Could artificial intelligence cause the deaths of one billion people?
Bill Gates warned that the use of artificial intelligence for malicious purposes could threaten the lives of billions of people. Meanwhile, cybersecurity expert David Linthicum considers a scenario resulting in the death of one billion people to be theoretically possible, but believes its likelihood is low unless numerous serious errors occur simultaneously.
Could artificial intelligence lead to the death of one billion people?
In an interview with NBC News, Microsoft co-founder Bill Gates stated that modern artificial intelligence tools could become an extremely dangerous weapon in the hands of malicious individuals. He also emphasized the necessity of regulating this technology at the legislative level.
The publication Tom’s Guide turned to David Linthicum, an expert in artificial intelligence, cloud technologies, and cybersecurity, to discuss this warning.
According to Linthicum's commentary, for a disaster of such scale to occur, AI must be able to influence critical infrastructure, weaponry, supply chains, or biological systems. Furthermore, it must be capable of acting without human intervention.
In the expert's view, factors such as major errors in interconnecting systems, serious management flaws, and the deliberate use of the technology for harmful purposes would need to align for this to happen. Therefore, he does not consider such a scenario to be inevitable or likely to happen in the near future.
Linthicum noted the probability that Gates meant terrorists or unstable states could use AI. However, according to his assessment, there are major practical obstacles on the path from a malicious idea to executing a large-scale disaster.
At the same time, the expert does not deny today's risks associated with artificial intelligence. He included data leaks, fraud, deepfakes, improper automation, and biased decisions among current problems.
His main concern is the connection of insufficiently refined AI systems to real corporate and government systems. If some organizations deploy the technology without the necessary control mechanisms, this could cause harm.
To reduce risks, Linthicum suggested granting AI only the authority necessary to perform its tasks, isolating systems during operation, and tying high-risk actions to human approval. He also noted the importance of logging all system actions, conducting regular testing, and maintaining the ability to roll back changes.
According to the expert, relying solely on an emergency shut-off button is not enough. If a human detects the problem too late, damage may already be done before the system is turned off. Therefore, security should rely on multiple control layers and be taken into account from the earliest stages of software development.