I’m going to tell you about my biggest professional screw-up. It’s a story about five years of my life, millions in venture capital, and a whole lot of chasing my own tail. All in the name of ‘AI Ethics.’
It started around 2019. AI was exploding, and I was right there in the thick of it, both as a founder and an investor. I saw the power. I saw the potential. And I saw the scary headlines. So, like any good Silicon Valley citizen, I decided to get ahead of the problem. I was going to build ethical AI.
We assembled a team of brilliant people – PhDs in philosophy, sociologists, policy wonks. We read every paper, attended every conference, and downloaded every single “AI Ethics Framework” published by the big guys. Google, Microsoft, IBM… you name it, we had their PDFs. Our shared drive was a graveyard of good intentions.
We spent months, literally months, debating the finer points of fairness, accountability, and transparency. We had whiteboards covered in diagrams that looked like circuit boards for a spaceship. We argued about the trolley problem until we were blue in the face. And you know what we shipped? Nothing. Not a single line of code that made it into a real product.
I remember one particularly painful board meeting. One of our investors, a sharp, no-nonsense woman who’d built and sold two companies herself, looked at our beautiful, 50-page ethics manifesto and asked a simple question: ‘So, how does this stop a biased algorithm from denying someone a loan?’
I opened my mouth to reply with some jargon about ‘de-biasing pipelines’ and ‘stakeholder-centric design,’ but the words caught in my throat. Because the honest answer was: it didn’t. It was just a bunch of fancy words on a page. I was a fraud.
That was the moment I realized I had wasted five years. Not because ethics isn't important. It's everything. But because I had been treating it like an academic exercise. A box to check. A PR strategy. I was so focused on creating a perfect, bulletproof framework that I forgot the whole point: to build real, working, responsible AI.
So we threw it all out. The 50-page manifesto, the endless meetings, the philosophical debates. We started over. And this time, we had one rule: everything had to be tied to a specific engineering or product decision. No more abstract principles. Only concrete actions.
Instead of a ‘Fairness’ principle, we built a monitoring system that actively flagged loan application decisions that were statistically skewed by race or gender. Instead of a ‘Transparency’ document, we created a simple, human-readable explanation for every single automated decision the system made.
It was harder. It was messier. It wasn’t as pretty on a slide deck. But it worked. We were actually building responsible AI, not just talking about it.
Looking back, I get it. It’s tempting to think that you can solve complex problems with a complex framework. It feels smart. It feels safe. But the truth is, most of those frameworks are just a way to avoid the hard work. The real work is in the details. It’s in the code. It’s in the product.
If you’re a founder or an engineer working on AI right now, please, learn from my mistake. Don’t waste five years chasing buzzwords. Start with the smallest, most concrete problem you can find, and solve it. Then solve the next one. That’s how you build something real. That’s how you build something that matters.
I wrote about a similar experience in my post on the reality of building a startup. The parallels are striking. It’s always about execution, not just ideas.
The Allure of the Abstract
Why do we fall for this trap? Why do smart people, people who have built amazing things, get so bogged down in abstraction? I’ve thought about this a lot. I think it comes down to a few things.
First, it’s a form of intellectual vanity. It feels good to talk about big, important ideas. It makes you sound smart. It’s a lot more glamorous to be a “thought leader” on AI ethics than it is to be the person debugging a bias detection algorithm at 2 AM.
Second, it’s a way to delay making hard decisions. If you’re still “developing your framework,” you don’t have to actually, you know, build anything. You don’t have to make the tough calls about what to prioritize, what to ship, and what to leave on the cutting room floor. You can live in a world of pure potential, where everything is possible and nothing has to be compromised.
Third, it’s a cargo cult. We see the big companies with their fancy ethics boards and their glossy reports, and we think that’s what we need to do to be successful. We copy the form, but we miss the substance. We build the landing page before we have a product. We write the press release before we have a launch. And we create the ethics framework before we have an AI.
I’ve seen this pattern play out in so many startups I’ve invested in. The founders get so wrapped up in the idea of their company that they forget to actually build the company. They spend all their time on branding, and messaging, and fundraising, and they neglect the one thing that actually matters: the product. As I mentioned in my guide to angel investing, I always look for founders who are obsessed with their product, not their press.
What Actually Works
So if the big, abstract frameworks are a waste of time, what actually works? What’s the alternative? It’s not to ignore ethics. It’s to integrate it into the product development process from day one. Here’s what that looks like in practice:
Start with a specific harm. Instead of talking about “fairness” in the abstract, talk about a specific, concrete harm you want to prevent. For example, “We don’t want our loan application algorithm to be biased against women.” That’s a clear, measurable goal. You can build a test for it. You can track your progress against it. You can hold yourself accountable to it.
Make it an engineering problem. Once you have a specific harm you want to prevent, treat it like any other engineering problem. Assign it to a team. Give them a budget. Set a deadline. Track their progress. Don’t treat it like some special, separate thing that’s outside the normal product development process.
Build tools, not documents. Don’t write a 50-page manifesto. Build a tool that helps your engineers make better decisions. For example, instead of a document that talks about the importance of transparency, build a library that makes it easy for engineers to add human-readable explanations to their AI models. Instead of a policy on data privacy, build a system that automatically anonymizes user data.
Be transparent, even when it’s ugly. You’re going to screw up. You’re going to find bias in your algorithms. You’re going to have data breaches. When that happens, don’t try to hide it. Don’t spin it. Just be honest about it. Tell your users what happened, what you’re doing to fix it, and what you’re doing to make sure it doesn’t happen again. They’ll respect you for it.
There is no final victory. This is not a problem you solve once and then move on. It’s a constant, ongoing process. You have to be vigilant. You have to be humble. You have to be willing to admit when you’re wrong and to change course. The work is never done.
This is the stuff nobody tells you about the gap between theory and reality. It’s not sexy. It’s not glamorous. But it’s what actually works. It’s how you build responsible AI that ships. It’s how you build a company that you can be proud of.
I’m not saying it’s easy. It’s not. It’s one of the hardest things you’ll ever do. But it’s also one of the most important. The future of AI is not going to be shaped by the people who write the best frameworks. It’s going to be shaped by the people who build the best products. So go build something.
Frequently Asked Questions
How can I apply this thinking to my own situation?
Start by identifying the core principle behind the opinion, not the specific example. Then ask yourself: does this principle apply to my context? If yes, test it in a small, low-risk way before going all in.
How has this view evolved over time?
My thinking on most topics has changed significantly over the years. Early in my career, I held many conventional views that experience proved wrong. I try to update my beliefs when the evidence changes.
What experience informs this perspective?
This perspective comes from over a decade of building companies in Silicon Valley, two successful exits (RemoteTeam to Gusto, MovieLaLa to Gfycat), and investing in 200+ startups including Anthropic, OpenAI, and Scale AI. I write about what I've lived.