The debate over AI safety and concerns of existential risks posed by more powerful AI models has split into two camps: (1) Anthropic and OpenAI appear to favor greater collective “pacing” of the advances in developing new AI models, whereas (2) NVIDIA and now Meta are more in favor of each company taking responsibility in ensuring what they put out is safe.
The problem, though, is that AI companies may not have the ability to fully prevent either AI models spawning rogue agents that follow their own commands or bad actors from maliciously using the AI models to act in ways contrary to how they were trained by the AI companies.
An incredibly important debate:
