Artificial intelligence models that emerged after the launch of “Chat GT” suffer from a transparency deficiency in their design, posing a risk to applications that rely on them, according to a study conducted by Stanford University.
A recent study conducted by researchers at the American University of California showed that the most transparent model among the ten models evaluated is Llama 2. Llama 2 is an artificial intelligence system launched by Meta group in July last year and can be reused.
However, this model is deemed insufficient as it only provides 54% of the required transparency level, which is inadequate compared to the level that needs to be achieved to be considered transparent according to the study authors.
The transparency level of the product “Chat GT-4”, which is the main product of the manufacturer “Open AI” of the famous robot “Chat GT” funded by “Microsoft”, does not exceed 48%.
Other less-known models received wide acclaim, including “Balm-2” from Google and “Cloud 2” from Anthropic (funded by Amazon).
In a statement, Rishi Bomassani, research director at Stanford University, pointed out that the desired transparency in “Foundational models” should range between 80% and 100%.
The study indicated that most companies do not disclose the extent of protecting the copyrights and users in training their models. Additionally, companies do not divulge the use of human labor to correct training data, which could pose a significant problem.
I deeply regret that “any company that does not provide information about the number of users relying on its models, or statistics related to the countries or markets that use them.”
The study participants explained that transparency indicators could be used in the future by policymakers and legislators. The European Union, the United States, China, the United Kingdom, and Canada have expressed their interest in enhancing technology and artificial intelligence transparency.
Last July, US President Joe Biden expressed his vision to industry leaders in this sector, stating that artificial intelligence represents a great promise with amazing opportunities, but it also poses a threat to our society, economy, and national security.
This topic topped the agenda of the Group of Seven summit in Japan in May, and Britain is set to host an international summit on artificial intelligence in November.
