The original title is "The Download: tricking LLMs, and reviving geothermal plants"

Short summary

MIT Technology Review's daily newsletter covers two topics: a fundamental flaw that makes large language models inherently vulnerable to adversarial attacks, and efforts to revive geothermal energy plants. The LLM security piece argues that full security against hacks is impossible due to how LLMs fundamentally operate.

• •LLMs have a fundamental flaw making them impossible to fully secure against attacks
• •The newsletter also covers geothermal plant revival efforts
• •Part of MIT Tech Review's daily Download newsletter

Generated with AI, which can make mistakes.