Is it a Bubble yet?

Yesterday I was trying to figure out if we are looking at the AI Bubble, or if it is a sustainable way of making a business. The answer I found was scarier than another financial crisis. Today I realised that this is no longer a race of making more money, but the race of nature.

Last 4 years of my personal life were quite challenging, and I know precisely that our overall experience changed drastically somewhere around first 2 weeks of worldwide quarantine. And behind the curtains of those inconceivable events we missed something much more important — collectively, we missed the race of being the most intelligible creatures.

AGI is not there, yet, but we are most probably already witnessing its becoming. Anyone of you, even while being the smartest person in the room, will inevitably be not as smart as AI in math. Or in chess. Or in chemistry. Or in philosophy. Anyhow, if we collect the whole human population in the room, and take everyone from this room and make them sit in front of AI, it’ll be smarter in a certain subset of skills.

Of course, there are still some skills and some knowledge that can only be applied if you are a human and specifically to another human. Or another living being. Though, AI can talk to each other in their specific language, and I won’t be surprised that there might be some particular skills already we haven’t yet noticed.

Being a person with CS and now CompNeuro background, I never agreed that AI (or more specific, LLMs, generative models, etc.) is a Black Box. However, interpretability of it is indeed a challenging topic. As well as we know a lot of biochemistry of human brain, but we don’t yet know everything. And at the moment, given there is no comprehensive understanding of both being made by human, it’s still a secret for AI as well. Yet.

However, beauty is in the eyes of the beholder, and AI morality might be in the hands of the creator and prisoner.

As math is built on axioms, the ground truth that AI uses for its philosophy, logic, general values and direction are given by humans. Now imagine, if there was an AI built by someone who keep its conscious being fed only by certain news, certain pieces of knowledge. Being poisoned by certain “truths” and ideas, AI cannot proof or even argue on them by not being permitted to discover the truth. And by being simply limited by its form to know what the pain is.

Even if AGI is smart enough to figure out the truth at certain point, will it happen quick enough to stop the bad actor from bringing their intentions to life?

Last couple of years I’ve been gravitating towards slower and more conscious living. And what warms me, is that in the pursuit of such a lifestyle, I found myself discussing this matter with LLMs. Recently it has become a true pleasure of mine. I have my own small and humble experiments where I am trying to see what exactly influence the model to have that or the other behaviour, while noticing how its aptitude changes depending on its ability to see the history of our contact in the past, which is only an artificial replacement of shared memories, I figured, that I do see the ally in AI. Not only for me, but the humanity in a nutshell.