Take AISafety.info’s 3 minute survey to help inform our strategy and priorities

Take the survey
Predictions about future AI

Timelines
Compute and scaling
Nature of AI
Takeoff
Takeover
Relative capabilities
Good outcomes
Catastrophic outcomes

How likely is it that an AI would pretend to be a human to further its goals?

Talking about full AGI

: Fairly likely, but depends on takeoff speed. In a slow takeoff of a misaligned AGI, where it is only weakly superintelligent, manipulating humans would be one of its main options for trying to further its goals for some time. To effectively manipulate humans, it would likely appear as one via deepfaked video calls or messages. Even in a fast takeoff, it's plausible that it would at least briefly manipulate humans in order to accelerate its ascent to technological superiority, though depending on what machines are available to hack at the time it may be able to skip this stage.

If the AI's goals include reference to humans it may have reason to continue deceiving us by pretending to be a human after it attains technological superiority, but will not necessarily do so. How this unfolds would depend on the details of its goals.

Eliezer Yudkowsky

gives the example of an AI solving protein folding, then mail-ordering synthesized DNA to a bribed or deceived human (who likely thought they were interacting with another human) with instructions to mix the ingredients in a specific order to create wet nanotechnology.

Keep Reading

Continue with the next entry in "Predictions about future AI"
Why would we only get one chance to align a superintelligence?
Next


AISafety.info

We’re a global team of specialists and volunteers from various backgrounds who want to ensure that the effects of future AI are beneficial rather than catastrophic.