Deep learning is great and all, but how do you implement in a production environment?
That’s a question that a lot of businesses are asking. Deep learning may look tempting when you think about how powerful it is. Research projects have shown it capable of replicating human vision and other types of thought processes. But it also can bring some complications to business processes.
In this edition of Talking Data we look at some of the business implications of deep learning, particularly when taking it from research projects to production environments.
About fifteen years ago, while covering application performance management, I had the great good fortune to meet Brend Harzog. Application performance in that first distributed age was like an onion — one with layer after layer that had to be peeled away in order to find the root cause of bottlenecks. Harzog, then a consultant and industry analyst, knew the domain well, and was patient, explaining that domain to a particularly curious journalist.
So, when I learned Brend had gone on to start OpsDataStore, a technology upstart intent on rolling up the diverse data sources in today’s data center into an understandable form, I was very fast to hook up with him for the Talking Data podcast. The topic: IT operations analytics.
It is Harzog’s argument that modern IT doesn’t need more big data as much as it needs tools that bring meaning to that data. While the topic may be a bit afield from our usual fare, we hope it is the start of an occasional series that looks at how big data is impacting some very diverse areas of interest. Listen to this edition of Talking Data. Let us know what you think about it.
Data science and artificial intelligence are converging toward one another, as businesses look to leverage anything they can to get a leg up on the competition.
In this edition of the Talking Data podcast, we take a look at how enterprises can support these two trends. While they may initially seem to be on somewhat different tracks, they actually have a lot in common. Some of the same businesses practices that go into supporting a strong data science program also enhance AI efforts.
One of the main things both analytics disciplines have in common is that they are different from traditional business initiatives, where ROI leads all considerations and projects must immediately prove their worth. With both data science and AI, projects can deliver value over time and need to be supported, even when they’re not directly bearing fruit. An experimental mindset is key.
Take a listen to this podcast to learn more about this and other topics related data science and AI.
Once upon a time .. access to a computer system was through an intermediary. The intermediary was an actual person – known as a machine operator or a key operator. This person worked 9-to-5, took off time for lunch, and input data into the System of Record, which was all about atomic transactions, and which, overtime, was all about relational transaction processing.
Then the Web happened, and nearly the entire population of the world became key operators. The input would come in great numbers, with unforecast spikes at unforeseen times. Web site operators began to collect data on the consumer army of key operators, to improve their Web experience. System of Engagement is a term that has come to describe that phenomenon.
Today, the System of Engagement has taken a place that is front and center in application development. For data management, that has meant wider use of non-relational NoSQL databases in operations. The systems handle vast amounts of data quickly, and thrive where relational alternatives now stumble.
The relational database still stands as the record of transactions. It is the final step – an important, but smaller, part of an overall system.
The impact of NoSQL and Systems of Engagement is the topic of this edition of the Talking Data podcast. We look at recent product updates in the space, and are joined by DataStax CEO Billy Bosworth for a free-wheeling look at the NoSQL terrain. Listen or download here.
Spark continues to gain attention as a unifying platform for analytics, machine learning, data streaming, and SQL querying. Surely, it is advanced technology, and, just as surely, Spark and its users will have to go through some maturation process. The steps in that process were on display in February at Spark Summit East 2017 in a snow-bound Boston. That and more is discussed in this Talking Data podcast episode by Trea Lavery and Jack Vaughan.
Yes, the BI army will continue to churn out those weekly and monthly reports, as in days of yore. But something else is going on too. Analytics is being connected to operations in near real time. Too, the path of data in organizations does not end when the data gets to the data warehouse. With the rise of data scientists, data is a two-way street. The shape of data management in the face of these changes is the topic of the latest edition of the Talking Data Podcast. Listen up!
A new project spearheaded by a researcher at Carnegie Mellon University is aiming to use AI to spot examples of fake news. But what at first may sound like a promising and relatively straightforward undertaking may in fact be a monumental lift.
The Fake News Challenge (#FakeNewsChallenge) is offering $2,000 to any programmer who is able to develop an AI algorithm that makes progress toward spotting fake news. But the challenges related to the project get at the very core of what makes true AI so hard.
Take a listen to this podcast to learn more about the Fake News Challenge, how researchers are trying to use AI to address the problem and why this task is so difficult.
In the year just past, data warehousing took a noticeable shift toward the cloud, as Microsoft, Oracle, Snowflake and others played catch up in pursuit of cloud leader Amazon. Just as remarkable, Hadoop distribution providers found themselves spinning up new versions of their offerings that were especially tailored to compete with Amazon’s style of Hadoop instances. These and other issues are under inspection in an edition of the Talking Data Podcast that looks back at 2016 in earnest and forward to 2017 in anticipation.
The year just passed would not have been the same without machine learning, deep learning and artificial intelligence. Trade and popular press alike zeroed in on these analytics trends with great fervor. In this edition of the Talking Data Podcast, TechTarget editors mull the meaning of it all. One question asked: Are we at the point where we can see adoption panning out at scale for machine learning and the like? To find out more, listen here.
Ed Burns looks back at the World of Watson – an IBM event held recently in Las Vegas – and finds much of interest to followers of AI. At WoW, GM said it would work with IBM to improve drivers’ experiences in traffic and at the gas pump. Staples rolled out a new version of its “Easy Button” that was hooked up to Watson Natural Language Processing and connected to the Staples order system. These and other use cases show IBM can read the tea leaves. Apple’s Siri and Amazon’s Alexa have often been front and center in AI discussions these days, and it appears the consumer angles is not lost on IBM, as GM, Staples and other Watson use cases indicate. “Household names seem to be out ahead in this,” said Burns. Hear more by clicking below to access this edition of the Talking Data podcast.