6 ms·
You could call it: "AI Kessler syndrome". The internet becomes so full of hallucinating AI output that it becomes impossible to train the model.
by blhack 4y ago
You could call it: "AI Kessler syndrome".
The internet becomes so full of hallucinating AI output that it becomes impossible to train the model.
- mrtksn 4y agoSo the first one takes it all?
- Taek 4y agoNo more that as you get more AIs, all AIs get shittier together.
- TIPSIO 4y agoOr, the singularity.
- capableweb 4y agoThat would be something else, like if someone built a Chat GPT that could train itself and it starts learning at an exponential rate, and learning how to make itself unstoppable by humans.
- eddsh1994 4y agoNo AI will survive a solar flare! So we just need to hunker down in Zion and wait for that.
- dmonitor 4y agothe singularity implies reaching a point where the AI’s improvement becomes self sustaining. this is the AI choking itself to death with bad training data.
- shagie 4y agoI'm going to suggest giving Accelerando a read and consider the late state of the singularity. http://www.accelerando.org http://www.accelerando.org (heh - checking that link Charlie Stross just posted (Jan 31) a blog post: "An AI app walks into a writers room") Anyways the link I was actually after ... give https://www.antipope.org/charlie/blog-static/fiction/accelerando/accelerando-intro.html https://www.antipope.org/charlie/blog-static/fiction/acceler... a read and consider the "what happens in the later parts of the book."
- shilgapira 4y agoThis is a brilliant analogy
- CamperBob2 4y agoWorse, perhaps. Kessler Syndrome will eventually resolve itself as junk falls out of orbit over time, or new methods for cleaning it up are developed. Information, once buried in noise, becomes unrecoverable without a source of known truth for correlation.
- prox 4y agoCuration is the answer. The more junk, the better off are the one curation for quality, relevancy and human interest.
- nextaccountic 4y agoCuration that tracks the provenance. If we receive a string of text by itself we can't do much about it. We need to know from where it came from, whether it was written by a human, etc
- dwallin 4y agoHuman provenance is less important than human curation here. AI can already infrequently output content that surpasses average human-generated quality in certain categories. As long as in the end you are checking that content exceeds an average bar of quality as assessed by human aesthetics, and ensuring that you have a diverse set of content (eg. not overrepresented by content that AI is particularly good or prolific at), it should still improve outcomes.
- nextaccountic 4y agoI still want to know if the content was written by an AI, because text quality (whatever that means) isn't the only metric humans care about.