Crossposted from the AI Alignment Forum. May contain more technical jargon than usual.

Many people worry about catastrophic misuse of future AIs with highly dangerous capabilities. For instance, powerful AIs might substantially lower the bar to building bioweapons or allow for massively scaling up cybercrime.

How could an AI lab serving AIs to customers manage catastrophic misuse? One approach would be to ensure that when future powerful AIs are asked to perform tasks in these problematic domains, the AIs always refuse. However, it might be a difficult technical problem to ensure these AIs refuse: current LLMs are possible to jailbreak into doing arbitrary behavior, and the field of adversarial robustness, which studies these sorts of attacks, has made only slow progress in improving robustness over the past 10 years. If we can’t ensure that future powerful AIs are much more robust than current models[1], then malicious users might be [...]

---

Outline:

(03:01) Mitigations for bioterrorism

(12:29) What reliability is acceptable?

(14:28) Mitigations for large-scale cybercrime

(17:23) Fine-tuning?

(17:46) Conclusion

The original text contained 3 footnotes which were omitted from this narration.

---

First published:
January 16th, 2024

Source:
https://www.lesswrong.com/posts/KENtuXySHJgxsH2Qk/managing-catastrophic-misuse-without-robust-ais

---

Narrated by TYPE III AUDIO.

“Managing catastrophic misuse without robust AIs” by ryan_greenblatt, Buck

LessWrong (30+ Karma)