Beyond the Mania and the Winter — Another Way to Think About Innovation

Dug up a note I wrote ten years ago. Felt like it still had something to it — dusting it off.

Written 2016.09.11

Foreword

Before 2015, capital markets were on fire and mass entrepreneurship was a movement. The past year has brought the opposite: cautious investing and a lot of dead companies. People started shouting that the capital winter had arrived, then sat down to lick their wounds. As it happens, I’m one of those failed founders — my first company also went under during this “winter.” But that failure had next to nothing to do with capital; the timing was just unlucky. The reasons are another story for another piece. With the startup wrapped up and a bit of breathing room to reflect, I decided to write down some thoughts on innovation, hoping to spark a discussion. As an outsider to the VC world, what follows is highly subjective. If anyone is offended, please bear with me; if I get something wrong, please correct me.

The Three Stages of a Major Technological Wave

Based on what I understand from the social history of science, an industry-wide (or society-wide) wave of innovation usually moves through three stages:

Breakthroughs in core theory and the buildout of infrastructure ↓ Development of foundational tools and application modules ↓ Extension into business models and use cases

These waves are made of countless individual innovations, all tightly intertwined. Often the boundaries between stages aren’t clean — the specific innovations depend on each other and propel each other forward.

Take the internet and mobile internet:

Before 1990 we were mostly in the first stage, driven by academia, the military, and research institutions. Most of the work concentrated on the theory of network transmission and organization, on algorithms and protocols, and on the underlying hardware. By the early 90s, when the WWW opened to the public and consumer infrastructure started getting laid down, the first stage gradually wound to a close.

The second stage spanned the next decade or so, driven by technologists and a handful of forward-looking commercial organizations. They built out the foundational modules — the building blocks — that the industry and its upstream and downstream players would eventually depend on: web technology, a complete email system, languages and frameworks for internet development (PHP, AJAX), publishing tools (Blog), online payment methods (PayPal), streaming media (adult sites, YouTube, Skype), cloud platforms (AWS), networking equipment (routers), end-user devices (phones that could connect). Entering this stage, the commercial value of the new technology was getting clearer, but the technical bar was still high; capital usually drove things forward and engineers did the building. The first version of any foundational module was usually built inside a specific use case, which meant those modules also doubled as the first applications. Because they were highly portable and broadly needed, they got abstracted out into general-purpose components and pointed the way for everyone who came after — in technology, methodology, and standards. At this point the industry started to grow fast.

After the dot-com bust in the early 2000s, the third stage of the internet came into focus. I like to call it the “Lego phase,” driven mainly by capital and entrepreneurs with business instincts. Because the foundational toolkit was complete, the technical bar had dropped sharply (you could build a basic, decent-looking site or app in less than a day) and visible profit opportunities had multiplied. Lots of non-specialists started to imagine, applying these tools to existing industries — reshaping how organizations work, or making processes more efficient (business-model innovation): Airbnb applied the web and online payments to short-term rentals; Facebook used the web to connect people; bilibili used streaming to build a UGC anime community; new internet companies started running their servers on cloud providers.

From around 2008, mobile internet had largely finished its second stage and entered its third. Concepts that weren’t even particularly novel — O2O, P2P, IP-driven plays, SaaS — became the talk of the town. The innovations at this stage were mostly “business-model” innovations; the core competency was no longer the technology itself, but the understanding of (and process changes inside) the traditional industries the tech was being injected into (“internet plus”), or the exploration of new market possibilities the technology had unlocked (IoT).

The whole third stage of the internet peaked in the first half of the 2010s. Public participation was extreme (because the bar was low); capital warfare turned white-hot (capital became the new bar); everyone was shouting about subsidies, market capture, “don’t miss the window” (winner-take-all). In my view, vast imaginative space combined with a low bar inevitably produces extreme social enthusiasm, extreme failure rates, and extreme resource waste. Third-stage innovations get even harder to judge when they aren’t anchored to efficiency or productivity gains. By that point, the innovation space around a given opportunity is more or less exhausted, and the social impact is close to maxed out.

Capital investment usually gets its best ROI in the second stage; overall social returns get maximized in the third.

In the second stage you tend to see: participants with “hard skills” (high-quality talent), high product-development difficulty (high bar, less competition), tools that meaningfully improve some process’s efficiency or productivity when they succeed (clear value), foundational tools that are broadly applicable and continue to evolve through the second and third stages (long monetization tail), and decision factors for success that center on feasibility, usability, and performance (relatively easy to judge). The catch: because the work is hard to grasp, innovation led by pure technologists often lacks imagination — and the imaginative, well-resourced capital crowd often can’t yet understand it well enough to participate. The output of this stage also doesn’t usually have everyday large-scale use cases yet, so ordinary people may not feel its impact directly.

In the third stage you tend to see: success stories with extremely broad reach (huge imaginative space, also a source of bias), winners who succeed because of unique understanding of a specific industry (hard to estimate in advance), low participation bar and vague requirements producing a flood of projects, competition that turns into a race for capital and market position. The whole thing is part imagination feast in full bloom, part casino full of mixed quality. The decision factors for business-model success are relatively complex — user habits, human nature and group behavior, UX, brand and marketing, monetization model, all sorts. To cope with this, capital markets have produced a dazzling array of micro-level theories trying to make sense of the chaos — admirably so. Setting aside whether overall capital returns are good, this stage is the one most able to ignite mass enthusiasm for innovation, and the output is typically usable directly by society. We should also see that this kind of boom is usually driven by the fervor of capital and founders together. Behind the fervor: a fear of missing out, because by this point innovation has been widely understood, anyone can riff on it for a minute, and everyone is afraid of missing the “wind” that has actually been quietly blowing for years.

Where We Stand

Mobile internet’s third stage is winding down. Narrow AI is charging through the second stage. Some branches of life sciences are seeing the dawn of the first stage’s victory. VR/AR? My instinct is that it’s more like display technology when it first appeared — it’ll disrupt some industries, but its spillover value won’t match the others above.

AI has been through 60+ years of highs and lows. Its theoretical core, neural networks (machine learning), was already in place in the 1980s and showed promise — but couldn’t be made commercial, and got shelved. In recent years, hardware progress and the rise in compute have made the theory practical. Combined with high-profile demonstrations from the big players (IBM Watson on Jeopardy!, AlphaGo on Go), public interest came back, and the first stage’s infrastructure work finally wrapped up.

Narrow AI is now in its foundational-tools stage: technologists are racing to productize AI algorithm libraries, image-recognition tools, NLP, pattern recognition frameworks, autonomous robot control, big-data analytics services, and so on. Looking ahead to AI’s third stage, you can let your imagination run: chatbots become believable virtual employees that retailers pay to deploy, replacing human service reps; image recognition becomes an API that exporters use to inspect container cargo and hospitals use to read medical imagery; NLP, logic engines, pattern recognition, and big-data analytics combine to produce a wave of B2B service providers offering live, data-grounded advisory and decision support to enterprise customers; autonomous control gets better at driving vehicles, surgical instruments, and household work… By that point, narrow AI is a cheap resource like water, electricity, or information, seeping into every existing industry and triggering its own third stage of change.

More possibilities in life sciences are surfacing too: gene-editing experiments are stacking up wins; tissue induction and regeneration, lab-grown organs are entering clinical trials; AI is showing potential to accelerate new-drug discovery; simple brain-computer interfaces have already been demonstrated… These foundational theories getting validated, and the infrastructure starting to take shape, opens new possibilities for humanity: the regeneration and enhancement of life. Boldly extrapolating to life sciences’ second stage: technologists will start building gene-editing tools that non-specialists can operate; successive generations of 3D organ-printing devices ship; bandages and sutures get replaced with regeneration-inducing materials; early drug synthesis and screening get fully handed off to reaction simulators and AI running 100x faster; libraries of human structural models and gene dictionaries get filled out; committees start drafting standards for brain-computer interfaces…

A Few Words to “Angels”

Most angel investors have already accumulated meaningful personal wealth. They aren’t short of opportunities to make more. So if making money is all they want, there are probably better paths than being an angel. The real angels usually have ambitions beyond the money. In my view, an angel who wants to push society forward wouldn’t walk away from spotting frontier technologies and helping build foundational applications just to fight it out in the well-known third stage. An angel who wants to back the next generation wouldn’t overlook the quiet, grounded pioneers of new fields just to spend their time figuring out which glamorous pitches are real. The reason “angels” are called angels is that they fly high enough to see far, point the way forward, and offer hope.

Closing

Most innovation happens in a sequence that’s both order-dependent and mutually influential. An innovation is always evolving and improving in interaction with other innovations; today’s innovation shapes tomorrow’s available resources, the maturity of upstream and downstream, cost, technical feasibility, social and cultural context, regulation, user acceptance, and more. Analyzing innovation in isolation produces fallacy. At the level of a specific company things get even more complex — a business may straddle all three stages, or be a mix of innovations sitting in different stages. The framework above is just a heavily simplified one for thinking; applying it directly would be naive. The examples are entirely my personal view; don’t take them on faith — I just hope to spark a little discussion.

Capital’s mania and winter are surface phenomena. Behind them are just ordinary economic patterns and historical experience. In the world of innovation there is no winter. There are only opportunists trying to read the weather.