AI Chatbots with Outdated Data: Government’s Executive AI Assistants Unveiled

In a meeting held today by the Vice Presidency for Science, Technology, and Knowledge-Based Economy, specialized AI assistants for government agencies were unveiled. However, the event highlighted significant shortcomings in these chatbots, which still face a long road to achieving practical efficiency.

According to Digiato, the Vice Presidency has taken on the responsibility of overseeing the progress of AI projects within government institutions. These systems are designed to process both general and specific information tailored to the needs of each agency. Nevertheless, today's ceremony revealed critical weaknesses, including outdated data and insufficient information. Some chatbots were found citing laws and documents from as far back as 2016 when responding to inquiries.

"I only answer relevant questions!"

Hossein Afshin, Deputy for Science, Technology, and Knowledge-Based Economy, showcased various prompts during the meeting to demonstrate how these chatbots respond. The results indicated widespread issues in their accuracy and presentation of information. Many struggled to understand questions correctly or provide coherent answers.

Some of these assistants have been intentionally restricted in their capabilities, a limitation that Afshin criticized. He remarked, "AI must address user needs effectively. Merely reiterating regulations isn't true smart technology; it should also offer general assistance."

Initial Release of Executive AI Assistants

A notable aspect of these projects is that the AI assistants will not source their data from news outlets or public databases. Instead, their information must be uploaded by administrators, creating heavy reliance on government agencies' willingness to share data, a commitment whose extent remains unclear.

Language Model or Framework?

Today’s meeting resembled more of a trial than an official launch, a test of performance that arguably should have been conducted privately with developers alone. The reason: these systems currently lack the readiness and competence required for deployment in government operations.

Given the government's intention to use these AI tools for decision-making processes, precision and accurate responses are crucial. This heightened sensitivity complicates their development further. During the session, a representative from the Ministry of Foreign Affairs disclosed difficulties with their "AI Media Assistant," which failed to perform adequately in practice.