AI driving humans extinct: Conspiracy theory or near-future possibility? A recent survey of over 2,700+ researchers has found that we could be closer than we think to an all-out tech war. Here’s what you need to know about findings and extinction risk.
According to the survey, over half of the researchers surveyed state that there is greater than a 3% chance of a total AI takeover that could drive humans to extinction. Others noted the possibility of “extremely bad outcomes,” which varies subjectively.
Per The Byte, findings also reported that the average person responding to the survey believed there to be a 10% chance that machines could replace humans in every task in just a few years’ time — with that risk jumping to 50% by the year 2047.
While the future of AI, humans, and sentience may seem bleak at the moment, other respondents held a more positive core belief; stating that there can be equal possibility for good outcomes from advanced AI.
What Do These Expert Worries And Concerns Mean?
While the 3-5% figure can seem concerning, it can be used to see how the current culture impacts our perception and awareness of AI. Many experts may anticipate danger, but it doesn’t mean that there is any at the moment.
The Byte quoted a prominent expert, Katja Grace, from the Machine Intelligence Research Institute; stating that “It’s an important signal that most AI researchers don’t find it strongly implausible that advanced AI destroys humanity…I think this general belief in a non-minuscule risk is…more telling…than the exact percentage risk.
The survey also notes that subject matter around AI is incredibly dynamic and complex, which has prompted many answering the survey to “perform poorly” per The Byte. The survey goes on to note that the survey respondents only have AI experience, and that the AI experience they do have does not translate to forecasting experience as you’d see with higher-level subject matter experts.
Despite this caveat, many experts and industry leaders find the survey results helpful, as science is comprised largely of educated guesses.
The survey goes on to state that while educated guesses are, by nature, unreliable, they are helpful in this context; as respondents are replying and informed by their in-field expertise. The data given could be used as a single cog in a broad system of information and evidence from expert sources across topics, including economic analyses and forecasting expert insights.
The Byte experts suggest that rather than anticipating a large-scale extinction event led by AI, people can instead focus on solving the current concerns around AI usage today. These include areas of concern including public opinion swaying, virus creation, deepfakes, and more as AI continues to advance.
Others have raised alarm about the racial bias that AI can have across users and contexts, which can be exacerbated by directed attacks on AI’s machine learning process.