Recombinant Innovation - the Importance of Joining Dots

These days you cannot browse a tech website, read a management magazine or browse your LinkedIn feed without mention of the latest AI-related news, be it the latest Generative AI model, some new demo that looks like magic, or some luminary forecasting doom or a bonanza of riches. As if out of nowhere, artificial intelligence is everywhere. Henry Kissinger is talking about its use in warfare, Bill Gates thinks it can help reduce inequality and offer better healthcare to the poor, the European Commission is sharpening its regulatory pencils, while in the meantime, Elon Musk is warning that machines will take over the world. Looking at the number of searches on Google for the term “AI” over the past five years shows that the launch of ChatGPT triggered an inflection point in interest in AI.

**Google searches for “AI” in the UK 2018-2023**

Artificial Intelligence as a Result of the Collective Brain

So what happened? Why did AI suddenly burst into the public consciousness? It is not as though computer intelligence is a new idea or concept. Just consider HAL, the self-aware, and ultimately murderous computer from 2001: A Space Odyssey, penned by Arthur C. Clark way back in 1968. The answer to this sudden snowballing of innovation can be explained through the theory of the “collective brain”. This posits that just as an individual’s thoughts emerge from neurons interacting as a whole, our can societies and social networks can be modelled as a collective brain. In this analogy, each person is analogous to a node, a neuron, that is interconnected with all the other people in their direct or indirect network, resulting in the emergence of collective, or societal intelligence. In the case of innovation, connections are made through formal and informal networks, through the publication of journals, attendance at conferences, ad-hoc discussions (those famous watercooler moments), writing of blogs, and debate on social networks.

HAL 9000 as imagined by Stanley Kubrick

In a great paper published in the splendidly-named Philosophical Transactions of the Royal Society, Michael Muthukrishna describes [1] how breakthrough innovations are the result of cultural recombination, where different elements of culture are recombined in new ways, giving the appearance of inborn genius. He says that these innovations are born “at the social nexus where previously isolated ideas meet.” Muthukrishana argues that it is no coincidence that great scientific discoveries have often been made by multiple people at around the same time. Newton and Leibniz separately invented calculus, while Charles Darwin and Alfred Wallace came up with remarkably similar theories on the role of natural selection as a driver of evolution. In both cases, the scientific and cultural building blocks were already there, ready to be adapted, evolved or recombined into something new. As many people had access to the same collective intelligence, it is therefore unsurprising that multiple people made the same breakthroughs at similar times in history. This is exactly what is happening in the AI space, where a combination of sophisticated deep learning models across multiple domains (natural language processing, image synthesis), massive training sets and large-scale computing have collectively provided a tipping point in capability.

The iPhone – Result of Steve Jobs’ Genius or Recombinant Innovation?

Look at any number of technology breakthroughs, and you see a similar pattern emerging, with the same invention emerging multiple times at the same time. The motor car, for example, is attributed variably and separately to Karl F Benz, Gottlieb Daimer and George Seldon [2] in the 1880s, but this built on prior innovations such as the carburettor and the spark plugs in the previous two decades. Similarly, there were 22 prior inventors of the incandescent lightbulb before Thomas Edison’s commercial success. As a more recent example, the invention of the smartphone is attributed to Steve Jobs’ undoubted product genius, the epitome of the “great man” theory of innovation. This (gender-biased) theory posits that visionary individuals with unparalleled insights are responsible for sparking major changes.

The theory that the iPhone results from Steve Jobs’s genius however does not stick up to scrutiny. At its inception in 2005-06, Jobs was not keen to become a phone maker and was concerned that it would distract from the iPod, which already had a touchscreen and was a cash cow for Apple. Jonny Ive saw the potential of using multi-touch displays as the primary interface for a mobile phone. This stood in stark contrast to the Blackberry, which was the leading smartphone of the time, and had a full physical QWERTY keyboard.

The building blocks for a revolution in mobile phone technology were coming together from all over the world. Semiconductor companies, including Samsung and Qualcomm, were building high-performance processors based on an architecture by the UK company ARM that was capable of running slimmed-down versions of the Linux operating system that was previously limited to full-fledged computers. In the meantime, the advent of 3G technology made the mobile internet a reality rather than a pipe dream (though the first iPhone made use of 2G), while thin-film transistor technology and multi-touch capacitative displays allowed Jonny Ive to realise his ambition for a fluid touchscreen UI. None of these technologies were invented by Apple, but Apple was the first to package them together in a form factor that persists to this day. As the ‘collective brain’ was available to anyone with the financial and commercial resources to exploit them, Google, who were already working on Android devices, launched their own touchscreen smartphone within one year and the world wasn’t the same again.

First Generation iPhone (2016) and Android Smartphones (2017)

Characteristics of Recombinant Innovation

So what is recombinant innovation? In my view, there are two stand-out considerations. First, it creates new solutions or products by pulling together the capabilities already existing in different, unrelated fields. As described above, the smartphone is the talismanic example of recombination of our age, but look deeply at most recent tech innovations and you will see similar recombinant patterns: smart home (cloud connectivity, low-energy wireless networks, energy-efficient, low-cost IoT processors), social media (ubiquitous high-quality smartphone cameras, graph databases, low-cost cloud storage), ecommerce (web browsers, payment networks, fulfilment networks, secure internet communications) and streaming media (smart devices – TVs, smartphones, speakers, low-cost cloud compute and storage, machine-learning recommenders, low-cost high-speed internet connectivity).

Secondly, recombinant innovations are much more likely to be disruptive than those created by a single function. If we refer to this alternative as incremental innovation, not because it is intrinsically focused on delivering small improvements, but because it optimises or improves existing products. Increasing the resolution of a smart TV from HD to 4K is an incremental innovation, as are the progressive improvements to an electric vehicle’s range. Incremental innovation may be the result of breakthroughs in screen technology, but from a customer perspective, it simply provides a better version of what came before. Disruptive innovations can instead up-end existing markets (e.g. ride-hailing apps, dating apps, streaming media, messaging apps). They are nonlinear because their adoption and performance follow the bottom left of a traditional S Curve. You’d be hard-pressed to find a disruptive innovation that was not the result of recombining many existing solutions.

The Innovation S-Curve. Source LSE

Consider the advent of electric cars. In my opinion, the launch of Tesla’s Model 3 was the key disruption point in the auto market. For the first time, an electric full-sized saloon (or sedan, for US readers) was available for cheaper than the average price of a new car (around $48,000). This marked a tipping point, as for the first time a ‘no-compromises’ family fully-electric car was available for a comparable price as other diesel or petrol vehicles. Tesla brought in numerous technologies from battery providers such as Panasonic, compute units from AMD and Nvidia, as well creating in-house technology (such as the drive units) to produce a vehicle architecture that was significantly simpler and cheaper than that of any incumbent car manufacturer.

I wouldn’t like you to leave this blog post thinking that recombinant innovation will always lead to disruptive outcomes. The opposite is true, as only a minuscule fraction of innovations will be disruptive. Nevertheless, recombinant innovation is highly likely to be more impactful than incremental innovation, be it at a team level or across an entire industry.

Creating a culture for recombinant innovation

Given all this, how can companies and teams set themselves to be best placed to create a culture that encourages the recombination of ideas?

1. Encourage networks within your company – Break down silos

The most obvious means of fostering recombination is to encourage teams to connect with other teams, both within and outside the company. Within a company, organisational silos are the enemy of recombination. Whether the silos are the result of functional specialisation (e.g. different teams creating and operating different parts of the systems – apps, databases, tools etc) or whether they instead arise from having separate business lines with their own P&L lines, siloes inevitably act as barriers towards the type of collaboration and networking that are essential for recombination to thrive.

There are a number of approaches to help teams work across silos. One is to explicitly create multi-functional or multi-disciplinary teams, focused on solving a given outcome. Not only is this team topology more likely to result in a successful outcome, but the greater diversity within teams lowers the cultural barrier between different teams. It will thus be easier to make meaningful connections with people in other teams.

Where companies are organised along functional lines, this often creates a mindset that focuses on getting the best outcome for one’s own team, even at the expense of encouraging new ideas or win-win outcomes. A paper by Edmundson et al. [5] neatly summarises this problem as follows: “You know you should swim farther to catch a bigger fish, but it is a lot easier to swim in your own pond and catch a bunch of small fish.” As an antidote, the authors suggest that functional teams should seek to include people who excel at creating connections. They call these cultural brokers who reach outside their silos, building relationships between teams. These brokers can either act as a go-between or help develop different teams’ abilities to work with each other.

For the rest of the team who are not intrinsically brokers, team members can still benefit from different perspectives by engaging their more limited networks frequently and with purpose. In a Harvard Business Review paper, P. Leonardi describes [6] how when SpaceX engineers were tasked with radically reducing the cost of the Falcon 1 rocket, they engaged with other engineers within the same company, who had different ideas on approaches to cost reduction. By combining and triangulating these approaches, they achieved a per-unit cost of $7m compared to the alternative

2. Focus on the outcome, not the means

One common trap that tech teams fall into is to become more focused on how best to apply their existing tools and knowledge rather than focusing on the nature of the problems to be solved. In many ways, this is a natural human behaviour. Highly-skilled knowledge workers are aware that they provide value through their specialised knowledge so are more likely to look as to how they can apply that knowledge to solve different problems than to look outside their comfort zone. This is, of course, a trap as it easily leads to a place where over time a gap will emerge between the teams’ skills and the company’s objectives. This is never more true than when an industry is undergoing severe disruption. Imagine the predicament of being a photographic film chemist in the early 2000s or consider the impact that electrification and digitalisation are having upon the skills required in the automotive industry. By focusing on outcomes, or by relentlessly focusing and working backwards from the customer’s needs, to use Amazon’s terminology, [7] teams can ensure that they are driven less by what they know today, instead focusing on the knowledge and skills they need to acquire (i.e. have a growth mindset) to achieve their customers’ desired outcomes.

3. Avoid over-specialisation

As technology has become ever more complex and as the breadth of scientific knowledge has become unfathomable, there has been a corresponding tendency for individuals to specialise in ever-narrowing fields of expertise. David Epstein, in his wonderfully sweeping tribute to the under-appreciated importance of generalists in helping navigate today’s complex environments [8], describes how academic publications are hyper-specialised and increasingly narrow. This is particularly true in fields like medicine, where specialisation means that the holistic care of a patient becomes ever more challenging, as practitioners often focus on “small pieces of a larger jigsaw puzzle in isolation.”

Ensuring that your teams contain generalists can go a long way to encouraging the recombination of ideas. Epstein states that many great innovators are systems thinkers (see a previous blog post here), who have an “ability to connect disparate pieces of information from many different sources.” Bill Gates supports this view and attributes Microsoft’s success to the fact that they hired people who had real breadth within their field and across domains [9]. It is therefore important to not only hire for breadth but also to encourage the generalists among you to meander across different areas, bringing their systems thinking approach to solve your trickiest problems. This is tougher than it might appear. For example, Epstein explores the impact of multi-disciplinarity in the publication of research papers, a space for which data is publicly available. Papers that draw on inputs from different areas are less likely to be published in prestigious journals and more likely to be ignored in the short term, though in the long run will accumulate more citations and have a greater impact. It appears that not being able to fit into a pre-conceived category is an obstacle to publishing a research paper.

4. Encourage networks outside your company

Research [10] suggests that it is 30% more costly to successfully discover and utilise ideas created in another firm other than your own. There are numerous frictions that can impede intra-company recombination, including framing contracts that encourage the pooling together of ideas, resolving the thorny issue of who owns what intellectual property concerns, as well as a multitude of strategic considerations such as whether a collaborative relationship can evolve into a competitive one. Open Innovation [11] is the over-arching term which describes how companies make use of external sources of ideas, expertise and technology to complement their internal capabilities. There are a variety of approaches to achieve this. Large companies, such as banks, automotive manufacturers and even government organisations are increasingly engaging with start-ups through Open Innovation Platforms (see here for a list) to help them discover start-ups, engage in crowdsourcing, and find new approaches to the challenges they are facing. Such platforms help reduce the speed, size, technology and cultural disparity that exists between multi-billion dollar automotive behemoths and much smaller and nimble startups, without tying down two parties in relationships based on equity investments.

Open Innovation however only really works if there is a genuine internal commitment to the concept that the space outside the company is as valid and valued a source of innovation of ideas, technology and components as internal teams. This is anathema to many large organisations, for whom “we can do it better” is a source of (often misplaced) pride. The stronger the technical credentials of the organisation, the deeper runs a “not invented here” culture.

5. Make the most of technology building blocks

In the digital space, we are living in a golden age of recombination opportunities. By definition, recombination involves putting together disparate building blocks, as if creating a Lego model. Extending this analogy, the builder’s digital Lego set has never been so big, nor as easily available or cheaply accessible.

Developers have long relied upon open-source software for their undifferentiated software needs. However open source is no longer limited to software. For example, open-source AI models and training data sets are easily available on platforms such as huggingface.co. Governments and companies alike are making data sets available through APIs, either freely or through subscription, with data sources as diverse as geolocation, satellite, demographic, healthcare and environmental data. Governments are beginning to regulate some sectors to make data available in order to foster greater competition. In the UK, the competition authorities have mandated that the largest banks implement open standards for Open Banking to allow customers to securely share their data with other financial providers. Software-as-service (SaaS) which provides technology as a subscription service, from anything from payments to communications services, is now an industry generating in excess of $250 billion / year. The big three cloud providers (AWS, Google and Microsoft) all offer extremely extensive software and platform-as-a-service offerings to builders. When taken together, all these modules, SDKs, models, services and so on, greatly reduce the friction of recombination and will likely continue to be a driver of disruption of innovation.

Conclusion

Hopefully, this rather meandering post has articulated why I believe that combining different ideas and technologies lies at the heart of innovation and why siloed thinking is the enemy of creative innovation. I have built on an earlier post on cognitive diversity, but I have only scratched the surface of this topic. I have deliberately steered clear of debates about when innovation is truly disruptive, and have only skimmed the surface of how to reduce intra-organisational silos or facilitate open innovation. Nevertheless, I hope it has given some food for thought, and it will certainly act as a springboard for further related musings, at least when I next get time to put pen to paper.

Recombinant Innovation – the Importance of Joining Dots