of here

https://www.lesswrong.com/posts/Ndtb22KYBxpBsagpj/eliezer-yu...

> I can't grasp all the motives of those people preaching "safety training" and "alignment problem."

That's simple: most want heavy regulation and AI licenses so there's less competition.

A few others just have big heads from the "baby AGI" hype.

But the only thing any kind of "safety license" will hurt is the AI consumer.

> To decrease biases in LLMs, one should clean up the datasets from these biases.

That's oversimplification of what biases are. You can't clean up biases, they're built into reality and context of the source texts. Everyone is biased to use certain words depending on time, location, history, etc. You want more objective stuff? That's bias. You want data without specific biased things? That's bias too. A neutral dataset does not exist.

That's why I like when "alignment" is used. It's just "how much does the output confirm to what I want out of it" rather than some idea of being uncensored, unbiased, unrestricted, etc.

I agree to some extent with what you say. But you can minimize the impact of biases to the best of your abilities. When I write "biases," I mean any data that increases the amount of falsehood. In this regard, the priority will be to exclude any kind of dataset from the internet crawl or related to political or social life or religions. The problem with datasets starts from the very low level. For example, the Wikimatrix, which is aligned text of English and other languages uses sentences from the scriptures. And this is wrong, especially when the text in the scriptures differs (Catholic vs. Orthodox). Text that should be the same actually has a totally different meaning, which screws the translations. In this regard the synthetic datasets may be more preferable.

>> You want more objective stuff? That's bias.

Yes would like to have models those are biased towards objectivity.

> text of English and other languages uses sentences from the scriptures. And this is wrong,

Wrong in what context? When you're discussing what was said in some Catholic sermon, its content will be based on the local translation, not whatever the oldest written version we can find. (which in itself would be a biased choice) Then which parts to include and which to ignore in the context of religion is another matter of consensus depending on the group you're talking too. Then we've got words which have changed meaning over time and without having an explicit translation, inferring from the original would just result in misunderstandings. I believe you can only choose in what ways you can be wrong in this case. There's no no-context objective version here.

>> Wrong in what context

Read above. I wrote what exactly is wrong with it.

>> There's no no-context objective version here.

Exactly. And that is why such texts must be excluded. https://arxiv.org/abs/2306.11644

I get what you wrote. What I'm saying is that including them is wrong in one context, right in another, and in some you're going to fail regardless which one you include. At that point, wrong is not wrong, but a choice.

>> What I'm saying is that including them is wrong in one context, right in another

There is no "correct context" to include erroneous data that could mess up your weights. You're shifting the discussion into opinions territory, which is exactly this:

https://www.youtube.com/watch?v=Zh3Yz3PiXZw

Reply on news.ycombinator.com