‘We can get you the capital by end of day’ was just one of the many emails that the Rocket AI team had received after unveiling a website about their new startup. Launching in 2017, Rocket AI claimed to be the new global leader in neurologically-inspired applied machine learning, building their systems around a ‘patent-pending technology’ called Temporally Recurrent Optimal Learning™. Since the launch party the night before, Rocket AI had taken over the internet, with millions of web impressions, over two hundred job applicants and five Silicon Valley investment funds vying to invest. It was every young startup’s dream. The only problem was, Rocket AI wasn’t real.
Conjured up as a birthday party theme by a group of friends the night before, Rocket AI was the brainchild of a tipsy conservation held during DeepMind’s annual party at the popular AI conference NIPS (Neural Information Processing Systems). The website RocketAi.org was made over breakfast the following morning, with invites mailed out to a surprise launch party that resulted in over 200 attendees only curtailed by the police at 3am.
It was only when major tech news started tweeting about the ‘biggest new AI company’ that we realized that people hadn’t worked out the acronym of Temporally Recurrent Optimal Learning spelt out our aim, which was, quite literally, to ‘trol(l)’. In the final hours before the party, we had gathered together to discuss whether a prank like this might cost us our careers. We weren’t sure, but after some deliberation, decided to go ahead with it anyway.
By 10pm the mansion was full to the brim with a queue of people trying to get in. Dr Anders Sandberg, from Oxford’s Future of Humanity Institute toasted the crowd at midnight with an absolutely brilliant totally improvised five minute speech about why TROL was the breakthrough the AI field needed. There were rumors of RocketAI having raised $80m in seed funding. One academic slipped his card into one of my colleague’s pockets whilst whispering in his ear ‘hire me’. People came to me telling me they were already confirmed to invest, asking if I wanted to invest to. I wasn’t sure who was running the joke anymore. It was confusing, amazing, shocking and brilliant.
Even after the NIPS conference week had finished, people were taking to Quora to ask what else had happened during NIPS apart from Rocket AI. The hottest name on everyone’s lips and tweets was #RocketAI. AI researchers continued on the joke, even after we raced to beat aggressive reporters by explaining our prank in a Medium post. It was too good to give up, too good a reflection on the parody and hype that had filled the AI industry. Everyone was laughing at themselves and it was healthy.
By no means was anyone surprised that people had taken RocketAI so seriously, enough to offer us ludicrous amounts of capital. We had all seen similarly bullshit companies play-off hype and go on to raise millions. Our team will always remember being brought in by another fund to do due-diligence on a ‘Natural Language Processing’ company raising $2.5m. The start-up, run by a former sales guy, told us straight-faced that he had discovered the ‘holy grail’ and ‘solved NLP’. He had apparently found a new way to ‘parse text directly to meaning’ from a dusty book in an old academic library ‘that he couldn’t tell us about’. At one point, I had to leave the room because I had such uncontrollable giggles that I thought I was going to explode. The people who thought RocketAI was so obviously a joke clearly hadn’t spent enough time working in a hyped tech field.
That ‘NLP’ company raised all their money.
Amongst the many problems projected onto Silicon Valley and the tech industry, the blur between real and fake is too often overlooked. Much less commented about than gender issues, or the uncontrolled power of big tech players and how they control our data. And yet, shouldn’t the possibility of a lack of stable genuine reality be of a higher concern? RocketAI, in all its good humor, horrified a bunch of us, concerned at how easy we knew to play the game to make perception into reality. Sitting in conference calls, feeling slightly confused, whilst big name lawyers and investors said to us “Well, you should just launch RocketAI as a real company off the back of all the attention you guys have.” And someone did, in Berlin, launch a ‘Real’ RocketAI explicitly to do exactly that. One well-known billionaire hedge fund manager reached out to us directly, telling us that our manipulation of reality demonstrated that we would be good with financial markets. Nothing was real, he told us, just people’s perception.
RocketAI started as a joke that ended up giving us deep insight into the social narratives that drive business. We thought the prank would cost us our careers, but instead it opened us up to a whole new sector of people who had not only saw perception as reality, but had capitalized off it. We unlocked a master key to a whole new world.
Even as I write this, I myself face an internal conflict with the necessary counterargument. But of course narratives exist, that’s why PR companies exist and press releases and even Twitter. Only the worst sort of naive person would think that the biggest drive of action is a tie to fundamental integrity and not a drive for capital and power. The negative connotations of ‘capital and ‘power’ aren’t meant to be inferred here, in a capitalist society they are our biggest drives. Of course, one could argue, we should expect any reality to be multi-faceted, many-layered, with each perspective only visible to a select few. Your vantage point depends on where you fall in market dynamics and how the overall system eventually benefits from your contribution.
We all already know that we each individually live in a limited facade of reality but it's not always conscious or explicit. Perhaps it’s too horrifying to consider how little we know- that our unknown unknowns drown out our factual truths a billion-fold. We hold onto the hubristic idea that our model of the world is complete and that our actions are logical steps within a well-ordered world. A world where you work hard and achieve goals in a linear progression. Career ladders, relationship benchmarks, benchmarks in progress, even time itself. The human condition desires order over chaos, negentropy over entropy. We want to believe in the domination of chaos, denying the laws of the universe to conjure up our arguments and beliefs. We want to believe we are full-fledged autonomous and ethical agents within structured models of reality.
We want to believe that there’s order over chaos, despite our perception of order being only a sub-branch of larger more complex models. Some have more access than others and in that lies a narrow definition of power. The strategically powerful will keep their access implicit in order to not offend those existing on the lower tiers. The best thing for the lower-tiered to believe is that they have full access and are free, so that they work diligently and enthusiastically.
A designer takes a start-up, where the CEO sells to him that his designs will have radical impact on X industry and that the designer is fulfilling their career goals. The CEO believes they are at the top of the model chain, designing the narratives to fulfill those below, when consciously or unconsciously, the CEO is fulfilling the trends and market dynamics of venture capital to please investors or customers to keep his company going. Each customer and investor believes that they are the top of the pyramid, handing out money to whoever they please. One could reduce this hubris to the art of collaboration within a system, but in reality it’s a delusion. The collaboration is a public narrative, whilst in reality everyone believes they are masters of their own model. And of course, at every level, everyone is an ‘expert’.
The most horrible truth is that we are all Rocket AI.
A few months after Rocket AI in the spring of 2017, a few of us sat round a dinner table and compared notes on the lessons we had each learnt. It had become a much deeper philosophical experience than any of us had expected. Personally, I looked back at my 20s and realized that I had had no idea what I was talking about. I’d only ever had a limited facade of an insight into anything. I was ashamed and embarrassed that anyone had ever called me an expert in any domain, and I began to wince when other people either deliberately or passively allowed the same.
Valid experts often don’t let on that they have access to a model that you don’t, because they capitalize on you not knowing. The people who parade their ‘expert’ status may believe they have some greater access information than most, but more often they implicitly think that by waving the expert flag they can capitalize off other people who believe that they are an expert, rather than it ever being true. I started to see things differently. RocketAI had taught some hard lessons. I got a custom neon sign for my apartment that said Everything is A Scam.
The pillars of status and expertise start crumbling when you start to see how incomplete people’s models are, of themselves, the people around them, the world, when you realize that everybody is making it up as they go along. And within this crushing reality comes fear, the loss of a hope of some sort of parental ‘elite’ that will keep you and the rest of the world safe (2017 was great at learning that). I'm horrified that there isn’t some incredible smart and powerful group with a secret world plan. That actually the destiny of things falls directly at your feet, that your contributions shape reality and that you have a responsibility for the world you want to see. That’s horrifying, akin to being ripped out of a womb too early and asked to solve all the greatest challenges. In that moment of mental nakedness, that dark moment when you realize that there isn’t anyone coming to save you, we panic. We look for any narrative we can find to fill the void. A potential causation for why we get more spiritual or religious with age is that we need some perceived sense of order to stay sane.
The flipside of there being no ‘real’ order is unbridled opportunity. Once you perceive that everybody’s making it up as they go along, life becomes a video game. A quest in playing different roles well. A game in working out the mechanics of credibility and using them to your advantage across any and all disciplines. Academics criticize Elon Musk for not having enough technical depth in the industries he disrupts because he offends their model of expertise. What they don’t realize is that Elon Musk is playing a different game. He no doubt realizes that nobody knows what they are doing and thus is able to go into a field and build a sufficient enough model of everybody’s inaction in order to revolutionize it. The fact that he can do it, successfully, demonstrates the opportunity from this as a realization. One could say a similar belief seems to tie the successes of all the so-called PayPal Mafia. They all went on to dominate radically different fields. At some point, I wonder if they collectively realized that nothing was real, that anybody could do anything and then...went their separate ways and capitalized off all of it.