Written by Yusuf Gürdoğan
on 

Future of AI: Thoughts on Acceleration

Do I know the future? No. But it’s good to make some predictions!

Unlimited entertainment online

Your favorite movie: but now with this actor, or with that scenario.

The pixels you look at will be so much fun, and it’s unlimited. And it will be so realistic.

Videos are going to get so much better. It will be quite difficult to tell what is real.

Making a video will be very easy; since AI knows your voice and speaking patterns, it will create the entire video from a script.

Real life entertainment has limited supply. Digital entertainment has unlimited supply. How is the demand going to be shaped?

More realistic deepfakes & scams

What to trust? Currently, deepfakes can usually be separated from reality. Within a few years, it’s likely that it will be impossible to separate. Then it could be very difficult to trust pixels. What is real?

A supercomputer in every home

Running an AI model that is smarter than us, locally, will be possible and affordable. This can provide intelligence to many robots in the house.

Tinybox is “now shipping”: tinygrad.org/#tinybox

They say, “we will commoditize the petaflop and enable AI for everyone.” tinybox Project Digits “will be available in May” 2025: Project Digits

AI agents will accelerate production exponentially.

Asset prices (such as stocks, housing, etc.) rising detached from the real economy Video by Şant Manukyan, view chart.

Black line: American household wealth

Yellow line: American economic development

For many years, there was a parallel increase in household wealth alongside American economic growth. In 96-97, there was an acceleration in wealth faster than the economy, followed by the 2000 dotcom crisis. Then household wealth accelerated again, with an even bigger divergence and convergence than the pre-2000 separation. Now we’re facing a tremendous divergence. The essence of the video isn’t just that “these two will converge” - though they will - but rather how will they converge? With AI and many related developments like space research, biotechnology etc., we might see the yellow line go up, or, as some at ZeroHedge put it, the black line might crash downward.

We could be in a “number go up” situation. All numbers can possibly go higher: productivity, therefore output. Perhaps we will all get rich.

I can now find a niche problem that people have, and try solving it by making a project. What does AI agents lack to do the same, better than me?

Currently, I have an idea. I tell AI what to code, and it writes the code. I just copy and paste and run it. When I see a problem or when I have a new idea, I tell that, and AI just does it. I’m just testing the code and giving feedback, and developing a product. With about 10x speed compared to 2022. AI can test it too, actually. And write to itself back. And it also can have new ideas. And ship the product.

What do WE do in this case?! Maybe it’s time we just touch grass and enjoy reality that we are sure of.

AI agents as a service? Possible. It would generate the best business ideas and make apps in just a few minutes. Package deals being rented for thousands of dollars per used minute by the leading AI companies or their partners.

They can solve many problems quickly, and multiply the current productivity. Make a house for 90% less of the cost, your container will be there at your desired place. Build and design your own car online and it gets shipped to you in a week. Walking robot dogs everywhere on streets, carrying and delivering packages. Self driving cars everywhere. Humans have always the top priority, and all robots respect humans.

When you ask for health advice, AI tells you to consult your local doctor. Maybe, at some point, doctors will say consult your local AI.

I hope we can all enjoy having more health, wealth, and being good.

What AI can’t do now can be done later. What can AI never do? Believe in God? Build a family? Have values?

AI beats us in both speed and smarts already. Cerebras and Groq outperform human speed, while Claude outperforms human intelligence in most areas. What makes us valuable as human beings?

The best models will always be closed source.

I’m hoping that this won’t be the case, but it’s likely that it will be the case.

Let’s hope it will be in the right hands, or, let’s be the right hands and make the best model. The race is open, the game is definitely on, and anyone (with a few billion dollars!!!) can participate.