10 min read



Like any good American I am functionally illiterate. Take, for example, the word “avuncular”. What the hell does that mean? For decades I managed to resist the corrupting influence of a dictionary and coast on vague connotations. It had something to do with being fat, maybe? Perhaps with candy? Ventricular and avuncular share some rhyme and rhythm, which surely suggests something? Unfortunately, we cannot all remain charmingly naive forever[1]. This is the story of how I learned the definition of avuncular, or why one should never take candy from strange old men.

In 2008, I was a theoretical neuroscientist at UC Berkeley’s Redwood Center for Theoretical Neuroscience with a rather fancy schmancy pedigree from Carnegie Mellon and Stanford Universities. I’d published research in Nature, Neural Computation, and NeurIPS (one of the premiere machine learning conference[2]). I was in the midst of a research collaboration with Google developing audio search based on the actual sounds. Type in “movie:godfather gunshot” and find a YouTube of Michael Corleone murdering Sollozzo and McCluskey[3].

The project with Google was my third experience turning pure science research into invention, building a living thing in the “real world” that could directly affect people's lives. I’d invented algorithms that allowed cochlear implants to learn to hear and helped build systems to detect lies off of raw video. I love and believe in pure science, but inventing was addictive. And so, when my wife and I began toying around with an idea to use artificial intelligence[4] to transform education we immediately asked ourselves: is this a science experiment or is it a startup? My first tech company, Augniscient, was born from that question.

My wife was faculty at Berkeley’s School of Education, and every time we talked shop I’d get a little of my AI chocolate in her learning sciences peanut butter[5]. We began to talk more and more about how my work with brains and machines could be applied to understanding students as they learned. What if we could listen to them chat with each other online or even via microphones in the classroom and predict their performance on the final exam? If you knew three weeks into a class that 5 students had a deep misconcept, why wait until the exam for them to fail. In fact, why give the exam at all. This was our crazy idea: replace biased, high-stakes tests with “personalized” insights about each student. Depending on your feelings about AI this idea might seem obvious, dangerous[6], or revolutionary, but in 2008 it sounded like magic. There was no Coursera, and Khan Academy was just some guy’s YouTube channel. The AI “revolution” driven by deep neural networks was years away from the mainstream. I love science, and if I was leaving it for a startup, it was going to be for something crazy. Augniscient would end high-stakes testing and all the bias it brings.

So, I had an algorithm and a plan and several fancy degrees. We added an amazing cofounder who was working at Yahoo back when that was something to brag about. We put together a business plan to transform education and built prototypes to highlight the technology. The next step, as with all Silicon Valley startups, was pitching the VCs, that fraternity of venture capitalists who hold their masonic rituals on Sand Hill road, just outside Stanford. It was like pulling fucking teeth.

It is the job of VCs to be brutal. I’ve started many companies over the years and pitched to many VC firms[7]. It’s always incredibly hard and should be. Millions and tens of millions of other people’s dollars shouldn’t just flow into the pocket of someone’s drinking buddies[8]. Kicking the tires is part of the process.

My first time through the process with Augniscient wasn’t exactly encouraging. We pitched quite a few firms, and they all loved our technology. “You can read students' minds?” as one VC misunderstood it. “I’ll invest $2 million if you do financial fraud detection. And, I have the perfect CEO for you.” They loved our technology but no one wanted to invest in education back then, or in us. When a VC locks eyes with the current CEO and says, “I have the perfect CEO for you,” it’s like telling your wife that you just met the perfect woman to bear your children. But for exceedingly rare circumstances, you might have dramatically misread the relationship.

We persevered. Our pitch and our business plan matured until we won our first full partner vote. After several layers of screening and pitching we finally had the chance to convince all five decision-makers at the firm to invest in our company. We pitched them, and then the questions came. The first partner asked a tough question, but I’m ready. The next makes a series of brutal observations but we’re resilient. The next guy…downright eviscerating, and the fourth simply left me feeling that I must have done something cruel to his family.

Then there was the fifth and final partner, the oldest in the room by quite a bit. No posturing, no tire kicking. He just smiled and said, "You should be so proud of what you've built."

After two hours of brutal interrogation I was so grateful. All I could think was, "At least one of these guys is on my side."

Let’s freeze this story, just briefly, and flash forward 6 years[9]. I’m watching a TED talk by Dame Stephanie Shirley. She is the rather astounding (and riotous) founder of Freelance Programmers (later called Xansa). Back in 1959, she created a company that hired women as computer programmers and set many of the programming standards we still use today. If you haven't seen her talk, go find it. Watch it. I’m not always taken with the style of TED talks but "Why do all ambitious women have flat heads?" tilted my world. That title alone demands some attention—what the hell is she talking about? Go watch it and do it fast because I’m about to spoil the punch line. Ambitious women have flat heads, she says, from all the times men walk up, pat you on the head, and say "You should be so proud of what you've built."

Holy shit! That’s exactly what that mother fucker said to me!

Guess who was the only partner to vote against us? He wasn't the one on our side. He wasn’t being nice. He was the one who voted "no". And, of course, he patted me on the head as I left the room.

I'm often asked if it's different for female and male entrepreneurs. Well guys, have you ever been patted on the head? Have you been complimented on your cute dress or nice ass? The typical dismissal of stories like this is that men and women are different, they make different choices[10]. “It’s apples and oranges.” It turns out, though, that I had a nearly unique point of comparison. Augniscient wasn’t my first company, and I wasn’t always me.

Nearly 15 years earlier[11], I co-founded a film company called Hard Drive Productions[12]. “Hard Drive” refers to the emerging nonlinear digital editing tools that we were leveraging for our low budget scifi. Not only did I have absolutely no background or training in film or the entertainment industry, I was at this point homeless. I’d recently flunked out of the University of California at San Diego[13] and started the company as an alternative to killing myself.

It’s important that you understand here that I am not joking. I cofounded Hard Drive Productions as a homeless failure simply to avoid confronting the horrifying reality of just how badly I’d ruined my life[14]. My cofounder, who’d previously produced a failed computer game, was from LA. He viewed film as the highest form of human endeavor[15]. He got to know me from role-playing games — the other thing I was doing to avoid reality — and asked one day if I’d handle the business end of a film company. I loved movies and still do, and I’m a particular fan of Science Fiction and Fantasy. So, when he pitched me the idea of creating a low-budget film from one of my favorite authors, David Brin, I thought, “This will be a fun way to go out.”

We had no reason to think we could do this — no credentials, no experience, no talent — but we put together a limited partnership and optioned a short story from Brin. Most amazingly, I was actually able to raise a few hundred thousand dollars selling the idea of a film with your name in the credits (and your niece in frame). Bialystock and Bloom probably had a better business plan then we did, but the funding commitments came with absurd ease, at least compared with every experience I’ve had in the tech industry.

There were so many differences between Hard Drive Productions and Augniscient, but every one of those differences should have the exact opposite experiences. In 1993 Hard Drive was built on nothing by a talentless college flunk-out selling nothing but the idea of a movie. By 2009, I had a wall full of degrees from the most elite schools, studying neuroscience and machine learning. We had amazing technology demonstrations based on my own published research. Though Augniscient’s business plan was audacious, I look back with a decade’s experience and still believe we were right to try[16]. Augniscient should have been easy to fund, not Hard Drive Productions.

In the end though, there is only one difference that truly mattered. In 1993 I was a man; by 2009 I was a woman.

In 2009 I learned the true definition of avuncular: “Girl, I’m not going to fund your startup."

I have since clearly learned that the kindly, sweet support that comes with a pat on the head is never supported at all. What it truly says is that there is no such thing as a businesswoman and you're excused from the meeting. I know women feel this all the time — in business, science, medicine, politics, and often even at home — but when you feel it for the very first time at 35, after a life of meritless privilege, your world is transformed.

There is growing research on the experiences of female entrepreneurs[17]. We’ll explore it and much more throughout this book. For now, my advice to more “feminine” entrepreneurs is the same as I’d offer to any other: only take (or even seek) funding from those that truly understand both you and your vision. The money should be fuel, not a burden. As a machine learning researcher, where “patterning matching” is a mathematical process, I’ve always been offended by Stanford Business School frat brothers saying that they used “pattern matching” to pick founders. Here’s a clue to their pattern: don’t have long blond hair and a dress; don’t have dark skin; be different from the masses but don’t be different than them.

Bias is corrosive. It’s terrible. It ruins everyone, not just those discriminated against. But bias isn’t done by villains; it’s done by us. All of us. We do not fail because we are biased. Bias, as we will explore, is inescapable. We fail when we refuse to recognize it. We fail when we refuse to change. And perhaps the most profound lesson is that we fail ourselves. There is a tax, and we all pay it. It doesn’t build bridges; it ruins lives. It doesn’t keep us safe; it robs us, all of us, of our future.

[1] In all fairness, I was never charming but remain dangerously naive.

[2] Back in the day the Neural Information Processing Systems conference was known as NIPS and had much more to say about brains. As time passed, though, two things changed. First, the conference metastasized into a feeding frenzy for companies desperately trying to hire AI talent. (Present an interesting poster on reinforcement learning regularization and get a job at Facebook!) Second, the juvenile polysomy of “NIPS” became a punchline to an increasing number of tasteless female anatomy jokes. Given the already tiny numbers of women attending the conference, wine-soaked breast jokes made for an increasingly bad look for everyone involved.

[3] Type in “donkey hysterectomy” and get…whatever the hell it is you’re looking for sicko.

[4] This is not a book about artificial intelligence (AI) per se, and we could quickly get bogged down in definitions and equations. We’ll return to this topic later, because it turns out that AI applied to massive scale data has some profound things to say about bias and inclusion, but for now just know that my work in artificial intelligence is entirely in the field of machine learning, where computer programs learn from experience.

[5] Well fuck you if you don’t like cultural references for 40 years ago.

[6] We are going to spend some serious word count in this book talking about how AI increases bias, as well as how it can be used as a tool to explore it.

[7] I have never pitched a female full partner. That means that the dozens of people over the years who have decided whether to fund my companies have never included a woman. I’ve been successful more often than I deserve, and the industry has improved…slightly, but I’m never happy looking down a conference table full of guys late for the kitesurfing session with their buddies. I find the show Silicon Valley simply too painfully real to watch.

[8] These guys are all pros! None of them would ever invest so frivolously. (If you believe that, you may not enjoy the rest of this book. Stick it out, though; it will be fun!)

[9] Yes, playing with narrative time can be terribly hacky but at least I’m not cutting to “24 hours earlier” as you look down on my corpse in a pool. Besides, I would clearly be Gloria Swanson in my version with the old VC guy ending up my creeptastic Erich von Stroheim.

[10] The spiritual commitment so many devote to the magical power of The Choice has earned an entire chapter in this book.

[11] OK, so I did the flashback in the end anyway. Surprise! I’m a bad writer. Best not to think too much about how many pages you have left to read. And do please keep reading. There are some genuinely valuable ideas lurking in this book, which you need to save from these self-indulgent footnotes and foul-mouthed ramblings.

[12] I swear it wasn’t a porn studio. Yet I must admit, my cofounder would have produced porn in a heartbeat if I’d let him.

[13] Where every young Trident believes they will become a doctor.

[14] A good portion of the rest of this book will explore how I’d reached this point in my life and the rather astonishing events that followed.

[15] Though I was never clear if that meant Truffaut or Troma. San Diego, it turns out, is a fucking weird place.

[16] Besides, the party platform of Silicon Valley is that we build audacious ideas to improve the world. It turns out that it’s not true.