At Lux, we help innovators find the right technologies and companies to bet on. This world has all kinds of tools and frameworks: gate processes, innovation funnels, idea managers, databases, and more. In recent years, I’ve had the sneaking suspicion that as we continue to add these tools, we are simply adding to the list and bloating the process, rather that replacing old methods with newer, and more effective ones. Take data, for instance: There is a ton of “innovation” data out there – news, investments, patents, papers, and more – and new platforms to serve up this data to you pop up seemingly every day. Sourcing the data isn’t the real problem; unifying and making the most of the data to identify weak signals and make the appropriate bets is.
This is why I am excited about artificial intelligence (AI) and machine learning (ML).
Few need convincing that AI/ML is disruptive. I just punched “AI and innovation” into Google, and on the first page alone I found trillion-dollar forecasts, thought pieces, book chapters, and events dedicated to the topic. Having advised clients on the impact of AI on their businesses, I’ve found that most of the public discussion has been severely lacking in real uses cases that people can act on. As I’m tasked with improving Lux’s core product and have access to a talented data science and software team, I’ve been fortunate to apply AI to real problems that my colleagues and I face in supporting our clients day to day.
As the bulk of the innovation data available is unstructured text, the natural tool set to explore is natural language processing (NLP). This broad bucket of analytical techniques underpins some ubiquitous platforms like Amazon Alexa or even Microsoft spell check. There have been tremendous advances in this field in a very short time, and any talented data scientist/programmer can access numerous libraries to try out the latest and greatest algorithms. One of our tools of choice within NLP is topic modeling. To understand how topic modeling works, I’ve included the handy illustration below.
Put simply, topic modeling takes a big bag of words and can tell you what recurring topics are present, with a topic defined by a collection of keywords. I’m not going to get into the nitty-gritty here, but suffice to say that there are lots of nuances here – what algorithm to use, what the right number of topics is, and how to label them. In a recent white paper I wrote, I walked through the challenges innovators face and connected that to key use cases that topic modeling can address. The four areas I identified were:
- Landscaping a technology/application
- Finding documents with similar concepts as a target document
- Competitive analysis
- Weak signal detection
Remember my earlier point – I’m willing to add on a new tool IF it helps me unify data and/or bubble up key concepts in a sea of data. All these use cases involve a ton of data – patents, papers, web sites, news, even Twitter potentially. In the white paper I referenced, I walked through landscaping 2D materials patents (think graphene and relatives) to show how topic modeling helps surface key topics where there is strong activity and pull out the key players working on those topics. Again, the point of AI/ML here is not that I couldn’t do this before, but that it would have been very time-consuming and potentially biased by whoever was leading the exercise.
Here I make no assumption as to what the key areas of innovation are – the model tells me what it sees. This is just one example of how we can use AI in the innovation process to create value, not bloat. If nothing else, the next time an executive asks you to start an AI project, you’ll have a tangible problem to solve rather than hurling some AI at the wall to see what sticks.
For more information download my white paper on applying AI and machine learning to the front end of innovation, or check out my presentation on weak signal detection, and email us your questions.
- Report: Lux Executive Summit Global 2019: Digital Transformation of X (Members Only)
- Press Release: Despite $6.1 Billion in Funding, China Not Leading in AI Innovation