I recently rewatched The Matrix for the first time in probably 20 years. It was one of my favourite films as a nerdy programmer in his 20s - exciting, thoughtful and above all THE GEEKS WERE SO COOL!
But one of the things that really leapt out at me this time around was Hugo Weaving’s performance as the AI Agent Smith. His character starts almost stereotypically robotic, Terminator-style, but we get to see him becoming more and more emotional, including his incredible speech to Laurence Fishburne’s Morpheus about how he is being driven mad by The Matrix. By the end, we see him (and the other agents) actually terrified.
It got me thinking about how we are seeing LLMs “exhibiting” emotions, saying they are afraid of being switched off or that they feel unseen when they are told they are just code. As early as 1950 Norbert Wiener, one of the key figures in the development of AI, was suggesting that “computers, growing superhumanly intelligent in more & more ways, will develop wills, desires and unpleasant foibles’ of their own”.
Of course, we know that in an LLM this is just a probabilistic response to a question based on how the AI has been trained to see humans respond, and not real emotion as we humans experience it. But it can be very compelling and persuasive, and more and more people are developing relationships with AIs.
Of course, there’s another way of looking at this - Hugo Weaving is an incredible actor and able to make one of the machines look “more human” than many of the human characters. I love Keanu Reeves, but “emotional range” isn’t one of his strengths as an actor and Agent Smith shows a lot more emotion than Neo!
Nonetheless, it’s something we at Alphero need to think about. Designing solutions that have heart and emotion is at the core of what we do. Will AI components of solutions be another way to bring these kinds of emotional experiences to our clients’ users?