“Traditional ML isn’t dead. It’s expanding into many more specific use cases, requiring greater precision. While those are still evolving, GenAI has been growing at a much faster pace,” said Sumeet Tandure, Head of Sales Engineering (India, Commercial & South) at Snowflake.
While speaking at Cypher 2024, India’s biggest AI conference by AIM Media House, Tandure clarified how generative AI has quickly overtaken traditional machine learning (ML) as the dominant force in AI. “Generative AI has become the default meaning of AI today,” he said. “Tools like ChatGPT have captured both the consumer and enterprise market’s attention, but the shift goes deeper than just hype.”
However, he emphasised that traditional ML still has its place, especially for use cases that demand high precision. He agreed that the adoption rate of GenAI has far outpaced previous AI technologies. “What took machine learning nearly a decade to achieve, generative AI has accomplished in just two years,” Tandure said. “Over 65% of enterprises are actively exploring or deploying GenAI right now.”
“If you are building traditional ML, predictive ML, or discriminating ML, there are enough use cases for that—examples include financial planning, market prospect analysis, CRM, and so on,” said Tandure, adding that there are a huge number of use cases around traditional ML, from feature engineering for these traditional ML models to training the model and then deploying it. “You can manage the entire ML lifecycle and MLOps process within Snowflake.”
He added that three key areas, sales, marketing, and security—along with aspects like risk management, forecasting, pricing, advertising, and supply chain and manufacturing, represent some of the traditional machine learning applications.
Navigating Enterprise Challenges with Generative AI
Tandure didn’t shy away from the challenges of scaling Generative AI in enterprise environments. He pointed out that while GenAI experimentation is widespread, taking these models from pilot phases to production at scale presents significant obstacles. “In a consumer setting, a ChatGPT error might not have serious consequences. But in enterprise use cases—whether it’s financial decisions or medical records—accuracy and trust are non-negotiable,” he said.
A major concern, according to Tandure, is ensuring that GenAI models produce high-quality, reliable responses. He emphasised the importance of preventing “AI hallucinations”—a term used to describe AI-generated outputs that are inaccurate or misleading. “Enterprises need to know that the answers they’re getting from AI are grounded in their trusted data sources,” he explained.
Bringing AI to the Data, Not the Other Way Around
Snowflake’s product approach also includes making AI more manageable by integrating AI models directly with the data. “Instead of sending your data to the AI models, we bring the AI models to your data,” Tandure said, emphasising that this eliminates concerns around data security and network inefficiencies.
Snowflake provides a single, fully managed platform that powers the AI Data Cloud, allowing organisations to securely connect and work with data globally across any type or scale to develop AI, applications, and more in the enterprise.
Tandure said that the platform leverages Snowflake’s data cloud, along with tools like Streamlitand Cortex AI, to build AI applications that are easy to deploy and scale. “We’ve built a complete ecosystem that allows enterprises to go from data to deployment without needing to worry about the technical complexities behind it,” he said.
Snowflake Cortex AI is a suite of large language models (LLMs) designed to understand unstructured data, answer freeform questions, and provide intelligent assistance. It includes foundation models such as Snowflake Arctic, Meta Llama 3, Mistral Large, and Reka Core.
“A solid data strategy is the foundation for a robust enterprise AI strategy. That’s the first aspect you need to address. Once you have the data in place, you can efficiently build AI applications and use cases on top of it,” he added.
Moreover, Tandure said that for businesses looking to commercialise their AI applications, Snowflake offers a marketplace where companies can sell AI-powered products. “We’re helping enterprises not just deploy AI internally but also monetise it,” he added.
He also shared several real-world examples of enterprises using Snowflake to scale their Gen AI efforts. One notable case was Siemens Energy, which utilised Snowflake to consolidate millions of research documents into a single platform. With a custom-built AI chatbot interface, researchers at Siemens were able to ask questions and receive precise answers from the database.
Tandure concluded his talk by conveying a clear message: ‘Make AI reliable, efficient, and easy to deploy.’
The post Traditional ML is Not Dead appeared first on AIM.
