(US & Canada) VIDEO | We’ve Built Our Own ChatGPT-like Prototypes With Promising Results – USPS OIG CDO and AIG

Ben Joseph, CDO and AIG at USPS OIG, speaks with Michael Hughes, Chief Business Officer of Duality Technologies, in a video interview about the initiative USPS OIG has taken to leverage technology, operationalizing generative AI models and the organization's approach to leveraging AI and ML.

Joseph shares that there have been massive technological successes in the past two years and appreciates the solid leadership support. Adding on, he discusses experimenting and putting value in the hands of customers. In terms of AI and ML, the organization has devised the strategy of going with the crawl, walk, and run.

Firstly, Joseph continues, the organization has started upskilling its existing personnel while bringing in new people. The second focal area is tech acquisition and he states that the organization has brought in the best of tools and modernized the architecture.

Joseph stresses that the organization needs to know how to use the tools, what type of architecture needs to be put in place, and then create a safe experimental space for data scientists. He mentions a target state architecture that is in the process of being built out.

The third crucial aspect Joseph says is governance and processes that involve the policies, guidelines, and standards required to be put in place, especially with the advent of generative AI. Therefore, the agency has come up with guidelines that are acceptable and specific to the organization’s needs.

Additionally, Joseph mentions working with partners and the Chief Information Officer to ensure enabling tools like Chat, Cloud, and Gemini for business operations like the usage of Copilot. He opines that one should be able to touch and feel the solutions built, and considers architecture as a solution.

Speaking of challenges, Joseph discusses the overexposure of OpenAI, and how it can be differently interpreted. Therefore, the organization has decided to put it aside till it gains more confidence in its usage.

Addressing the generative AI space, Joseph mentions using open-source models to build organizational ChatGPT-like prototypes. To do this the agency ingested its data and tried to ask questions, which showed promising results.

Then, the agency did a feasibility test to see where it stands and decided to wait it out with generative AI due to the high cost of operationalizing.  As far as the traditional AI and ML space is concerned, the organization has made different efforts and seen results by incorporating them into current tools empowering end-users.

Elaborating further, Joseph mentions anomaly detection, utilizing traditional NLP, and making headway into Computer Vision. In conclusion, with AI and ML, he affirms that it is better to buy a product off the market and build on top of it.

CDO Magazine appreciates Ben Joseph for sharing his insights with our global community.