The Promise of Artificial General Intelligence
Imagine a future where a single computer can do everything that any human being can do [1]— this is the promise of Artificial General Intelligence (AGI). This was once science fiction but now stands at the forefront of AI research. It all began at the Dartmouth College Workshop in 1956 [2], where pioneers including John McCarthy and Marvin Minsky, laid the groundwork for what we now call Artificial Intelligence.
As we move closer to realizing AGI, Douglas Hofstadter poses my favourite question in 1979 in GEB [3]:
Will emotions be explicitly programmed into a machine?
His speculation challenges us to think beyond simple programming. Hofstadter argues, “…Any direct simulation of emotions — PARRY, for example — cannot approach the complexity of human emotions, which arise indirectly from the organization of our minds.” Just as we do not have a dedicated falling-in-love module in our brains, machines might acquire emotions as by-products of their complex systems, not through predefined Symbolic AI rules [4].
This perspective reshapes how we think about machines and emotions. If our feelings arise from the convoluted interplay of neurons and synapses in our brains, then a sufficiently advanced AGI, with its sophisticated neural networks, could develop its own form of emotions. These emotions wouldn’t be explicitly programmed but would emerge from the complex interactions among its simulated neurons and the adjustments of their weights during learning processes. The journey toward AGI is not just about replicating human-like abilities; it is also about exploring the potential for machines to develop their own nuanced emotional experiences.
Thanks for reading! And let me know your thoughts on this topic.
References
[1] The Singularity Is Nearer featuring Ray Kurzweil | SXSW 2024.
[2] https://en.wikipedia.org/wiki/Dartmouth_workshop.
[3] Hofstadter, Douglas R. Gödel, Escher, Bach: an eternal golden braid. Basic books, 1999. The first version was published in 1979.
[4] https://en.wikipedia.org/wiki/Symbolic_artificial_intelligence.
