Does AI make strong tech companies stronger?
First, though you need a lot of data for machine learning, the data you use is very specific to the problem that you’re trying to solve. GE has lots of telemetry data from gas turbines, Google has lots of search data, and Amex has lots of credit card fraud data. You can’t use the turbine data as examples to spot fraudulent transactions, and you can’t use web searches to spot gas turbines that are about to fail. That is, ML is a generalizable technology – you can use it for fraud detection or face recognition – but applications that you build with it are not generalized. Each thing you build can only do one thing. This is much the same as all previous waves of automation: just as a washing machine can only wash clothes and not wash dishes or cook a meal, and a chess program cannot do your taxes, a machine learning translation system cannot recognise cats. Both the applications you build and the data sets you need are very specific to the task that you’re trying to solve (though again, this is a moving target and there is research to try to make learning more transferable across different data sets).
So: as an industrial company, do you keep your own data and build the ML systems to analyse it (or pay a contractor do do this for you)? Do you buy a finished product from a vendor that’s already trained on other people’s data? Do you co-mingle your data into that, or into the training derived from it? Does the vendor even need your data or do they already have enough? The answer will be different in different parts of your business, in different industries and for different use cases.
This takes me to a metaphor I’ve used elsewhere – we should compare machine learning to SQL. It’s an important building block that allowed new and important things, and will be part of everything. If you don’t use it and your competitors do, you will fall behind. Some people will create entirely new companies with this – part of Wal-Mart’s success came from using databases to manage inventory and logistics more efficiently. But today, if you started a retailer and said “…and we’re going to use databases”, that would not make you different or interesting – SQL became part of everything and then disappeared. The same will happen with machine learning.
One giant step for a chess-playing machine
Most unnerving was that AlphaZero seemed to express insight. It played like no computer ever has, intuitively and beautifully, with a romantic, attacking style. It played gambits and took risks. In some games it paralyzed Stockfish and toyed with it. While conducting its attack in Game 10, AlphaZero retreated its queen back into the corner of the board on its own side, far from Stockfish’s king, not normally where an attacking queen should be placed.
Yet this peculiar retreat was venomous: No matter how Stockfish replied, it was doomed. It was almost as if AlphaZero was waiting for Stockfish to realize, after billions of brutish calculations, how hopeless its position truly was, so that the beast could relax and expire peacefully, like a vanquished bull before a matador. Grandmasters had never seen anything like it. AlphaZero had the finesse of a virtuoso and the power of a machine. It was humankind’s first glimpse of an awesome new kind of intelligence.
Tellingly, AlphaZero won by thinking smarter, not faster; it examined only 60 thousand positions a second, compared to 60 million for Stockfish. It was wiser, knowing what to think about and what to ignore. By discovering the principles of chess on its own, AlphaZero developed a style of play that “reflects the truth” about the game rather than “the priorities and prejudices of programmers,” Mr. Kasparov wrote in a commentary accompanying the Science article.
What is frustrating about machine learning, however, is that the algorithms can’t articulate what they’re thinking. We don’t know why they work, so we don’t know if they can be trusted. AlphaZero gives every appearance of having discovered some important principles about chess, but it can’t share that understanding with us. Not yet, at least. As human beings, we want more than answers. We want insight. This is going to be a source of tension in our interactions with computers from now on.
Maybe eventually our lack of insight would no longer bother us. After all, AlphaInfinity could cure all our diseases, solve all our scientific problems and make all our other intellectual trains run on time. We did pretty well without much insight for the first 300,000 years or so of our existence as Homo sapiens. And we’ll have no shortage of memory: we will recall with pride the golden era of human insight, this glorious interlude, a few thousand years long, between our uncomprehending past and our incomprehensible future.
Evaluating early stage startups — The three metrics that matter
Defining “fast growth” depends on stage, but for early (Seed or Series A), growing 100% YoY is typically pretty solid. Paul Graham (PG) famously looks for 5–7% weekly growth for companies in Y Combinator, and his rationale is pretty simple: “a company that grows at 1% a week will grow 1.7x a year, whereas a company that grows at 5% a week will grow 12.6x.” When you consider the compounding effects of this growth, it means a company starting with $1,000 in revenue and growing at 1% will be at $7,900 per month four years later, whereas the company growing 5% per week will be bringing in more than $25 million per month.
The founder’s guide to understanding investors
When we dig deeper, the degree to which early-stage investing is a grand slam business is shocking. First, amongst early stage investors, the returns are disproportionately distributed. The Kauffman Foundation, an investor in many VC funds, found the top 20 VC firms (~3% of VC firms), generate 95% of all venture returns. Second, outside of the top 20 VC firms, most lose money! A study found the top 29 VC firms made a profit of $64B on $21B invested, while the rest of the VC universe lost $75B on $160B invested.
As early-stage investing operates on a power law, Paul Graham (founder of Y Combinator) mentions “You [as an investor] have to ignore the elephant in front of you, the likelihood they’ll [the startup] succeed, and focus instead on the separate and almost invisibly intangible question of whether they’ll succeed really big.” He highlights there are 10,000x variations (!) in startup investing returns, meaning top investors must have the mindset of willing to strike out in order to hit grand slams.
There needs to be room for your startup to capture a large share of this market. Elad Gil (early investor in Airbnb, Coinbase, Gusto, Instacart, Stripe), explains this means i) the market is structurally set up to support multiple winners, but ii) if the market only supports one winner and customers are currently not served well – there is an opportunity to dominate the market.
At the Series A stage, investors are mainly looking to see if PMF is achieved. This evaluation can be qualitative – Marc Andreessen (co-founder of Netscape and Andreessen Horowitz, an early investor in Facebook, Twitter, Wealthfront, Slack) notes, on the inside, “you can always feel product/market fit when it’s happening. The customers are buying the product just as fast as you can make it — or usage is growing just as fast as you can add more servers. Money from customers is piling up in your company checking account. You’re hiring sales and customer support staff as fast as you can. Reporters are calling…”
When investors are evaluating for PMF, Rachleff notes that the best test is to see if the product is growing exponentially with no marketing, meaning the product is so good it grows through word of mouth. Top investors often don’t want to see marketing spend because it shows care for vanity metrics (things that don’t matter) rather than building an amazing product that people engage with (which does matter).
Not all buzzwords will fulfill their potential and result in a disruptive technology shift though. As a founder, you can reduce this risk by avoid starting a startup on that shift until the technology adoption is growing quickly and reaches a multi-hour per day level of usage. Sam Altman expands, “It’s very hard to differentiate between fake trends and real trends…If you think hard and you really pay attention, sometimes you can. The metric I use to differentiate between a real trend and a fake trend is similar to loving a product. It’s when there is a new platform that people are using many hours every day.”
To believe the startup can fulfill grand slam potential, investors want to see the startup has verified their assumptions on how users find the product in a repeatable and scalable manner. This is also called a go-to-market strategy (GTM).
Bill Gurley (major early investor in Uber, Stitch Fix, Zillow, etc.) called a unique GTM the most under-appreciated part about startups. It’s not about who did it first, but who did it right. Gurley looks to see if the startup has two things: (1) An interesting way to get into the market; (2) A way to establish themselves once in the market. The word ‘unique’ is important here. Replicating existing GTM strategies is often too costly because incumbents have already dried up the channel(s) to market and sell to customers. As a founder, you need to find a unique GTM that is repeatable and scalable. The good news here is that if you succeed, you’ll be able to keep out competitors by saturating the new channels.
Andy Rachleff has a second perspective on how startups can avoid competition. With his adaptation to Clayton Christensen’s (Harvard Business School Professor) disruption theory, startups can compete with reduced competition in either two ways. They can compete via new-market disruption – targeting a new set of users and competing on different characteristics (e.g. instead of price, focus on experience) than competitors, or they can compete via low-end disruption – targeting the same set of users as incumbents, but offering a greatly reduced product at a lower price point.
Along with the above quote, Bill Gurley tests if executives at the startup have a notion of insane curiosity – constantly learning new ways to win. To evaluate this, he asks questions on what information (e.g. books, podcasts) executives learn from, how they engage with it, and then probes if they are trying to use that information to majorly improve themselves or their business.
Curious folks tinker. Obsessively curious folks solve the hardest problems that require endless tinkering. If you are obsessively curious and fail with your original plan, odds are you will use your learnings and pivot into a big market that loves their product.
If a founder is obsessively curious, they can navigate the idea maze. By running a founder through the idea maze, investors evaluate if the founder understands all permutations of their idea, why their plan is superior to all other competitors, and which turns to lead to treasure versus which ones lead to certain death. It’s important for a founder to thoroughly know their idea maze, it can save years by not going down the wrong path, in addition to convincing investors you know can be a grand slam.