The challenges and opportunities presented by transformative Generative AI technology are posing dilemmas for IT leaders tasked with delivering innovation.
After the initial explosion of publicity around this technology, digital experts are now collaborating with business managers to navigate a transformative path that results in tangible business success.
It hasn’t taken long, however, for all parties to discover that the development of game-changing Generative AI strategies is potentially complex, costly and high risk. Businesses are taking a distinctly hard-headed approach.
A roundtable event hosted by EPAM and AWS heard that IT decision makers are being pragmatic when faced with these challenges, reducing the scope of their Generative AI projects to ensure return on investment. Much of the conversation focused on the financial sector.
Focusing on low-risk proof of concept
Participants said they are searching for Generative AI use cases they can develop quickly and cost-effectively in the cloud. This approach will enable them to demonstrate proof of concept without committing significant time and resource.
Several roundtable attendees said they are investigating how Generative AI can enable non-specialist employees to query data using natural language prompts. This dispenses with the need for code and specific nomenclature.
Generative AI use cases in the room included enabling customer service representatives to access and interrogate mission-critical data across multiple systems, without the need to toggle between multiple screens.
Another attendee said they wanted to achieve a more cohesive approach to risk management by using large language models to aggregate data and achieve a single view of the truth.
Collaborating to overcome data and regulatory risk
Attendees agreed that two of the biggest barriers to Generative AI adoption involve data risk management and regulatory compliance.
Much of the data required to build large language models already resides in legacy IT systems. Data aggregation, migration to the cloud and operationalization, however, expose this data to new risks.
In one well-documented example, engineers at electronics giant Samsung accidentally leaked proprietary information by inputting it into ChatGPT.1 Samsung and many other multinationals have since banned employees from using ChatGPT to prevent this from happening again.
Collaboration is the key to safe and secure data migration to the cloud. In addition to working with key internal stakeholders (e.g. risk, compliance, security and governance teams), it is also advisable to partner with a leading cloud provider, such as AWS, which has market-leading risk management and data security baked into its solutions.
Moreover, industry verticals each have a unique and fast-evolving regulatory landscape. Partnering with a specialist capable of monitoring this landscape, such as EPAM, adds significant additional value.
Humans in the loop to verify Generative AI outputs
Attendees also discussed hallucination: one of the biggest risks associated with Generative AI. This occurs when algorithms identify and base their outputs on purely random patterns found in large language models, potentially causing errors.
Until a solution can be found, companies are adopting a human-in-the-loop approach, which involves humans manually validating Generative AI outputs. This human-centric approach reduces risk, but it can also cut productivity gains and undermine overall solution value.
Taking a discrete and pragmatic approach
Companies’ pragmatic approach to Generative AI means they are reducing the size and scope of their projects to help manage risk in areas such as data use and hallucination. The unexplainable nature of Generative AI algorithms is also causing enterprises to avoid seeking outputs that require auditing.
Five years from now, such challenges are likely to be overcome, but for now, EPAM’s London roundtable demonstrated that the trend is for discrete and pragmatic Generative AI adoption.
Discover how EPAM and its cloud partner AWS can help you begin your Generative AI adoption journey.
1CSO, Samsung bans staff AI use over data leak concerns: May 2023, https://www.csoonline.com/article/575215/samsung-bans-staff-ai-use-over-data-leak-concerns.html