Wrong Challans Issued Due to AI Traffic Camera Errors in India

The use of Artificial Intelligence (AI) powered traffic cameras for the automated enforcement of traffic rules is increasing in India, but significant flaws in this technology are coming to light. According to a report by The Times of India, a large number of incorrect challans have been sent to motorcycle and car drivers due to errors in this system.

Instances of misidentification of objects by these cameras have emerged. According to the report, in some cases, the camera identified a guitar as a pillion rider, leading to the imposition of a fine. Similarly, a turban worn by a person was identified by the system as a bare head (without a helmet), which was treated as a violation of traffic rules.

There has been a sharp rise in complaints from motorists regarding the accuracy of this technology. These incidents have raised serious questions about the accountability and transparency of the system. To reduce the hardships faced by motorists, it has become imperative to ensure accuracy alongside the development of the technology.

On the other hand, significant steps have been taken to develop indigenous AI capabilities in the country through the 'IndiaAI Mission' under the Ministry of Electronics and Information Technology (MeitY), Government of India. According to a report by The Hindu, the ministry recently launched three indigenous foundation models: EKA-1, Dhīr 2B, and Dhīr 5B. These models were unveiled in the presence of IT Secretary S. Krishnan and other senior officials.

The EKA-1 model, developed by Soket Labs Technology and Research, has 26.3 billion parameters and has been trained on approximately 7.5 trillion tokens. It covers all 22 scheduled Indian languages, including English. Abhishek Aparwal, CEO of Soket, stated that this model has been specifically designed keeping in mind the requirements of India's banking, financial services, and cybersecurity sectors. This model can operate on local devices such as smartphones even without mobile data or Wi-Fi.

Dhīr 2B and Dhīr 5B, developed by ZenteiQ AiTech Innovations, are open-source models that support English and 10 Indian languages. Prof. Shashikumar, CEO of ZenteiQ, stated that these models can run on various computing platforms such as GPU and TPU. Due to its lightweight architecture, it can function on mobile phones without cloud connectivity, thereby improving data privacy and security. These models can be used in defense, government, industrial, and educational sectors, especially where connectivity is low.

The government stated that the development of these models was made possible due to subsidies received for national AI compute infrastructure and research expenditure. According to Secretary S. Krishnan, such indigenous models will strengthen India's technological and strategic autonomy. He added that these models would be made available to researchers, students, and the general public through platforms like AIKosh to enable further experimentation.

The IndiaAI Mission is a project of the Government of India aimed at enhancing Artificial Intelligence capabilities in the country and preparing indigenous Large Language Models (LLMs).