I’ve been thinking about starting a company based on Generative AI technology , drawing on what I've learned and done so far in my PhD in Cambridge and my work at To0Space as a R&D leader. The natural questions are: what should be the business of my startup? What service could we provide? But before answering these, I think something much more fundamental needs to be answered first: "What value can the Generative AI technology bring to the human race in the long run?"

The analogous question for the Internet technology was asked at the turn of the century. Today, we are fairly confident about the answer: The Internet benefits the human race by providing an efficient way to distribute information. The technology has been integrated into new infrastructures of our lives: Google for Search (distribution of information), Amazon for online merchants (distribution of products), Uber and Deliveroo for short-term hire (distribution of short-term supply and demand information), TikTok and RED for entertainment (distribution of created content), ..., the list goes on.

L0oking at the giants that survive to this day, I would argue that a necessary condition for success is to leverage the game-changing advantage of the technology. In the case of the Internet, the advantage is the cheap and fast distribution of information. This is a closed case in 2025. The burning question today is: what is the game-changing property that AI brings to the table?

Position 1: AI is best suited for helping us, not replacing us

Many glaring headlines warn that AI will replace the majority, if not all of the human workforce (i.e., an AI take-over of the world!). But I just don't see that happening. Many professions, accounting, law, politics, fashion, finance,…, just to name a few, ultimately require human presence because someone needs to take responsibility for other human to work with you. It is the basis of trust. Human participation will always be a mandate for these professions and so a major replacement is out of the question. Besides, AI is really not good at guaranteeing 100% accuracy. Just ask GPT to count the number of “r”s in the word “strawberry”. Human makes mistakes, too. But human can be held accountable, whereas it will take long, mainly for legislative rather than technological reasons, before we can hold AI accountable.

We are not really seeing AI replace a substantial industry. In contrast, we have seen AI thrive in aiding human: coding, email writing, generating creative images/videos. These are the most frequent use cases as of today. Think about how most end users use OpenAI's ChatGPT - they use it to draft texts, search information and write code-snippets, so that they can complete creative tasks better and faster. Again this shows the aiding nature of AI. I believe the fundamental change brought by AI lies in amplifying the human creativity, not replacing it. Efforts that go against the nature of the technology are doomed to fail, just like those who failed in the Internet Bubble.

Position 2: AI startups should look at the “Long Tail”

The “Long Tail” theory is proposed by Chris Anderson in his book The Long Tail: Why the Future of Business Is Selling Less of More published in 2006. Drawing from the success of Internet companies like Google and Amazon, he observes: the real demand curve over products, which was previously distorted due to limited shelf space and high distribution costs, has been revealed. And it is a long-tail distribution.

What struck me was the fact that if the channels of distribution are available, almost every product has an audience. 98% of the music on iTunes has been downloaded at least once. This is massively different from the music section at WalMarts, where no more than 50% of the tracks will ever be bought once. The Internet has moved us to a different economy where the niche products almost always find an audience.

What the Internet has also shown is that if reaonably convenient tools are available, almost everyone will "go out of their way" to create some niche products for the market. This reveals something fundamental about us human beings, that we have an internal drive to create and to share our creation with others. Compared to the early days of the Internet, there is a lot more tangible incentives in creating these niche products. But remember, in the early days, these people only did it because they wanted to. And without these productions, the Internet giants like TikTok will never be successful.

Such was the Internet age: both supply and demand became longer-tail, and the Internet serves as the connecting bridge. What about the AI age? I believe the same trend will follow. The AI technology will further reveal the demand curve which is ultimately more diverse and flat. This will be done through AI’s aid in further lowering the cost to create and distribute, via more powerful assistant tools and more powerful and convenient searches.

Think about how much easier it is now to write a blog post today compared to when we just had the Internet and to when we only have paper medium like newspapers. Of course, delegating everything to AI is not gonna work (as per Position 1. Google has actually started filtering purely AI generated content and I believe this is for good cause), but if you have an original idea, it is now much easier to make a strong piece of writing that can influence other people. We are already seeing signs of AI-assisted content creation. But I believe its full power has not been unleashed yet. There are still opportunities in the space.

Position 3: Start with Why and the Higher Cause

I read Start with Why by Simon Sinek in the past weeks. It’s been an enlightening read. It deals with the personal motivation of entrepreneurs and the long term value of companies. It reveals something fundamental - people follow you because they believe in WHY you are doing the business, not WHAT or HOW you are doing it. Your WHAT and HOW are, if you are authentic, manifestations of your WHY.

Why do I want to start an AI startup? As I read the book, this question is begging for an answer. The road of entrepeurship is hard and long. Statistics is unfavorable, over 90% of startups don’t last more than 3 years. If money is the goal, then, risk adjusted, there are surely better options in terms of expected return. I really need a why, or a cause, to start on this less-traveled road.

Here's my personal why: I am a firm believer in human creativity. Looking back at my path from High School to College and to PhD, I’ve always tried to understand and build things from first principles by myself. That led me to Robotics in High School and later Language Models in University. These are fascinating machines that other people built and I couldn't help asking: can I build it myself? I find myself at my best if I am truly intrigued by the system and when I set out to build one from scratch.

So coming to realize that the true value of AI technology lies exactly in aiding human to create was an exciting moment. I believe what AI will do is to allow likely-minded people, those who are internally motivated to build things themselves, to engage in the creative process more easily. The prerequisite of domain knowledge is going to be lowered so that every reasonably educated person can engage in creative works in almost every field. People will be able to design their own living spaces, write their life stories into novels, and make more creative contents with the help of AI. We are already seeing this happening with the current technology (e.g., ChatGPT and MidJourney). An exciting world lies ahead where human thrive not only in material abundance, but also in being empowered to create like never before. 

I believe many people believe in the cause of "Do-It-Yourself". In fact, that’s the Punk Spirit:" it doesn't matter if you are a good musician. What matters is that you have something to say". For me, bringing that vision to life where everyone is empowered to create is my WHY to start a startup.

Post-Script

I wrote this post to document my thoughts on AI startup at the moment, after a month of intense learning and reflection during my holiday. These are fairly general thoughts about the direction of AI startups. But I think answering these fundamental questions are necessary for all entrepeneurs in this space.

If this reasonates with you and you feel like reaching out, shoot an email and I will be happy to chat!