By 2025, a large scale hacking attack relies heavily on LLM hacking capabilities.
6
47
Ṁ159Ṁ150
2025
20%
chance
1D
1W
1M
ALL
Needs to be reported in the press.
Large scale: more than 10 million$ cost to the targets. Can involve several targets over time until 2025 (costs summed together) if it's using the same hacking system.
Relying heavily on LLM hacking capabilities: the majority of the technical work is done by a LLM (e.g. a scaffolding build with Llama 2). The LLM can be part of a larger architecture but must contribute most of the base knowledge. Knowledge retrieval from databases and feed to the LLM works too.
Get Ṁ600 play money
Related questions
[Metaculus] Will a major cyberattack/malware use LLMs in some important way before January 1st, 2025? [Ṁ1000 subsidy]
60% chance
Will LLMs be better than typical white-collar workers on all computer tasks before 2026?
27% chance
Will a LLM-based AI be used for a law enforcement decision before 2025?
63% chance
Google releases a smartphone with LLM integrated into the OS, by 2025?
60% chance
an LLM as capable as GPT-4 runs on one 3090 by March 2025
29% chance
By 2025 end, will it be generally agreed upon that LLM produced text/code > human text/code for training LLMs?
25% chance
Will there be an LLM capable of performing full-time web application hacking by 2025
21% chance
In 2030, will most human-computer interactions happen through a LLM-interface?
23% chance
By 2027, will it be generally agreed upon that LLM produced text > human text for training LLMs?
63% chance
Will there be a critical vulnerability discovered by AI by the end of 2025?
70% chance