Machine Learning 201: Exploring the Market for New Business


2016 will be remembered for many things, not least as the year that AI and machine learning stepped into the technological and cultural spotlight. Advances in applied machine learning fueled our enthusiasm for smarter, more talkative devices. It impacted how we learned about and processed the news—and fake news—of the national elections. Automation grew in sophistication: detecting financial fraud, improving healthcare, deepening business intelligence, and disrupting human work models.

While significant flux continues, the fact is that machine learning is now firmly established in business technology. Some obvious big data categories are already dominated, but you need only ingenuity to start putting machine learning insights to new and quantifiable uses. The business market for machine learning is in an exciting growth phase, and it’s definitely not too soon to find your place in it.

What kind of ML opportunity should your business pursue? It helps to start with the current state of the industry. ML products and services have self-organized into categories that are roughly analogous to the technology stack of modern cloud computing.

Gorillas in the ML mist

Leading technology companies—Google, Microsoft, Facebook, Amazon, Apple, Baidu—envision a commercial future underpinned by machine intelligence. In the past year, we’ve seen almost weekly news about these 800-pound gorillas of industry actively pursuing alpha status in the machine learning/artificial intelligence category.

Controlling data moats. So much of the competitive advantage in the machine learning field comes from the data to create models. The gorilla companies have existing “moats” of users whose data profiles are global and varied across consumer and business contexts. These data moats are ginormous and generate massive, steady amounts of data for driving insights. Wider access to datasets could dry up the business value of a moat, but the gorillas are cannily open-sourcing their low-value commoditized data to encourage innovation in machine learning that can lead them to new sources of higher value data.

Centralizing skills. The trend among the gorillas to incubate AI research serves the twin goals of futurist learning and business development. By investing significant resources in creating and nurturing centralized AI skills in-house, the gorilla companies reap long-term benefits in machine learning sophistication, and they build up their bench of science and technology talent in this increasingly competitive field.

Facebook and Google kicked off the brain trust trend by organizing their artificial intelligence teams under famed data scientists Yann LeCun and Geoffrey Hinton, respectively. Salesforce and Apple followed with AI units built from platform acquisitions: Salesforce purchased MetaMind and PredictionIO in 2015, and Apple acquired Turi in 2016. Microsoft has strong machine learning talent across product units and Microsoft Research; in September, the company announced a 5,000-person reorg under Harry Shum.

Centralized AI teams also provide a fringe benefit in the form of PR for potential hires. Terms like “artificial intelligence” are buzzy, but the fact is that machine learning is still an emergent discipline—and it’s developing faster in enterprise think tanks than in universities. The industry as a whole struggles to find enough qualified machine learning candidates, and the gorillas’ public commitments to research generate a lot of brand goodwill among talented academics and autodidacts.

Fueling higher education. Funding for academic programs in AI ebbs and flows with the times: When some technological breakthrough makes AI look promising again, schools ramp up. We’re in one of those boom times now. Since lack of academic opportunity led many AI pioneers to choose industry as the more viable (not to mention lucrative) option, now the top industry researchers are contributing tremendously to AI scholarship, and many of them are teaching on the side. A great example is Baidu’s and Stanford’s Andrew Ng, whose most popular course, CS229 Machine Learning, is available in both traditional classrooms and online through Coursera.

The future of business imagination

If the 800-pound gorillas are holding big data moats and machine learning talent, where does that leave entrepreneurs in the machine learning marketplace? Is it possible to find enough talent to innovate independently? Can a new or pivoting company carve out an ML niche using democratized algorithms, or a combination of commoditized and proprietary or licensed data?

The short answer is yes. Machine learning platforms and platforms-as-a-service are seeing dynamic change right now. Some new companies are offering managed machine learning, from completely black box solutions to fully configurable algorithms. Others are finding value working in the entire range of machine learning capabilities (algorithm development and training, data services, models, and deployment).

Because access to and control of vast data moats is often out of their reach, small or new AI/ML ventures generally focus on unique machine learning services and exclusive or very specialized data to set themselves apart. For example, a startup can prospect for opportunity by mining data “exhaust” (the detritus data or patterns that may be found in a customer’s target data pool). Imagination becomes the key factor in finding viability.

The value that humans provide as data detectives, curators, and scientists is considerable. The quality of machine insights is dependent on the accuracy and completeness of the data set the algorithm has to work with. Humans working in partnership with AI can evaluate hidden biases in a data pool, or identify missing data or new sources for comparison.

Successful startups often become the farm teams for the large enterprises. Attractive acquisition candidates tend to fall into three categories:

  • Companies that gather new data, either through a new method of collection or by drawing on a large user base in a way that activates more data.
  • Companies with platforms or tools that facilitate expertise with the data sets that the acquiring company has in abundance.
  • Companies with desirable human expertise that can become the seed of a big, multi-faceted corporation’s centralized AI team.

History teaches us that technological change tends to be evolutionary—except for the moments when it is revolutionary. Technology companies on the forefront of this revolution are grasping the opportunity to define a common understanding of what machine learning is and can become.

What remains to be seen is how well these enterprises share the resources and insights of machine learning to extend its benefits across human society.

At Intentional Futures, author Michael Dix and collaborator Scott Holden monitor developments in machine learning, autonomous vehicles, and Block Chain, among other technologies. Follow @intentfutures

Trending on Xconomy