Niranjan ThomasGeneral Manager, Developer Platform & Solution Engineering at Dow Jones
About the talk
Artificial intelligence models feed on data, but what do you do if your data pipeline is running dry, or you need information that's not present in corporate repositories? It's time to turn to data vendors - some of these are new kids on the block, while others are established names that are re-inventing their business models to accommodate the new data economy.
Dow Jones fits into the second category: primarily known for its publishing and financial information services, that include the Dow Jones Industrial Average, the company is creating new revenue streams by offering data for AI training.
At the AI Summit New York in December, we caught up with Niranjan Thomas, general manager for Developer Platform and Solution Engineering at Dow Jones, to find out more.
We're here at the Javits Center for the second day of AI Summit New York and I'm speaking to Neon John Thomas who's the general manager of one 00:04 of the businesses that Dow Jones runs and it's 8 its professional product about your work. And then what is Dow Jones interest in machine learning 00:14 you have a really powerful set of apis and and data feed products 00:23 across that fact either and use wise and every Screen Compliance product line. My role is to really make sure the developers who are using apis both 00:33
understand have a great experience are in a successful in building application around those apis and seeds and its applications really powering the 00:41 most important decisions that they make it within their businesses. That's my roll a joint in machine learning this company right into the 00:50 provider better. We employ. I am machine learning in the production in the creation of You'll see what we customers who invite themselves using 00:59 artificial intelligence and machine learning without data so many places for Sweden at 4. And also we do not customers environment. Okay, and then you 01:08
deal with the news information me the information and I'll structured and then and just generally to walk around the show floor. There's a lot of 01:17 conversations about unstructured data and and the value that I'd there at the challenges in in in in understanding unstructured a thin, you know, like 01:23 getting somebody out of it business. So we have list of 01:32 sanctioned individuals and and Bad actors. If you like an instructed to I-26. We also have completely unstructured data, right so news articles, which 01:42
we can use wise for us and we can consider the articles is unstructured content. 01:52 To tag in code the content. So that customers can find the companies in the people when the subjects on topics that they're interested in. Right but 02:04 beyond that we also a finding more and more as machine learning in artificial intelligence become so prevalent within a costume is environments that 02:12 they themselves when I use machine learning without data sets and they want to bring a different level of structure to that data in some cases at 02:19
custom. It's a free camping the reinsurance face a very very Domaine specific knowledge of the day looking for weeding out data, so it might be 02:26 something very specific to Property and Casualty or life and health insurance. Now that wants a logical models in the structure of what they looking 02:33 for is highly specialized and they use machine Learning Without items 02:40 machine learning. 02:44 Self-realization off of these DVD's work clothes pension rights. Are we moving from a world which software 03:04
is developed in a way where you explicitly give the machine all of the instructions in terms of what to do to a world where you give the machine some 03:14 of the instructions or no instructions at all in the machine learns from the data. Unstructured or otherwise, like I said, it's two pots that juani's 03:23 how do you create discipline and certainty in that world with machine is making decisions and then I discipline and structure in that then how do you 03:31 style it as you describe right to visit to really important question and a number of different areas operationalization considers things like the 03:39
skill sets and capabilities within an organization it considered some of the processes, right and the disciplines to ensure that there is no inherent 03:47 bias in some of the decisions of the Machines of making the government. So there are clearly Cymbalta considerations around the decisions that 03:54 machines are making soap. Like a sequel considerations avoiding bias, right? And I think you know it is it is about really putting all three together 04:03 when we talked about operationalization people process and the right. I mean you've been in several tunnels are like over these two days 04:11
are like the most interesting outtakes 40 points, but the same people talk about you show us what I think I think we really moving 04:21 beyond the technology in a little bit. So what we're really focusing on now as an industry and has a broader ecosystem is really the really exciting 04:30 business challenges that we can solve together. Right? I mean, I am always amazed at some of the amazing applications of Dow Jones data by what about 04:40 Customs I gave you an example before the reinsurer using instruction use content from efectiva news content from a database to 04:48
use machine learning in artificial intelligence to better understand some of the risks that I might face is a reinsurer we've got In a training space, 04:57 we've got clients that are executing trades using a newswise product. All right, and there's increasing sophistication in the models that they 05:05 building those clients are able to bring in more diverse dated who made the news and correlating the news with many many other instructions phone to 05:12 divert as well. So I just satellite imagery and other forms of alternative data compliance business right decision on who they want to do business 05:21
with on naughty business with based on their own internal rhyme with all the regulatory environment in which they operate the applications are 05:30 absolutely right across the professional okay. Mmmm. It's a professional business. But what about for example social media? So is there something 05:37 you're looking at right now, or is it something you you might consider in the future, social media perspective in the past. We have looked at me 05:46 constantly evaluating the data sets that we bring into a product with that being used why I spect evil risk and compliance. So we're always on the 05:53
hunt as a date of business to identify new and value-added. It's for a customer. But we also recognize that customers have their own choices around 06:01 idea that they making all the time first-party data that may exist in their environment or data that they made for Cutie independently of Dow Jones 06:09 and what's really important for us is to fold do we have the right data sets available? We have products throughout these are apis or a user 06:16 experiences a mobile apps excetera easy for customers to link their data to Aldo today 06:24
as well around the earth cleansing and scrubbing data and linking data. It's just hard work and it's not getting any easier. I 06:33 want we are always on the hunt to do is make sure that we are the easiest way to sit to work with and we remove a lot of the pain points that we 06:43 provide high-quality to our customers whether we provide a direct sales only make it easy to link outdated to the other day that says that's a real 06:49 priority for us. Critical in terms of 06:56
Buy this talk
Buy this video
With ConferenceCast.tv, you get access to our library of the world's best conference talks.