Big Data has been a hot topic since the Information Age took the world by storm in the 21st century. What quickly followed was the birth of the “data scientist” as an occupation to manage this influx of information, a job title our past VP of Product, DJ Patil coined in 2008.
Having witnessed the growth of data science as a discipline, our Chief Data Scientist, Vitaly Gordon is excited for its future. Vitaly currently leads our data science team at SalesforceIQ, where they are developing sophisticated machine learning tools for the intelligent Customer Success Platform.
We sat down with Vitaly to get his perspective on what he envisions for the future of data science. For part 2 of this conversation, read the post here.
“Data Scientist” is a relatively new term, but that doesn’t mean data scientist jobs didn’t exist before. There were data analysts, data mining, engineers. Google was doing it, Amazon— there were companies with data roles before it was called “data science.”
And that’s for a reason. It is new. As many people mention, it’s the Information Age. It’s similar to social media marketing managers. It isn’t as if social media didn’t exist before, but as the industry changes and the need grows, it forms a specific job function: the social media manager. It’s the same situation for data science.
Let’s look at where data science is now. In the last few weeks, Google created an AI that beat Go world champion Lee Sedol.
In ’97, IBM’s supercomputer Deep Blue beat world chess champion Garry Kasparov in a six game match.
But this time, it’s different from the match in 1997, because Deep Blue was an expert system — which I’ll explain further in a moment. Deep Blue was taught by chess grand players to play chess, but the DeepMind Go system taught itself how to play Go. There were not as many experts involved, so this is a mind blowing breakthrough.
With chess, it’s more about who can think more moves into the future. For machines, it’s actually not as difficult as Go, which is far more complex. Go is complex enough that it’s actually impossible to run through every possible move. Instead, what differentiates expert and novice Go players is their sense of intuition for the game.
When Google’s AI beat Lee Sedol, it showed us machines can now imitate this sense of intuition.
So what’s next?
The thing is, we’re not even realizing how big data science will become. - Vitaly Gordon
You can think of so many things: computerized lawyers, computerized medicine, even computerizing VC decisions when they invest in startups. And you can get even venture into art and music, and computers writing papers indistinguishable from human-written papers.
There’s a philosophical debate on AI. Is AI teaching computers to be more like humans? Or is it about making them the best they can be at a task?
So data science is so much more than what it is now, you can see the wide range of applications in the future that haven’t even been explored.
Artificial intelligence is a broad term that simply means: any intelligence run by a computer. Machine learning is a subpart of this, but you don’t need machine learning to have an AI.
For example, what is an AI that’s not machine learning?
Search is a good example. GPS also. In GPS, when you’re given a destination, it computes the best path for you. It doesn’t necessarily require machine learning, since what it’s doing is making computations quicker than humans are able to rather than demonstrating the ability to learn from data.
For example, the Google AI that beat Go champion Lee Sedol, didn’t have to consume data to learn and improve.
You can teach a machine with techniques other than data.
No, for example, you can work with many customers and each customer has a small set of data. That’s data science where you’re working on many small data problems at once.
But, there are some problems that will only occur with larger data sets. Take language and translation for example. It takes humans years to understand language, there are so many use cases to learn. You probably experience this when learning a second language — it takes years to master all the rules and exceptions. Some problems you can easily deduce insights, but for things with more structure? That’s where Big Data comes in.
Before machine learning and data science, we had something called “expert systems,” as I mentioned before, which were essentially groups of experts that instilled machines with knowledge in the forms of rules and logic. This used to be very state of the art, and they would produce a lot of rules.
What companies realized was, at a certain quantity of data, machines will learn more than they would from experts.
With Big Data, machines actually get a richer repository of options to compute from.
In part two, we’ll discuss with Vitaly on what we can do to prepare ourselves for the future of data science. How will you and your business approach a world that is run on smart tools and intelligent applications?