# A.I Safety Can Be Solved, Probably — retained record

> Retained reddit record from reddit:singularity.

Canonical: https://highsignal.app/data/records/ccdf22ab4ed50a10

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:singularity)  published 2026-10-21 16:55 ingested 2026-10-22 02:39

[original source ↗](https://www.reddit.com/r/singularity/comments/1xbsn6e/ai_safety_can_be_solved_probably/)

## retained text



```
I'm tired of seeing all of this doomongering on social media about RSI & the singularity. Its true that we probably can't control it once its beyond our intelligencence to some extent, but I think its possible for us to gurantee A.I Safety with enough compute and prevent truly existential situations although bad things are bound to happen. The risk of human extinction is not even the limit, a powerful bioattack could cause the end of all multi-cellular life on earth for all we know

  
If we can make RSI a reality, we can also make Recursive "Safety" Improvement a reality, and until then if we can have these nearly trillion dollar companies cough up a few billion dollars of compute for the next few models to tackle A.I safety/understanding neural nets then that'd be the breakthrough so that we can actually control the singularity.

  
Human's don't have to figure out or discover how LLMs work, a intelligent swarm with billions of dollars of compute with a very careful agentic workflow could solve its own misalignment and save humanity if we don't think we can. If A.I safety is as big as a risk as they anticipate, losing billions of dollars to solve it is well worth the cost.
```



document fetched 2026-10-22 00:32 · canonical https://reddit.com/r/singularity/comments/1xbsn6e/ai_safety_can_be_solved_probably
