Meta has used artificial intelligence (AI) to automate the product risk assessment procedure that has been conducted in improving the function of the platform and modifying algorithms. In consequence, development efficiency is anticipated to extend, but there may be concern that unexpected negative effects and risks will increase.
The NPR reported on Monday (31) that it has automated update functional updates and algorithms in Instagram, Facebook, and WhatsApps, that are the core services, to automate as much as 90%of the product risk assessment procedures which have been carried out through the change of function updates and algorithms.
Within the meantime, to be able to prevent personal information infringement, protection of youngsters, and spread of harmful content, human beings are replaced by human beings’ privacy and integrity review.
In line with the interior document obtained by the NPR, in the longer term, the AI ​​system is ‘immediately approved’ by the AI ​​system just by creating a straightforward questionnaire with a straightforward questionnaire. On this process, you may mechanically analyze the danger areas and response tasks, after which check that the event team has met it.
These changes are advantageous to cut back the product development cycle and advance the time of launch, and engineers can autonomously launch their products.
Nevertheless, it was identified as an issue that human ethical judgments were step by step excluded. A former meta executive said, “If more functions are released faster in the alternative and opposite, the danger will likely be increased.”
In reality, sensitive areas resembling privacy, AI safety, child protection, false information and violence contents were also included within the automation goal. Previously, all major updates needed to be reviewed by skilled risk evaluators, but now the product developer will determine whether to request a manual review.
Meta explained that this measure is to simplify decision -making, and “complex and recent issues are still involved in human experts, and automation is being made just for low -risk decisions.” As well as, the European Union (EU) users explained that the identical level of surveillance is maintained in consideration of strong regulations resembling the Digital Service Act (DSA).
Nevertheless, Meta has continued to deregulate the platform by terminating the very fact check program and relieving the hate speech policy. In a recent announcement, “in some policy areas, large language models (LLM) are making higher judgments than people.”
“AI may help to cut back overlapping reviews and speed up redundant reviews, but human checks and balance should be balanced.”
By Park Chan, reporter cpark@aitimes.com
