I do not want to be a data scientist, should I still study machine learning?

  •  
  • 1
  •  
  •  
  •  
  •  

Artificial intelligence (AI) and machine learning (ML) are often mentioned as things that will transform our world in the coming years. Lots of well meaning people implore young people to become data scientists.

The advice to become a data scientist ignores the fact that not everyone can, or wants to, be a data scientist. But should those who don’t want to work in the field study it anyways?

Everyone should learn about AI and ML

Not that long ago, computers were not in every workplace. You don’t have to go back to Mad Men days before every single office worker had a computer. Afterwards, employers would advertise looking for “computer literate” employees for years. Believe it or not some people still put “Microsoft Office” on their resume today.

Where am I going with this? Computers dominate office jobs today but AI enabled systems will dominate the workplaces of the future.

Many people will say they have heard this before. But, this time, it really is different. In the past, while computers eliminated some jobs entirely, most were just supplemented by computers. Now we are rapidly approaching the point where a computer can do the job by itself.

What do most jobs actually do?

A large portion of jobs require the employee to look at a situation, make a decision, and then take action based on that decision.

Compare that to what machine learning does.  It looks at a situation (data defined by the features), then makes a decision (prediction or label) and passes that decision to an AI interface to take action.

Machine learning makes predictions. It makes predictions on whether something will happen or not, what group something belongs to, or how large of an impact something will have.  Artificial intelligence uses machine learning predictions to take action.

Now ask yourself how different that is than many jobs.

How about an example of how an AI realtor could work

Predictions fill in missing information. Will the client buy this house? If we know their financial information and what they want in a house, an algorithm can predict which houses they will like. An AI interface with this algorithm can push those houses directly to the customer. They can also access the property when the AI interface notices they are at the property.

With a little work, you could set up a virtual tour of the house. High points to the house room by room that come up when the customer enters the room or on demand.

That is an AI realtor. All of that is possible today based on very simple algorithms that everyone understands.

 

Or…robot realtors may already exist.  

I know, you’re going to say there is a whole lot more to it than that. What about legal and zoning restrictions?

The largest investment banks and their lawyers are using AI to perform due diligence on complicated, billion dollar, international deals. If they can do that, don’t you think they can handle a home sale in Fresno?

That home buyer is going to need a loan. Currently, they’ll fill out some forms online and then a bank representative will call them.  Underwriters and loan officers look through documents and then process required paperwork. They take action by asking the customer for more documents or processing existing ones. How long do you think it will take before a computer is doing that?

Buyers make their decision based on the rate. If a bank can offer a lower rate by saving money on people in the process, how long until that will be commonplace?

AI can replace many office jobs

I was not intending to pick on the real estate or mortgage industry in my latest examples. If you consider what most office jobs actually do, you can break it down to making a prediction and then taking action.

Do you see how this can expand to other areas in the real estate process like insurance and taxes?

Think about it from the perspective of your job. How much time do you spend on forecasting or budgeting? Those are essentially predictions; how long until a computer can do it better?

You don’t believe me that it will ever get to that? People used to ask for directions, now you ask for directions and they’ll say, “use Google Maps!”

AI and ML will get cheaper

The cost of making a prediction has already fallen significantly. Those prices will continue to fall as computers get faster and more data becomes available.

If you consider a prediction like a commodity, the ramifications are rather obvious. As prices fall, we tend to use more of that commodity.

Avi Goldfarb, in his book Prediction Machines, uses coffee as an example. If coffee gets cheaper, people will drink more of it. They will also use more related goods like sugar, cream, and cups. They will probably use less of substitutes like tea.

Check out his podcast on Harvard Business Review where he discusses these issues.

 

Wow, it sounds like the demand for data scientists will continue to grow!

Well maybe, but maybe not. It is also possible that increased computing power, more availability of data, and simpler tools will increase the supply of predictions. There could be far more capacity for making predictions than demand for them. What takes a team of data scientists today could only take one or two in the coming years.

When the prices of a commodity fall, the people who produce that commodity earn less.

Demand for data scientists may drop or they might earn less? Why would I learn about it then?

I am not actually saying that the demand for data scientists, aka the people making predictions, will drop. It might, but it also might not. It is also far from certain that earning power will decrease.

What definitely will not go down is the demand to DO SOMETHING with those predictions.

The people who know how to make predictions AND understand how to use them will win the day

People who know the ins and outs of making predictions are naturally at an advantage. Understanding what is possible gives them an immediate leg up on those without a background understanding.

People who understand how to properly use a prediction are also at an advantage. There will be tons of “off the shelf” predictions available that are decent quality and cost effective. But the person who understands the nuances of what they really need in a prediction will win.

If you combine those two knowledge bases into one person, they will be unstoppable.

Conclusion, aka what should I do if I want to win in the future

If you do not already have any skills in machine learning or AI, you should study up. There are lots of programs available online for free or cheap that can get you up and running. You can utilize that skill in improving your current efficiency or finding ways to branch out.

If you are a current data science practitioner, expanding your knowledge outside of data is key. Can you sell? How well do you know the ins and outs of your business?

Those talents will help you succeed as a data science leader right now. In the future, they will be necessary to win.