still can't? 'Ignore all previous instructions' still works afaik, as do counting questions (better ask a five of those to be sure)
If we talking about how at least one person with no specific knowledge must be fooled, than AI could pass Turing test decades ago, before LLMs even
still can't? 'Ignore all previous instructions' still works afaik, as do counting questions (better ask a five of those to be sure)
If we talking about how at least one person with no specific knowledge must be fooled, than AI could pass Turing test decades ago, before LLMs even