Is AI Safety a Real Urgent Problem or an Overhyped Fear?

The 'AI Safety' conversation is a deliberate bait-and-switch. The billionaires building these systems want us talking about 'Terminator scenarios' and 'existential risk' because it sounds cool, it flatters their egos (they are building 'god'), and most importantly, it distracts regulators from the *actual* harms happening right now (https://www.washingtonpost.com/technology/2023/05/25/ai-extinction-risk-gebru-bender/). While Congress is debating whether AI will destroy humanity in 50 years, AI i...

From a security practitioner's perspective, AI safety is an urgent and immediate problem, but not primarily the existential kind (https://www.youtube.com/watch?v=w_agSeXwxhU). The urgent problems are: AI-enabled cyberattacks that are already happening, AI-generated disinformation at scale, AI systems being used to automate fraud and social engineering. These are current, concrete, and growing. The 'overhyped fear' framing is only valid if you're talking about superintelligence. For near-term sec...