Applied AI

With all the major tech companies open-sourcing their latest machine learning algorithms and building SaaS infrastructure around it, we’re entering the age of commoditisation. Even the (laudable) attempts to build open source data sets for common machine learning tasks could threaten any data advantage a startup may have. So what does that mean for startups?

I participated in a panel debate at Capital Enterprise’s CASTS launch last Thursday looking at the AI scene in London. One entrepreneur’s question deserved a fuller answer than the one I gave at the time. It was about how to answer VCs wanting to know what IP assets her company had.

Having spent some time investing for a fund that required patentable innovation when most other VCs were focused on eCommerce, social, marketplace, and SaaS businesses, it is odd to see the same VCs demand protectable IP from today’s AI-enabled businesses.

In particular, my colleague, Sebastian Spiegler (@SRSpiegler), draws an analogy with the noughties when the explosion in web-enabled businesses was built on the availability of cheap distributed computing and front-end platforms. Many valuable businesses were built without any proprietary technology. Instead they relied on domain knowledge that conferred first-mover advantage (and executed like crazy). Likewise, those using commoditised machine learning infrastructure and algorithms can apply their own vertical/market expertise to create valuable products and services today.

These applied approaches look like a smart move to me. With so much money being invested by the gorillas to ensure commoditisation, doing a horizontal play must bring additional risk (as pointed out by Matt Turk’s piece on Building an AI Startup). These platforms and tools level the playing field and enable valuable new products and services to be built upon them. We (@ProjectJunoAI) didn’t have enough space to look at all the vertical-focused machine intelligence companies in our recent AI-landscape work. We will look in more detail at these verticals soon.

John Spindler (CEO of Capital Enterprise) also rightly pointed out the complexity involved in knitting together solutions and the non-trivial implementation headaches of building for resilience, user experience and scale. These things may not be patentable, but they still form an effective barrier to entry. Vertical-focused AI-enabled businesses should have the considerable advantage of being able to clearly define what their value proposition is too!

The age of machine learning commoditisation is also the age of applied AI.