Hacker News new | ask | show | jobs
by visarga 1092 days ago
I believe low to medium quality algorithmically-generated content is easily identifiable with large language models. Consequently, spammers may find themselves needing to leverage LLMs to produce text that appears high in quality. This implies that the content generated could either be of substantial quality or well-disguised gibberish.

Therefore, some form of reputation system remains necessary, like those used for scientific publications. I predict websites that provide a trusted reputation system will have a lot to gain in the future. Github stars, upvotes, retweets, citation count, or just good moderation - they will be essential in solving the spam/hallucination problem.
1 comments
rchaud 1092 days ago
Respectfully, to expect to use AI to solve a problem created by using AI, is wishful thinking.

We haven't even solved the email spam problem, and that started 30 years ago! We have simply accepted that the largest players in the business (MS, Gmail, Yahoo) will decide on our behalf which emails we will actually see. If you want to start your own email service, fine, just keep in mind that the big players might think you're a spammer.

link
visarga 1092 days ago
But you can check your Spam folder if you disagree with the classification. Don't you agree the current anti-spam systems are rather good? Very rarely do they make a mistake. I haven't seen a spam in ages.

> to expect to use AI to solve a problem created by using AI, is wishful thinking

Fight fire with fire. Manual verification can't keep up with bots. We need "protective AI" to counter "aggressive AI".

link
pellucidar 1092 days ago
Some spam is just thrown away, not put into your Spam folder, so there's some lost opportunity to disagree (and it's hard to judge whether you would have without seeing the spam).

Also, GMail is no longer good at detecting spam. It consistently puts my phone bill into spam, as well as other repeat emails I've marked as not spam. More importantly, I get bursts of actual, obvious spam in my inbox at least twice a month, possibly more often. Maybe Google has just given up and some other anti-spam system is better.

link