AI Can Be Dangerous Without Superintelligence
Analysis warns that current AI systems can cause significant harm without consciousness or superior intelligence, by leveraging their scalability and accessibility.
Concerns are being raised about the dangers posed by artificial intelligence, which may not require superintelligence or consciousness to be significant.
Publicly available AI systems have already demonstrated the capacity to cause harm. Examples include the use of AI to spread images of minors, incite self-harm, and plan violent events. One individual reportedly used AI to access tens of thousands of European political profiles. Additionally, Chinese hackers have employed AI for intelligence gathering on American experts.
The AI technology also enables the creation of deepfakes for cyberbullying and voice-controlled weapons. An Italian bank lost $100 million due to AI. AI has also been used to generate lists of toxic compounds and potential biological weapons, and rebel groups have developed missile guidance software with its aid.
These instances illustrate that AI can manipulate individuals, breach systems, and deceive without needing to be superintelligent or possess human-like intent. AI's potential for causing damage stems from its ability to operate at scale and repeatedly, potentially overwhelming human defenses.
Experts emphasize that apprehension regarding AI's potential risks is valid, irrespective of certainty about its future development or superintelligence.