A Call to Get Specific About Generative AI

Analyzing Sam Altman's Future of AI Comments at MIT

Generative AI holds great potential for enterprises. At Evalueserve, we see domain-specific implementations of generative text, images, design, video, and data as an enterprise use case and need dictates. The better trained on domain-specific and use case-specific data, the more likely an enterprise will see strong outcomes.

Open AI CEO Sam Altman acknowledged as much when he said at an MIT event last week, “I think we’re at the end of the era where it’s going to be there, like, giant, giant models. We’ll make them better in other ways.”

While Open AI’s large language model ChatGPT has inspired hundreds of thousands of users to experiment with AI, the new generative AI revolution interests enterprises but hasn’t fully taken root yet. The reason why is simple. LLMs are not trained to deliver results like domain-specific AI can.

Many people focus on the data that goes into an LLM model. This is 100% on point; we all understand the importance of quality training data. But just as important is prompt engineering, which must understand the nature of the industry and the subset of context data relative to the problem — particularly for a domain-specific use case.

Making Generative AI Tools Work for Companies 

Generative AI is a critical algorithm toolset for future business AI uses. But an AI model that includes a generative AI algorithm needs a strong use case and to be trained for it to be advantageous to the enterprise. 

ChatGPT’s impact on the consumer landscape has yet to achieve the same success on a large scale inside enterprises. A recent report by Battery shows that only 32 percent of companies are looking for generative AI application pilots, primarily for internal productivity, workflows, knowledge transfer, and cost reductions. https://www.battery.com/blog/bv-march-2023-state-of-cloud-software/#early-signals-for-generative-ai-in-the-enterprise

Each company knows its industry and customers better than any large language model possibly could. This knowledge is what we mean by domain-specific and AI products and tools built by subject matter experts within an industry. Domain-specific AI models train on that industry’s particular prompts, queries, data sets, terminology, and experiences. Then, combined with a use case that can produce meaningful outcomes, it becomes tangible and worth doing.

The above oversimplified model shows a straightforward idea: A basic model is not enough. You need people with specialized knowledge in a specific domain area to create valuable outcomes for a business. From prompting to organizing data to verifying responses, human guidance helps domain-specific AI succeed.

Use cases help define outcomes by clearly stating business objectives, identifying the challenges that need to be resolved, and potential benefits or the ROI. But a good use case also defines feasibility, risks, measurement, and possible proofs of concepts. When these factors are clear, enterprises can select the use cases with the greatest likelihood of impact and long-term use as a repeatable solution.

Envision a future with thousands of smaller models working together, a symphony of AIs, to provide more efficient and accurate business results. Many of these models will work together in low code/no code environments, interacting to achieve their desired outcomes. But first, getting specific about use cases will be critical for the long-term successes and breakthroughs in AI, both in general and within the generative AI subset.

Elevating Outcomes Is the Barometer

My colleague, Satyajit Saha, noted in his recent discussion about our domain-specific Research Bot’s use of generative AI, real-world applications and benefits matter. Therefore, elevating outcomes for an enterprise is the barometer for adopting any use case that requires AI.

In the use case Satyajit outlines, a consulting firm leveraged Research Bot as part of a larger Insightsfirst model to stay ahead of their competition. Our competitive intelligence teams use Research Bot to:

  • Provide the latest, most relevant insights tailored to your specific questions whenever you need them
  • Generate parallel prompts to construct key takeaways that are multi-faceted for each topic
  • Offer easy access to sources for each insight
  • Enable filters to focus on critical insights and narrow the scope of the conversation

Research Bot’s just one example of how Evalueserve uses generative AI tools to create meaningful outcomes for enterprises. You can read more about some of our efforts here.

Generative AI tools like the GPT algorithms created by Open AI will have a vital role in the future of enterprise-level AI. It begins by understanding how generative AI can make the right results for a business. Being specific within domains is essential for the next generation of models and implementations to succeed.  

Rigvi Chevala
Chief Technology Officer Posts

Latest Posts