Recent developments in machine learning
The machine learning panorama is one I follow closely in my capacity as a strategic advisor to BigML, and has recently produced some interesting developments.
Scaled Inference, a company made up mostly of former Googlers, raised 13.6 million dollars at the end of December as part of an investment round that saw the company valued at 60 million dollars: a company that is less than a year old is looking for money to build a product that it still doesn’t have and that says that won’t be ready for at least a year. Vinod Khosla, an advisor to the project and who has already invested in other similar startups such as MetaMind, is still on the look out for opportunities, aware of the sector’s future.
Seattle-based GraphLab is a startup dedicated to lowering the entry barriers to machine learning and that announced on January 8 in a blog entitled “2015: the year of machine learning” it was changing its name to Dato and a new investment round of 18.5 million dollars. This is a company founded in 2013 from open source software, with a portfolio of just four clients, and that says it aims to provide analysis tools to clever engineers able to build applications and connect up to a data base but who lack any training in machine learning.
Meanwhile, BigML continues its super-lean progress and now has positive cash flow and a very interesting portfolio of enthusiastic clients. It was recently classified by ReadWrite as one of “nine startups that made life better in 2014” and by Inside Analysis as one of the “ten companies to watch in 2015.”
I don’t know whether 2015 really is going to be the year of machine learning: my impression is that we’re talking about something that is developing slowly, almost by word of mouth, as managers realize the potential of the data they generate — all types of data , not just the transactional— and that they can realize this potential by using simple cloud-based analysis tools that are as simple to use as drag and drop and need not cost much. On many occasions, we have seen how directors become aware of these possibilities as a result of analysis of different types that achieve a certain visibility or by following their competitors announcements.
What happens when we teach a computer not to do something specific, but to simply learn? What are the implications of such a development? Are we sowing the seeds for the end of humanity, as Stephen Hawking has warned? Or will technology simply do all our work for us? More and more technology, and less and less jobs? Either way, this is a fascinating sector and one that I am very happy to enjoy a privileged position within that allows me to keep up to date with, and one that we are going to be hearing a lot more about in the near future.
(En español, aquí)
