One of the most important things that a lot of social media platforms have started looking into has to do with content moderation and the like. With vaccine hesitancy becoming much worse due to the widespread nature of misinformation and disinformation, it has become clear that the use of human moderators might not be enough to help improve the quality of content that users can end up taking advantage of. AI moderators can be effective but they require months of preparation and data so that they can learn about what types of content need to be cracked down on.
With all of that having been said and now out of the way, it is important to note that Meta has developed a new AI that it is calling the “Few-Shot Learner” which would make it so that moderation can be implemented in weeks instead of months. It requires far fewer training examples with all things having been considered and taken into account, and that means that the gap between AI development and content moderation can end up being greatly reduced which might help prevent the spread of misinformation and hateful ideologies that are currently so rife on numerous social media platforms.
It is a much more generalized system of learning, which is helpful because of the fact that this is the sort of thing that could potentially end up avoiding the use of specific labels and instead training AI based on examples with a wide range of uses. Based on the tests that Meta has conducted so far, FSL is doing a lot better than other forms of AI moderators by around 55% which might lead to a more intellectually sincere experience for the vast majority of social media users that are out there right now.
The presence of logic sequences in this AI is another fact that should be considered. It has allowed this AI to spot incidences of policy violations that other AI missed out on, and in some cases it even managed to exceed the performance of human moderators. AI needs to be improve if social media is to become a wholesome space for all, and this is a step in the right direction. However, Facebook is known for using AI and algorithms to often marginalize certain communities so that is something that people might be concerned about with FSL as well to an extent.
Read next: Facebook And Instagram Congratulate Their Communities With A Year Well Spent Via A Pair Of Year-End Appreciations
With all of that having been said and now out of the way, it is important to note that Meta has developed a new AI that it is calling the “Few-Shot Learner” which would make it so that moderation can be implemented in weeks instead of months. It requires far fewer training examples with all things having been considered and taken into account, and that means that the gap between AI development and content moderation can end up being greatly reduced which might help prevent the spread of misinformation and hateful ideologies that are currently so rife on numerous social media platforms.
It is a much more generalized system of learning, which is helpful because of the fact that this is the sort of thing that could potentially end up avoiding the use of specific labels and instead training AI based on examples with a wide range of uses. Based on the tests that Meta has conducted so far, FSL is doing a lot better than other forms of AI moderators by around 55% which might lead to a more intellectually sincere experience for the vast majority of social media users that are out there right now.
The presence of logic sequences in this AI is another fact that should be considered. It has allowed this AI to spot incidences of policy violations that other AI missed out on, and in some cases it even managed to exceed the performance of human moderators. AI needs to be improve if social media is to become a wholesome space for all, and this is a step in the right direction. However, Facebook is known for using AI and algorithms to often marginalize certain communities so that is something that people might be concerned about with FSL as well to an extent.
Read next: Facebook And Instagram Congratulate Their Communities With A Year Well Spent Via A Pair Of Year-End Appreciations