Hi, With the popularity of Generative AI, anyone can shed some insights on why production and enterprises would prefer Traditional Machine Learning and Deep Learning Technique over Generative AI given ready platform like AWS Bedrock, Vertex AI from Google etc？ Thanks
The key factor in deciding between traditional ML/DL models and Generative AI hinges on the specific requirements of the use case. For tasks where interpretability is paramount or that involve conventional classification/regression, traditional ML/DL approaches often prove more suitable. On the other hand, Generative AI excels in tasks that demand creativity. Additionally, despite the availability of cloud services, deploying traditional ML/DL models can be quicker and more straightforward for certain applications.
Thanks for the insights. In a business, cost would be deterministic factor sometimes. If I can accomplish a task using both, will the cost involve in fine-tuning a LLM Base Model etc be likely higher than that of a traditional,approach?
It’s computationally more expensive to train an LLM. As such, if there’s an effective traditional approach, it would be more economical to go for it business-wise. Unless there is a justifiable improvement that can be achieved by using the LLM. And that too depends on the business objective.
In today’s tech landscape, the rise of Generative AI has been remarkable, yet traditional Machine Learning (ML) and Deep Learning techniques continue to hold sway in production and enterprise settings. Despite the user-friendly platforms like AWS Bedrock and Google’s Vertex AI catering to Generative AI, businesses often prefer the reliability and well-established nature of traditional methods. Traditional ML and Deep Learning offer proven models, robust interpretability, and a track record of success in various applications. Moreover, many enterprises may find it challenging to fully integrate Generative AI into existing workflows, preferring the familiarity and seamless integration that traditional approaches provide.
Thanks for the great reply. I have the same thinking also. Your reply reinforce my thoughts. Yes. You are right. Generative AI products and apps does have difficulties in pulling off to big enterprise because of security and privacy and other issues. Thanks again. 
I think it really comes down to the use case. Generative AI is useful for things like content and open-ended tasks, but traditional ML still makes more sense when you need predictable results and better control. For businesses, cost, privacy, and accuracy are also big factors to consider.
It is also about cost and convenience, I don’t need generative AI capabilities for a task that can be achieved by much simpler and more explainable models. From what I have seen, a decent amount of companies are looking into how to benefit from the simple models, and if these fail for whatever reason, then we move on to bigger and better and more complicated models and techniques.