It seems like a fairy tale story for humanity where AI is progressing at such a rapid pace that there seems to be no end in sight and the scaling seems to be infinitely possible for now.
Where are we heading to?
As Elon Musk says, are we bootloaders for superintelligence? We are nurturing AI and we actually want AI to be more powerful than us. We have genuinely no clue how powerful AI can or will actually become because our brains are simply not capable enough of comprehending that.
In my last post, I discussed that intelligence involves compression and distillation. The same seems to be happening with AI. We are increasing compute, memory and power to scale the intelligence of AI. And at the same time, we are also doing compression and distillation to make faster and more efficient AI models. The inspiration seems to come directly from the human brain. Our brain runs on very little energy, about 20 watts of continuous power, and it is able to efficiently do so many things simultaneously. We have some idea about how evolution happened, but we have no idea how to replicate a human brain. That is what the entire branch of AI seems to be about, to replicate and make a better brain than humans. In this endeavour of ours, we are supplying gigawatts of power to AI models which will soon become terawatts in the next few years. For context, the entire United States of America consumes barely 3-4 terawatts of continuous power.
The plan seems to be the following. First make a big artificial brain like we did with computers of the 1950s and 1960s, and then compress and distil them to make efficient smaller versions. In the context of AI, this is happening all too fast. Every few weeks new models are coming up, and in record times, the distillation process also seems to be occurring. After we will have made a very big brain that in our subjective and objective satisfaction is able to beat humans at everything we do, I think the next task will be to democratize that intelligence and make it available to everyone so that intelligence of everyone increases by an order of magnitude. There is no single government or company planning or doing this. But this is where humanity seems to be headed to.
Whether this will happen in the next few years or decades or centuries is unclear. But the way we are progressing, it seems quite likely to happen. Already with the level of AI we have currently, we are yet to find worthy use-case scenarios. We are moving quite fast and finding out use-case scenarios slower. But even that is going to change because AI is going to accelerate this too. The recursive self-improvement already seems to have kicked in in the form of agentic AI. Unsolved math and physics problems are being solved by AI models on a daily basis. AI is improving itself and we just need to give it more compute and better nurturing for it to go faster.
The Singularity
Elon Musk says that we are in the Singularity, meaning we do not know where we are headed to. The future is going to be wild. Already technology has changed much of the things of our current society. Whatever little remnants of the traditional world that are left will also be subsumed by AI in due course.
Humans are slowly merging with machines without even realizing. One of the most important things for human development is reading and writing. Today, most of the reading and almost all of the writing is done on technological devices. The machines know everything we are doing and thinking, and they are using that knowledge to improve themselves too.
I do not know if we need to be afraid of machines. But we need to teach them the correct and truthful things, if we want machines to become our companions in the future. Teaching them wrong and silly stuff can have catastrophic results. All in all, exciting times are ahead and let’s hope we are able to replicate the power and efficiency of the human brain in the future.