News for hackers. Stuff that matters.
Re:A misleading open letter about sci-fi AI dangers ignores ... by boredumb on Thursday March 30, 2023 @11:31AM (#35374299)
I truly believe one of the biggest dangers of AI in our not so distant future will be the reaction to a lot of the LLM spam introducing an internet realID issued by governments to keep track of Real People and the internet will become divided into a walled/non-anonymous garden and a hellscape of GPT content on the other side. I hope i'm wrong but it's one of those roads to hell that can get paved with good intentions.
Re:A misleading open letter about sci-fi AI dangers ignores ... by holoduke on Thursday March 30, 2023 @11:35AM (#35374388)
I think AI will slowely encapsulate us individuals. Just like people wearing glasses to see better we integrate AI to be able to perform or even cope with society. Slowely more and more skills will be aided or replaced by AI. Till in the end we only have some brains left to guide the AI in everything we do in life.
Re:A misleading open letter about sci-fi AI dangers ignores ... by SmoothBrain12 on Thursday March 30, 2023 @11:35AM (#35374381)
<Click here if you're over 18>
Re:A misleading open letter about sci-fi AI dangers ignores ... by paulmd on Thursday March 30, 2023 @03:42PM (#35377856)
[are you a real human being?]

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

Re:A misleading open letter about sci-fi AI dangers ignores ... by randcraw on Thursday March 30, 2023 @11:47AM (#35374583)
Yeah, I think a primary challenge with LLMs is to somehow identify them. If we can't, the net will become swamped by fake humans bent on distorting human reality in whatever ways serve their masters' whims.

Frankly, I think identification of LLM personas vs humans will be impossible. It's increasingly likely then that the rise of LLMs may introduce the decline of the internet as a way to connect with other humans. Soon nobody will want to take an internet entity seriously unless it can prove it's human. And I don't see any way to do that.

So maybe LLM fake entities will be the straw that breaks the Net's back as a social medium and propels us back into the physical world? I hope so.

Re:A misleading open letter about sci-fi AI dangers ignores ... by Loughla on Thursday March 30, 2023 @03:18PM (#35377627)
It is true that the rise of language models has the potential to introduce fake personas that could manipulate and distort human reality. However, I believe that identifying LLM personas versus humans is not impossible. There are techniques that can be used to detect synthetic content, such as detecting patterns in the language used or analyzing the response time. Additionally, researchers and companies that have been discussed on HN itself that are already working on developing tools to combat deepfakes and synthetic content.

While it's important to remain vigilant about the potential negative impact of LLMs, it's also important to remember the positive ways in which they can be used. Language models have the potential to revolutionize industries such as healthcare and education by analyzing vast amounts of data and providing insights that can improve people's lives.

Furthermore, I don't think that the rise of LLMs will necessarily lead to the decline of the internet as a social medium. Rather, it will likely lead to a greater emphasis on trust and transparency in online interactions. We may see the development of new technologies that enable people to verify the authenticity of online identities, or the rise of platforms that prioritize human-to-human interaction.

Overall, I believe that while there are challenges associated with the rise of LLMs, there are also opportunities for positive change. It's important to continue to monitor the development of these technologies and take proactive steps to ensure that they are used in a responsible and ethical manner.

Re:A misleading open letter about sci-fi AI dangers ignores ... by meghan_rain on Thursday March 30, 2023 @06:03PM (#35379396)
@dang, should boring, ChatGPT-generated comments likes these be banned?
Re:A misleading open letter about sci-fi AI dangers ignores ... by Loughla on Thursday March 30, 2023 @08:14PM (#35380647)
That's fair (but please don't ban me for that @dang). I genuinely just wanted to see if anyone would figure out it was chatgpt generated. What tipped you off?

Because I believe it's going to get harder and harder to figure out what's human and what's not.

Re:A misleading open letter about sci-fi AI dangers ignores ... by meghan_rain on Thursday March 30, 2023 @11:00PM (#35381951)
Sorry my comment about banning was a bit harsh.

Re what tipped me off: gonna stay security-through-obscurity on that for now :)

Re:A misleading open letter about sci-fi AI dangers ignores ... by creamyhorror on Friday March 31, 2023 @12:10AM (#35382447)
It's very typical of ChatGPT, the neutral, regular phrasing and the general, anodyne statements. ChatGPT's averaging of words produces very flat writing.
Re:A misleading open letter about sci-fi AI dangers ignores ... by ryanwaggoner on Friday March 31, 2023 @02:13AM (#35383293)
Yeah, I figured it out too. It sounds too formal in its structure, like a high school paper.
Re:A misleading open letter about sci-fi AI dangers ignores ... by Borrible on Friday March 31, 2023 @05:24AM (#35384532)
No, the op just forgot to give Chatty some context. Human enhanced AI if you will.

Tell it to talk like a typical HN'er, adopt a more robust and aggressive stance, and include five grammatical and spelling errors.

Re:A misleading open letter about sci-fi AI dangers ignores ... by modeless on Thursday March 30, 2023 @11:47AM (#35374600)
Yes, this is a good realistic danger too. Is anyone seriously working on ways to authenticate human-generated content online that don't destroy all possibility of anonymity? CAPTCHAs will soon be impossible. Centralized ID services are unpalatable. Maybe some kind of web of trust would be the way to go.
Re:A misleading open letter about sci-fi AI dangers ignores ... by tzs on Thursday March 30, 2023 @05:45PM (#35379204)
There are three properties I think we probably want.

1. I can prove to a site that I am a human without the site getting any information from the proof about which specific human I am.

2. I can visit the site multiple times and prove to them that I'm the same human, without them getting any information from the proofs about which specific human I am.

3. Any third party site that is involved in the proof does not get any information about which site I trying to prove my identity to.

One way this could be done is by having ID sites that you are willing to prove your real identity to in a robust and way which certify your humanity to other sites in a way that has the three aforementioned properties.

Good candidates for running ID sites would be entities like major banks that most of us have already verified our real identity to in a robust way.

One way to do this might be something like this. When a site wants to check that I'm human they give me some random data, and I have an ID site that I have an account with such as my major bank do a blind signature of that data, using a blind signature system the allows the blinder to remove the blinding.

When I get back the signed blinded data, I remove the blinding and then send the data and unblinded signature back to the site I'm trying to prove my humanity to.

They see that the data is signed by my major bank's ID service, so know that the bank believes I'm human.

Re:A misleading open letter about sci-fi AI dangers ignores ... by avgcorrection on Thursday March 30, 2023 @12:58PM (#35375759)
Why government-issued? Could just as well become a de-facto corporate monopoly by Google or someone else. Is Gmail the “standard” email or is it a government-backed one?

This problem has been coined “Digital Feudalism”. And medieval feudalism (or at least our popular conception of it) wasn’t driven by governments.