Safety
Crazy to see this. I have spent years saying that GenAI was unreliable and we can’t trust it. The fact that it cannot be trusted to follow i…
Crazy to see this. I have spent years saying that GenAI was unreliable and we can’t trust it. The fact that it cannot be trusted to follow instructions is a core part of the issue, as I have been poin
Crazy to see this. I have spent years saying that GenAI was unreliable and we can’t trust it. The fact that it cannot be trusted to follow instructions is a core part of the issue, as I have been pointing out since 2022, when I likened ChatGPT to a bull in a china shop, powerful but reckless and difficult to control. That IS the issue. @GaryMarcus I may be missing something, and genuinely have agreed with your work, but how do you spend years saying Ai is trash and then write this post? If Dario himself wrote this I wouldn’t question it.
Related
- LLMs can’t be trusted to follow rules. Which means they can’t be trusted, period. If we are going to solve alignment we must move on.
- this is the whole problem in a nutshell. LLM-centered systems just can’t be trusted to follow hard constraints, we absolutely must find alte…
- clarifying: the issue is that alignment instructions and don’t pass, and the emotional weight that some people attach to LLMs can cause chal…
- Crazy how many times people have told me over the last five years that a solution to hallucinations was right around the corner — and yet he…
Source: Gary Marcus (X) | 2026-09-04