Capabilities of large-scale pre-trained AI models continue to skyrocket, as demonstrated by large-scale vision-language models like CLIP or ChatGPT.

These typical generalist models can perform reasonably well in tasks covering a large variety of fields, which paved the way for their widespread adoption. However, such versatility no doubt comes at a cost. Training and operating large-scale models consume extreme amounts of energy and time, which goes against sustainability goals and limits the types of computers they can end up deployed on.

Moreover, in practical applications, people want AI models to fulfil specific roles rather than be jacks-of-all-trades. In such cases, a model’s generalist capabilities might be useless and even counter-productive, reducing accuracy. That could change as there is now a way to leverage large-scale pre-trained models more efficiently by having them “forget” unnecessary information.

To that end, a new methodology is in development called “black-box forgetting,” by which one can iteratively optimize the text prompts presented to a black-box vision-language classifier model to have it selectively “forget” some of the classes it can recognize, according to a research team led by Associate Professor Go Irie from Tokyo University of Science (TUS), Japan.

Total Classification Rarely Required
“In practical applications, the classification of all kinds of object classes rarely ends up required. For example, in an autonomous driving system, it would be sufficient to recognize limited classes of objects such as cars, pedestrians, and traffic signs. We would not need to recognize food, furniture, or animal species,” Irie said. “Retaining the classes that do not need to be recognized may decrease overall classification accuracy, as well as cause operational disadvantages such as the waste of computational resources and the risk of information leakage.”

Schneider Bold

Although some methods for selective forgetting in pre-trained models do exist, these assume a white-box setting, where the user has access to the internal parameters and architecture of the model. More often than not, users deal with black-boxes; they do not have access to the model itself or most of its information due to commercial or ethical reasons. Thus, the researchers had to employ a derivative-free optimization strategy – one that does not require access to the model’s gradients.

Furthermore, they extended a method known as CMA-ES, with the image classifier model CLIP as the target model for this study. This evolutionary algorithm involves sampling various candidate prompts to feed to the model and evaluating the results via predefined objective functions, updating a multivariate distribution based on the calculated values.

However, the performance of derivative-free optimization techniques deteriorates quickly for large-scale problems. As more classes need to end up forgotten, the “latent context” used to optimize the input prompts grows to unmanageable sizes.

New Parametrization Technique
To address this issue, the research team came up with a new parametrization technique called “latent context sharing.” This approach involves decomposing latent context derived from prompts into various smaller elements, considered to be “unique” to a prompt token or “shared” between multiple tokens.

AI Learns to Forget

Selective forgetting aims to reduce the classification accuracy for classes to end up forgotten while maintaining the accuracy for the classes to remain remembered. The proposed method, which targets the image classifier model CLIP, achieves selective forgetting by optimizing the input text prompt, since the model itself is a “black box.”
Source: Go Irie from Tokyo University of Science


By optimizing aiming to optimize for these smaller units rather than large chunks of latent context, the dimensionality of the problem can greatly reduce, making it much more tractable.

The researchers validated their approach using several benchmark image classification datasets, trying to get CLIP to “forget” 40 percent of the classes in a given dataset. This marks the first study in which the goal is to have a pre-trained vision-language model fail to recognize specific classes under black-box conditions and, based on reasonable performance baselines, the results were very promising.

This method has important implications in the field of artificial intelligence and machine learning. It could help large-scale models perform better in specialized tasks, extending their already astounding applicability. Another use, for example, would be to prevent image generation models from producing undesirable content by having them forget specific visual contexts.

In addition, the proposed method could help tackle privacy issues, which are a rising concern.

‘Selective Forgetting’
“If a service provider is asked to remove certain information from a model, this can be accomplished by retraining the model from scratch by removing problematic samples from the training data. However, retraining a large-scale model consumes enormous amounts of energy,” Irie said. “Selective forgetting, or so-called machine unlearning, may provide an efficient solution to this problem.”

In other words, it could help develop solutions for protecting the “Right to be Forgotten,” which is a sensitive topic.

This approach not only empowers large-scale AI models but also safeguards end users, paving the way for seamless integration of AI into our daily lives.

Pin It on Pinterest

Share This