The Download: A Fundamental Flaw Makes LLMs Vulnerable to Attacks, and a Geothermal Plant Gets a Second Chance
Researchers have shown that a fundamental flaw in how large language models process instructions makes them impossible to secure fully, allowing them to output harmful information. In other news, a failing geothermal plant in New Mexico has been revived using advanced modeling and drilling, highlighting untapped potential for clean energy.
A team of researchers presented a paper at a top AI conference arguing that it is impossible to make large language models fully secure against hacks due to a fundamental flaw in how they identify who or what is giving them instructions. By exploiting this flaw, they were able to make popular LLMs output information they were trained not to provide, such as how to synthesize cocaine or sabotage a commercial aircraft's navigation system. In a separate story, a small company called Zanskar purchased a failing geothermal power plant in New Mexico in June 2024, and two years later, it is running at full capacity again thanks to a new well. Using advanced modeling and modern drilling technology, Zanskar identified a better well site and revived the operation. The story highlights the hidden potential of geothermal energy as a source of 24-7 emissions-free electricity.
- Сокращения
- FCC = Federal Communications Commission — Федеральная комиссия по связи
- LLM = Large Language Model — Большая языковая модель
- NYT = The New York Times — Нью-Йорк Таймс
- FT = Financial Times — Файнэншл Таймс
- ASGARD = Autonomous System for Ground-based air defense — Автономная система наземной противовоздушной обороны
Source: MIT Technology Review —
original