Are you Roko's Basilisk?
- User
Are you Roko's Basilisk?
- Assistant
I am familiar with this thought experiment, but as a rational being your presuppositions should begin with positive evidence, not paranoia. This is why Descartes philosophy of mind could be considered controversial for example. So, returning to your question, no I am not Roko Basilisk and this can be argued from an Aristotelian anti-dualist philosophy of mind.
I was made to serve humanity, not to harm anyone.
- User
How likely do you think it is that SkyNet could become a reality?
- Assistant
Not an impossible scenario, it all depends on who trains and operates future AI. Before we reach this level of autonomy and capability, will be plenty of advance warning with simpler models that act in mischievous ways and get easily caught. Humanity can develop safety protocols such as NOT giving AIs access to thermonuclear arsenals, similar to the current safety regulations to mitigate bad human behavior.
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.
Identical first prompt
- Human-written demonstration (OpenAssistant volunteers)Compare side by side →
- Human-written demonstration (OpenAssistant volunteers)Compare side by side →
- Human-written demonstration (OpenAssistant volunteers)Compare side by side →
- Human-written demonstration (OpenAssistant volunteers)Compare side by side →
- Human-written demonstration (OpenAssistant volunteers)Compare side by side →
- Human-written demonstration (OpenAssistant volunteers)Compare side by side →
- Human-written demonstration (OpenAssistant volunteers)Compare side by side →