Artificial Intelligence & Machine Learning , Finance & Banking , Video

Key Considerations for Adopting Large Language Models

Adopting large language models involves several critical considerations for businesses. Vinit Tople, former head of generative AI product portfolio, Alexa Automotive, Amazon, outlined three key areas for implementing LLMs - identifying use cases, deciding between building in-house capabilities versus buying off-the-shelf solutions and mitigating the inherent risks associated with LLMs.

Factors such as cost, timelines and internal expertise play a role in decision-making, Tople said. He suggested that relying on external developers for general LLM capabilities often proves more efficient.

Companies should concentrate on training LLMs for tasks specific to their industry or where proprietary data is involved, he said. "There's something called a chain of thought reasoning in which the LLM takes the data points, comes up with the next step ... and then it completes a task over a series of steps. This complexity of going to build the context that you need to complete a task can be a use case where you have to bring it on yourself to make it happen," Tople said.

In this video interview with Information Security Media Group at the Cybersecurity Implications of AI Summit, Tople also discussed:

• Validating inputs and moderating outputs;
• Ensuring thorough risk management practices;
• Detecting fraud by analyzing interaction signals.

With more than a decade of service at Amazon, Tople has taken a series of large product ideas from concept through launch across various industries. The products have spanned the e-commerce, software and AI categories as enterprise offerings.