Posted on: December 03, 2018, by: Ann Kelly
AI – Myths and Reality
Global business leaders believe that AI is transformational and either they incorporate it into the enterprise or get left behind. The reality - AI is still in its nascent stages of development. While there’s plenty of AI used in specific areas such as fraud detection, billing systems, and data integration, autonomous machines that replace humans is still a work of fiction.
Despite its popularity, there are still many misconceptions and myths surrounding AI and how it can be leveraged to gain insight.
Myth # 1 – you just give the machine the data; it learns and then delivers useful insight
In a business sense, the role of AI is to derive insight through better analytics and improve efficiency. While true, to gain insight, whether you are a human or a machine, you need context - metadata is the key to context.
According to Gartner, metadata describes various facets of an information asset in order to improve its usability throughout its life cycle. Metadata management is about the organization’s management of its data and information assets to address use cases such as data governance, analytics and enterprise metadata management (EMM).
It is important to note that this understanding of metadata goes far beyond just technical facets; it is used as a reference for business-oriented and technical projects and builds the foundations for describing, inventorying and understanding data for multiple use cases such as information governance or analytics.”
On its own, AI is not able to identify and extract context, however, incorporating a semantic platform like Semaphore, which creates rich metadata, enhances AI processing.
Myth #2 – we will quickly assemble the training set, then pour it into the machine and it will deliver actionable intelligence
From a simplistic perspective, machine learning is the construction of mathematical algorithms that can learn from and make predictions on data. Each concept/construct requires 3 sets of data; training, validation, and testing.
Each concept is initially fitted from a training dataset, which is a set of carefully selected examples used to train the concept. The concept is trained using a supervised learning method and produces a result, which is then compared with the target. Based on the result of the comparison and the specific learning algorithm being used, the parameters of the concept are adjusted. This is stage one.
The output from the stage one concept is compared against the validation dataset. The validation dataset provides a known control. If the machine passes the controlled comparison we can call it a stage 2 machine.
Finally, the test dataset is used to provide the phase 2 machine using data that has never been used in training, but which the answers are known, this gives a measure of its accuracy in real life.
Compiling the data sets is time-consuming, requires specialist skills, and is intellectually difficult as the more data dimensions required for the analytics task, the harder it is to gather and assemble 3 sets for each data dimension.
The second point is that we have disintermediated the business minds from the construct creation because we need quite a few maths graduates to get involved in the learning stage.
Myth #3 machines are going to be autonomous decision-makers soon
Today, AI is not mature enough to provide full autonomous decision support. It can, however, automate the next level of repetitive work, which is compelling from a cost and efficiency perspective. There is a data quality yield as well, the consistent application of digital-driven processes result in accurate and traceable outcomes. This data veracity is important, as the quality of information affects analytics and the fidelity of business decisions.
Machines need a narrow focus – they can drive cars but when it is snowing they are less reliable unless they can see the road, find the markers, and know how to manage conditions. Data in the enterprise is wide, not narrow so it is unlikely they will be autonomous anytime soon. To take advantage, it is best to leverage each to their strength: humans do creative and context and machines do the math.
Will AI replace human intelligence, disrupt the workforce and make business decisions – in the short term – no. However, if we allow humans to govern knowledge and domain models, which harmonize, extract and enrich data, organizations can apply the relevant context to all enterprise information and leverage it to break down complex tasks into units that can be automated. Smartlogic is helping customers do this today.
CEO and Founder
Data Governance in Pharmaceutical Science and Technology
In Life Sciences, data is the new currency but only if it can be mined and retained at the highest quality; used properly, it can drive insights that improve research, operations, and outcomes. Pharmaceutical and biotechnology markets are poised to exceed $1.2 trillion by 2022. As data volumes continue to explode, opportunities, as well as challenges, emerge:
Investing in innovative and cutting-edge technologies to manage today’s massive data volumes from disparate sources is key. Storing and recording data from multiple sources is one thing, harmonizing and using it to gain insight require more than traditional technologies can currently deliver.
To learn how one Smartlogic customer solved data governance challenges download our case study
Smartlogic Recognized in Gartner’s Critical Capabilities for Metadata Management Solutions
Smartlogic Semaphore Inc. makers of the world-class, Semantic AI platform Semaphore, has received above a 4.0 out of 5 across all use cases in the November 2018 Gartner Critical Capabilities for Metadata Management Solutions.
Smartlogic releases Semaphore 4.2 - an enterprise-grade, semantic AI platform
Smartlogic Semaphore Inc. announces the release of Semaphore 4.2, the latest version of their enterprise-grade, semantic AI platform used by Fortune 1000 companies worldwide. Semaphore 4.2 delivers key capabilities across model management, content classification, and fact and language services.
KMWorld 2018 - If you were unable to attend KMWorld 2018 this year you can still
KMWorld 2018 -
If you were unable to attend KMWorld 2018 this year you can stillsee Jeremy Bentley’s presentation using this link.
If you missed our webinar: Creating a Unified Customer Experience with SDL and Smartlogic you can listen to the replay here.
Smartlogic provides training so your team can get the most from the Semaphore platform. Smartlogic will be conducting our full set of Online Semaphore training courses this Spring. Schedules and timelines will be posted in January. To learn more about each course, view our events page at https://www.smartlogic.com/events.
To register for courses, email us at firstname.lastname@example.org or contact your Smartlogic Account Manager.
To learn about upcoming events, webinars and recent blog posts, follow us on:
US: +1 408-213-9500
US Federal: +1 703-956-2600
UK: +44 203-176-4500
Copyright ©2019 Smartlogic Semaphore Ltd