Back

 Industry News Details

 
How Do You Feel About An Algorithm Deciding If Your Startup Gets Funding? Posted on : Mar 29 - 2021

I’ve been keeping an eye on the use of machine learning algorithms, particularly by venture capitalists, to make investment decisions for some time now. They’ve been investing in machine learning companies for years, so applying their products to other businesses, once you have studied how they work, seems a reasonable proposition.

After all, what is the decision to invest in a startup based on? Basically, the fruit of a set of analyses and previous experiences that can be systematized and verified in different ways, while the experience corresponds, in reality, to the imperfect distillation, with its biases and errors, of a series of previous decisions, weighted by the results obtained in each.

That said, venture capitalists are not entirely objective: they usually allow multiple factors to enter the decision-making process, which include anything from the feelings generated by the company’s founding team, to more or less rigorous analyses of its capacity for future development.

Could we consider replacing or complementing this type of decisions — in many cases personal or consensual among the members of an investment fund — with the results of an algorithmic analysis based on previous decisions, labeled with the results obtained over time? Can a start-up’s success be predicted, or does it depend in many cases on factors that are impossible to foresee or whose results may be different each time, depending on specific personal or other factors? What happens when success, moreover, is significantly influenced by the synergies or contacts that investors can bring to the company’s project?

A few years ago, I gave a class entitled “When an algorithm decides whether you get a mortgage”. Increasingly, we will see decisions with major economic repercussions being made based on analytical processes of this type. Factors such as intuition and experience will never go away, but they will be increasingly influenced by what a particular algorithm predicts, supported by lots and lots of data from previous investment rounds and their results over time.

An algorithm is not an abstract entity, it’s an analytical construct based on mathematics and statistics, obtained from given data, and labeled with the result that data produced. Obviously, there is no guarantee that variables relevant to the final result and that we simply did not understand or detect, but we can approximate what part of the decision is susceptible to be explained by our analysis. Then there is the question of biases, since they may be determined by the data we feed or educate the algorithm with. View More