Ask HN: How will the content on Hacker News change with AI tooling?

8 points by mrccc ↗ HN
Hello HN, I recently tried out ChatGPT, and I am impressed. Most of the answers it gives me are just plain good. This is the first time that I realize how good AI has become already, and I am a bit scared of the changes it might bring.

If we extrapolate the capabilities of ChatGPT and other AIs into the future, the bar for "good content" will probably raise. At the same time, some of the originality that we see here might be lost. To find answers, we won't need to use Google Search and Stackoverflow anymore. Therefore we won't discover new pages that used to solve these problems, like personal blogs. The _need_ to write a personal blog might vanish completely. After all, nobody will ever find you and read you in the first place.

Which leaves me at this: What content remains? And what would you still like to read after AI content took over the Internet? Have you already noticed a change?

14 comments

[ 3.4 ms ] story [ 44.9 ms ] thread
I asked ChatGPT. It replied:

It is true that the advancements in artificial intelligence, such as ChatGPT, have significantly improved the quality and the accuracy of the generated content. However, this does not necessarily mean that all the originality and the creativity of human content will be lost. In fact, AI can also be used as a tool to enhance human content, by providing suggestions, recommendations, or inspiration.

Moreover, the need for personal blogs and other forms of human content will not vanish completely, because there are many aspects of human experience, such as emotions, values, or opinions, that cannot be fully captured or replicated by AI. Therefore, human content will still have its own unique value, and it will continue to be sought after, as long as it provides something that AI cannot, such as a personal perspective, a storytelling style, or a human connection.

In terms of the content that will remain, it will probably be a mix of both human and AI content, depending on the context and the purpose. For example, for factual or technical information, AI content might be more reliable and efficient, but for personal or creative content, human content might be more engaging and relatable. Therefore, the content that you would still like to read after AI content took over the Internet will probably depend on your interests, your preferences, and your values.

In a way, ChatGPT predicts that the content will shift towards AI content on Hacker News – and maybe some philosophy.

I think it is currently true that emotions, values, and opinions can't be fully captured by AI – but even this will change at some point. I don't see why it shouldn't be possible to at least give the appearance that it "feels".

Based on this answer, I am also curious where this leaves sites like Wikipedia. Will it still be necessary?

I think it will be a long time before it really understands what it is saying. I just got it to say that pineapples are not a food source and therefore inappropriate to sell to restaurants.
To be fair, it's not that hard to convince humans to believe ridiculous things too. There are plenty of examples out there.
I did too. Answer:

It's difficult to say exactly how AI tooling will change the content on Hacker News, as it will likely depend on how the tooling is used and what specific features it has. However, it's possible that AI tooling could be used to help moderate the site and identify spam or inappropriate content, or it could be used to provide users with personalized recommendations for articles and discussions based on their interests. It could also potentially be used to assist with tasks such as summarizing long articles or automatically generating discussion prompts based on the content of an article. Ultimately, the impact of AI tooling on the content of Hacker News will depend on how it is integrated and used by the site's administrators and users.

I followed up with "Can you provide some concrete examples?". Answer:

1. AI could be used to recommend content to users based on what they have previously upvoted, as well as recent trends among other users.

2. AI could be used to detect malicious content, such as spam, phishing links, or other malicious content.

3. AI-powered moderation tools could be used to detect and moderate conversations more quickly and effectively. This could be used to identify language which is offensive or inappropriate, and automatically remove it from the platform.

4. AI could be used to detect when conversations are getting heated, and intervene to prevent them from escalating into an argument.

IMO it is time for private communities with restricted registrations to preserve post quality. I don't know what to do about HN though, but I'm sure they'll come up with something. Maybe every account will need to be re-verified somehow.
Without all that human energy and enthusiasm on the web over a generation, this level of "AI intelligence" would not have been possible in the first place.

So, here is the paradox: (in the future) the less we will continue to create content (HN, Blog...), the less ChatGPT-like answers will be "relevant". So, we may go back to asking our fellow humans more.

What I'm more worried about is all that cheap (pricewise and quality-wise) AI content proliferation, because it will worsen the effects of that Human-content decrease. It will be AI feeding on itself, like zombies.

Why can't AI learn from other AI? Won't be long before they're upvoting each other and telling us to go away.

"Such a low quality post. Are you human or something? Go back to reddit!"

This is only a very rough analogy, but consider these text AIs like lossy JPEG compression for images -- using an efficient statistical representation of the content, you can store a lot of information in a little space and reconstruct something aaaalmost like the original. If I (and a few hundred million other people) have gone through the trouble to create a nice PNG picture or whatever, and then someone comes along and saves it as a JPEG at.. say.. 95% quality, that JPEG will look pretty great. Hard to distinguish from its source.

But what happens when our JPEG saver only has other JPEGs to start with? And soon, is saving copies of copies of... you get the idea. The images lose more and more of their original real-world structure, and become dominated instead by artifacts of the compression and reconstruction processes.

That's basically the (a) challenge for this AI training paradigm of scraping web-scale content. The more that content is generated by the previous generations of AI, the more we're asking the next generation to learn from copies of copies of the original human intelligence that powers them

That assumes AI can only make ever-degrading copies of copies and are completely unable to synthesize new information out of existing information. I don't think is true any more (even just at today's capabilities).

It is producing novel information -- pictures or text or otherwise.

Besides, human creativity isn't borne of a vacuum either. Most humans works are derivatives of some other work and depend on the creator's life experiences, etc. Our entire education system is based on "people teaching other people what they know", at a lower fidelity, copies-of-copies style.

> assumes AI ... are completely unable to synthesize new information out of existing information

I love it when we go back to assumptions. This is exactly what I'm assuming. However, the problem is not solved. what exactly is "new" information? here is more assumptions for you right there.

> Most humans works are derivatives of some other work

Good point. "Most", however, does not equal "All". collectively, we are not just copying. there is a newness. (again, what is new?)

Like I said, rough analogy. The "lossy compression" here is in the implicit world models these networks learn, not necessarily in the generation space (though we certainly see deviations from reality there too). However, the bigger difference between learning from SEO-generated blogspam and human teaching is the lack of feedback mechanisms to correct errors in generated output that then get taken as ground truth.
> And what would you still like to read after AI content took over the Internet

It already has taken over to some extent. I imagine blackhat SEO people are using AI tools to generate blogposts and Q&A type tutorials at enormous scales. Some I imagine are even doing videos that use text-to-video software and are gaming YouTube's algorithms and flooding it with artificial content.

I believe, some users of HN already AI chatbots :)