Discussion about this post

User's avatar
Dan ODonnell's avatar

I couldn’t help but notice (as a classically trained reader, the mention of AI may hurt your ears) that much of this discussion is what engineers building artificial intelligence models struggle with. Just got back from a conference in Texas where I spoke about AI and the future of marketing. How do you create a human-like machine? What knowledge do you give it to learn and understand knowledge like a human?

Like the creature in Frankenstein, who observes the De Lacey family through a cottage wall and tries desperately to understand human connection, warmth, and belonging through knowledge alone, AI systems occupy a similar space. The creature's sad story, learning language, reading Paradise Lost (understanding his own identity through comparisons of Adam and Satan), grasping philosophy and yet remaining forever far away from real human experience, leaves us asking an important question in machine intelligence: Can knowledge ever substitute for being alive?

Two schools of thought wrestle with this (early AI engineers in the 60’s spent hundreds of hours debating these theories). The Symbolist approach treats knowledge as you do… as connection with objective reality, trained in rules, logic, and structured illustrations. An idea that AI learns through detailed instruction, itemized rules, mirroring how you taught your children to read with intentional frameworks. Or, to function in a more human capacity, should building AI models follow the Connectionist approach, by contrast, builds knowledge through pattern recognition and neural networks, learning world truths from experience rather than instruction… closer to your description of unconsciously learning to hold one's breath underwater.

Your though about intention in pursuing knowledge applies to AI: systems trained merely to optimize patterns without solid grounding in the ‘basic goods’… truth, beauty, practical reasonableness… where risk becomes intelligent tools for harm. Your second condition for balance also feels prophetic: an AI that pursues only knowledge without understanding 5 senses and true experiences becomes the "brain in a jar" you warn against.

Maybe, the truly human-like machine isn't one that knows more, but one that knows wisely… understanding which knowledge to pursue and when to pursue it. Which combination of knowledge inputs allows humans (or machines) to be wiser?

I need more coffee... it is too early for this kind of thinking. :) Dan O'Donnell

3 more comments...

No posts

Ready for more?