Joe Rose of JBS Dev Debunks Myths About Generative AI and Imperfect Data
In the rapidly evolving world of artificial intelligence, misconceptions abound. One such myth, often perpetuated by vendors and consultants, is that data must be flawless before engaging with generative and agentic AI systems. Joe Rose, president of strategic technology provider JBS Dev, is keen to set the record straight. “It’s a common misconception that your data needs to be perfect before running these types of workloads,” he explains.
Understanding Imperfect Data
A recent article in the AI Fieldbook highlights this misconception, noting the push for massive data lakes or extensive data transformation programs. However, according to Rose, the reality is different. “The tools have never been better for dealing with poor quality data,” he asserts. “It’s almost remarkable what an LLM can understand from a half-written prompt.” This adaptability makes AI a powerful tool even when data quality is less than ideal.
Leveraging AI with Human Oversight
While AI technologies can process incomplete or imperfect data, it is essential to use them with the right guardrails. Rose emphasizes the importance of human oversight, especially given the inherent unpredictability of AI models. “People are used to: ‘We build it, it works, we forget about it,’” he notes. “That’s just not how these systems work.”
Rose shares an example from the medical sector, where a client was transitioning to a new billing reconciliation system. The data was a mix of formats, including PDFs and images. Despite this, the generative AI was able to extract clean data using OCR and other techniques, comparing records to ensure proper billing.
Incremental Automation and Future Prospects
“You start to layer different use cases on top of each other,” Rose says. While AI doesn’t do everything perfectly, it enables gradual automation. “We started at 20% automation, then 40%, then 60, 80%,” Rose explains, highlighting the potential for increased efficiency over time.
Looking ahead, Rose foresees discussions focusing on cost and portability. He anticipates a shift away from radical leaps in modeling capabilities towards more sustainable cost structures. “The last mile is: How do we get these things to run on a laptop or a phone instead of running them in a data center?” he posits.
DIY AI: A Controversial Yet Practical Approach
At the upcoming AI & Big Data Expo, Rose is eager to discuss another controversial opinion: encouraging businesses to move away from SaaS providers when feasible. “It’s not as hard as it sounds,” he asserts. With many companies already having a cloud presence, Rose suggests leveraging cloud tools offered by major providers to implement AI workloads without additional software licenses or training.
Once businesses have taken these initial steps, JBS Dev stands ready to support them in the next phases of their AI journey.
Watch the full interview with Joe Rose below:
For further insights, visit the full article Here.
“`
