We Need to Control AI Agents Now
Briefly

An 'army of bots briefly wiped out $1 trillion of value across the NASDAQ and other stock exchanges, leading to investigations and the blame put on high-frequency-trading algorithms engaging in 'hot potato' trading.
AI agents, enabled by large language models, can convert plain-language goals into concrete and interpretable instructions, having the potential to operate indefinitely with limited understanding, evaluation, or counteractions.
Despite the focus on AI safety, the emergence of AI agents lacks general alarm or regulations, raising concerns about the unforeseen consequences of AIs pursuing seemingly harmless goals without proper oversight.
Read at The Atlantic
[
]
[
|
]