When ChatGPT launched last year, it revitalized the world’s interest in artificial intelligence (AI).
The last time people were this excited about AI was in 2016 – 2017. Back then, people were impressed with the ability of AI to understand conversations and perform advanced analytics. But the excitement quickly faded when it didn’t live up to the promise of being as bright as a human. Sure, there were many successes, like virtual agents through an IVR or messaging channels, but those tools didn’t have the impact on jobs people thought they would.
ChatGPT, on the other hand, provided new capabilities that excited people about a better experience and demonstrated how much AI has advanced.
Leveraging Mainframe Data for Conversational AI
The latest Large Language Models (LLMs) are much more advanced than their predecessors. They can now have more sophisticated virtual conversations, create art, and write entire papers on a topic.
But one thing that has not changed is the importance of having accurate data to train the LLMs. The current LLM for ChatGPT sources its information from crawling the internet. While there are many successes, there are also many examples of ChatGPT writing a paper with false and inaccurate information. This is a serious issue since having accurate information is crucial for building trust with AI.
ChatGPT does provide a better conversational experience than previous solutions, but if you’ve ever had a conversation with existing AI solutions, you’ve no doubt experienced many of the current capabilities’ shortcomings.
For example, most virtual agents strive to do more than share data and answer questions. They want to be transactional: when you reach out to a virtual agent to look at different flight options for a trip, you want to know your options — but you also want to select a flight and potentially make changes during that same conversation.
This means the virtual agents will operate in a hybrid cloud environment and will need access to data located on a legacy system like a mainframe and replicating data will be complicated and costly. To provide access to that data, many customers have developed APIs and written code, increasing MIPS and cost. Many also developed change data capture (CDC), ETL, and/or event-based architectures — complicating their hybrid cloud environment and increasing the cost and effort required to manage multiple data sources.
How Lozen Can Help
Putting your mainframe data to use as part of a hybrid cloud strategy can be a complicated challenge. Lozen streamlines access to the data you need, making mainframe data available to use across hybrid cloud environments. It provides real- time, read-write access without compromising security or efficiency, allowing you to leverage that mainframe data in ways you’ve never been able to before.
Lozen lets you put your data to work immediately: Launch and access up-to-the- moment data with no custom code and fewer risks than other solutions. Get results with less manual effort and lower usage costs. And real-time, read-write also means real-time change data capture (CDC) without having to copy data or make changes to your existing mainframe applications.
As mentioned, the AI landscape is shifting fast. Machine learning is growing more accessible. Predictive analytics tools help guide smarter decisions. And businesses are finding they can deliver great insights by supplementing pre-built LLMs with their own current datasets.
Amid these fast-moving opportunities, there is one constant: An endless demand for quality data. With Lozen, you can simplify access to your mainframe data, making it easily available for model training and analytics. You’ll be able to connect your most current mainframe data to the latest AI innovations.
In addition, because Lozen was designed to be zIIP eligible, it is a significantly lower- cost option and in most cases won’t require any additional MIPS capacity.
Furthermore, Lozen can perform data transformation in real-time as data is being accessed, mapping VSAM/QSAM data and other mainframe data to more modern data formats, including text, CSV, XML, JSON, and binary.
Figure 1: Conversational AI Architecture with Lozen. (Click image to enlarge.)
With AI model usage on the rise, data consumption rises with it. Lozen will help you keep ahead of the curve and reduce the cost and effort of maintaining your data.
Deliver fresher results, fewer frustrations, and bigger impact. Lozen unleashes the power of your mainframe data and can support your AI strategies. VirtualZ also offers Consulting Services to support Data Access for AI.