Some thoughts on some problems in AI this morning as I consider the roadmap for DAISI. In some ways this is a case against centralization of compute. In another way, I just can't understand why the very intelligent folks at these companies are doing it this way when there are alternatives.
Tech Scaling Costs
The big guys like OpenAI, xAI, etc. cannot build new, cutting-edge AI data centers - it's impossible. Elon built "Colossus 1" in just 19 days, but that was largely due to some very fortunate real estate and technical circumstances. Colossus 2 started in 2025 and is still not complete, though it has become partially operational.
By the time they build these data centers, the depreciated lifespan of their chips is half gone. Not to mention that with new chips coming out every year that are 2 to 3 times better, it makes their original investments obsolete. Once the gig is up and they have to realize the capital expenditures in realistic timeframes of one or two years instead of five years, the funding will dry up and they will be stuck.
Top all of that off with the fact that every single chip that TSMC can produce in 2026 for NVIDIA has already been sold. Every. One.
There's not alternative that is comparable to Blackwell, so why would anyone make $100B investment in an inferior chip knowing the others are superior? Intel and AMD are trying to catch up but are still far behind. Even if they succeed, where are they going to build these new chips? There aren't fabs for them.
I won't even start with the sh!t show that China invading Taiwan would cause.
Chip manufacturing and data center obsolescence remains a massive bottleneck to global scalability. It's only going to get worse moving forward.
Unlimited Power Requirements
There's no good way to say this. Building nuclear or natural gas power plants to power centralized AI data centers is insane. The costs alone can add tens of billions to a project, not to mention the real danger of fallout.
Do we really want tech companies monitoring and maintaining these huge infrastructure projects? Do these companies really want the headaches associated with this? It seems like the answer to both questions is no.
So why are we doing it this way? In the race to win at all costs, they are burning money in effigy to the gods of expedience. There has to be a way to do this on today's grid without this nuclear option.
Consolidation of Compute
The powers that be at Open AI (and Microsoft by proxy), xAI, Google, Meta, and Anthropic have already become the gatekeepers of data and truth. Just look at what social media has become: a marketplace of consumer data for better marketing and controlling popular interests. Whether fads or politics, ideas are the real currency in today's world and there is a lock down on what is allowed and what will be tolerated.
The old adage, "If the product is free, then you are the product", rhymes pretty well with what is going on in today's AI space. What becomes the product once these multi-trillion-dollar investments become commoditized and price wars keep them unprofitable into perpetuity?
You. You become the product. What are you searching for in AI? How are you feeling? Need a confidence boost? They see you are planning a trip to Rome with your family... Here's an ad for a travel agent in your area or did you notice they have Italian wine on sale at Costco?
It's coming and the data privacy issues of the past are going to seem paltry by comparison.
However, this consolidation isn't just about privacy and targeted advertising. It's also about choosing who gets access to these near-AGI systems, who gets to build the companies of the future on these precious resources, and who plays gatekeeper.
I imagine a world where if a person wants to start a company that can actually compete, the first step is to prove that they can raise enough capital to pay these few tech overlords their ransom for enough compute power to run the AI that can compete with the existing solutions and products on the market. The scarcity of compute will be overwhelmed by demand soon and when that happens, a very few companies will be able to afford the skyrocketing costs.
Lost Purpose and Individualism
After AI and robotics eliminate the need for 40-50 million jobs in the US alone over the next decade, what will those people do to survive the hyper-consolidation of wealth that will be a byproduct.
Elon says that "working will be optional" in the next decade. You won't need to work because you will have everything provided for you: education, medicine, housing, food, childcare, or whatever else you need. Notice he doesn't say "everything you want" and that in this system everyone becomes the same, homogenized, lowest common denominator of human existence.
Since when did this type of welfare-by-another-name make people feel valuable or find their purpose? A man needs a purpose to find fulfillment - a reason to get up in the morning and take on the world. Raising a family and work, which I define loosely as the process of creating new things while working on teams, fills that purpose gap for most people.
If he is right, it means that individualized, purpose-driven lives could be gone in a decade. Everything to be provided to the sheeple by a small group of systems, controlled by power-hungry narcissists who also happen to be the wealthiest elites in the world. It's the apex of socialism or even communism when it reaches its full form.
Alcohol, drugs, zealotry, fetishes, and debauchery will fill this purpose gap for most people, especially men. I wish I could say that the best in people will come out and they will spend this newfound free time on music, art, writing, and philosophy to make the world a better place for everyone to exist, but that hasn't proven to be true in the past when large swaths of the population find themselves with idle hands.
The currency that will be bartered in this type of world will be the ability to provide fleeting moments of simulated purpose, or ways to dull the sense so that you don't have to face that reality.
Doesn't sound like a world that I want to live in or one that anyone will want to raise a family.
My Humble Solution
First off, I don't have all of the answers. My projects aren't going to solve the lost purpose of our species as a whole or even stop robotics from eliminating millions of jobs.
But the list above does articulate some of the problems that I think DAISI can help solve. I think we can help reduce some of the tidal forces that otherwise consolidate power and cripple our access to these technologies. I want to continue to enable people to build things, grow their own companies, and to have the means and desire to start new families.
A distributed AI system, like Daisi, is a way to build AI where regular people have a part to play in its success and can maintain a piece of control in the agentic world that is coming. There's no gatekeeper to the system and there's no overlords holding the keys to the kingdom.
Also, it works on today's power grid and with regular CPUs and GPUs that already exist, so no shortages, rationing of compute, and scalability issues.
I'm excited about it. If you think you can contribute somehow, reach out. It's my opinion that humanity needs all the help we can get.