Former safety researcher at OpenAI Steven Adler says that the AGI race is not just risky but also comes with a “huge downside.”