Panasonic Develops AI Technology to Combat Fake Detection

2023-08-03

Panasonic has developed FlowEneDet, an AI model designed to enhance reliability by detecting unknown objects and improving AI technology for commercial use. Panasonic Research America and Panasonic Corporation have developed a new technology to improve the reliability of their AI models by detecting "unknown objects" that are unlearned and inherently unidentifiable. The company has developed a new flow-based generative AI model called "FlowEneDet" to enhance reliability by extending pre-trained segmentation models. Panasonic also stated that they are committed to accelerating the implementation of AI in society and advancing research and development of AI technology to ultimately assist their customers. The Power of Responsibility in AI AI has a history of false detection, leading researchers and developers to seek improvements in models. In particular, in July 2023, a published study showed that an AI GPT detector incorrectly classified articles written by humans as articles written by AI, raising concerns about bias and academic fairness. As more technologies focus on increasing the reliability of AI, it is interesting to see how these developments will contribute to achieving more responsible AI within enterprises. Responsible and efficient AI has been praised for its revolutionary significance to businesses. This includes initiatives such as cutting-edge model forums aimed at promoting AI security research, establishing best practices for identity security, and sharing knowledge to enhance responsible AI development and utilize AI to address societal challenges. It is also important to consider how to improve AI to adapt to specific business needs. Panasonic's flow-based detector has a simple energy-based input, as the model can correctly identify trained and inherently recognizable objects by predicting the "reliability" of AI model recognition results. The company proposes FlowEneDet, a flow-based generative model that serves as a network to estimate the confidence of AI models in their recognition results. It is a flow-based generative model that represents complex distributions as combinations of "invertible transformation functions," accurately modeling the distribution of training data. This modeling allows for differentiation between "objects known to the AI model" and "unknown objects." According to Panasonic, FlowEneDet is a model that handles low-dimensional free energy in energy models, thus having a low-complexity structure and the ability to handle unknown objects. It can also simultaneously handle misclassifications within the distribution without significantly increasing training and evaluation costs. Source: https://aimagazine.com/ai-applications/panasonic-develop-ai-technology-to-combat-false-detection

RECENT AI NEWS

RECENT AI TOOLS