The Language of Deception

Chapter 5Consciousness, Sentience, and Understanding

Prior to addressing the legitimate risks associated with the latest generation of language models, I feel it is important to address a major distraction that has served to divert attention away from the actual issues. There seems to be an increasingly prevalent concern among society that AI has reached or will soon reach a level of complexity that would grant it consciousness, autonomy, and sentience. This notion is often accompanied by dystopian and science fiction–inspired visions of a future where machines become hostile and subsequently enslave or exterminate humanity. But is this even a legitimate concern?

I do not believe we should dismiss the concern merely because it resembles the plot of countless science-fiction books and films. In truth, science fiction has historically been a well-received medium by which futurists have been able to effectively publicize their visions and concerns about the future, often with more widespread awareness than could be achieved with technical papers focused on the same. And many of the old science fiction themes have, over time, manifested themselves as our modern reality. Therefore, I think it's appropriate to at least consider these concerns and draw an appropriate conclusion. This chapter seeks to establish an understanding of the concerns related to the emerging sentience of computers and assess the reasonability of them. Ultimately, through a careful analysis of the facts, we ...