TikTok tops Google as most popular website of the year, according to data
DUBAI: Earlier this year, former Facebook employee Frances Haugen leaked internal company documents and accused the social media giant of prioritizing profit over safety by failing to tackle adequately to the spread of disinformation and harmful content.
The whistleblower also said that 87% of disinformation spending is dedicated to English speakers, although only 9% of Facebook users are English speakers.
Now, Facebook’s parent company, Meta, has developed a new artificial intelligence system called ‘Few-shot learner (FSL)’ which it says is faster, more efficient, and works in more than 100 languages, including ‘Arab.
“AI has to learn what to look for and it typically takes several months to collect and tag the thousands, if not millions, of examples needed to train each individual AI system to spot a new type of content,” the company said in a blog post. .
The new FSL system works in weeks, instead of months, learning from different types of data, such as images and text. It does this through a system called “step-by-step learning,” where the curriculum begins with a general understanding of various topics, then narrows it down to fewer examples to learn new tasks.
“By taking advantage of the few-shot mode early in the process, we can find data (samples) to prime the model more efficiently, allowing us to label more efficiently. These samples can then be fed back into FSL, and as FSL sees more it just gets better and better, ”a Meta spokesperson told Arab News.
So far, the new model has only been tested on a few issues. Although he is trained on all integrity-violating policies that are part of Meta’s community standards, he has been deployed on certain events such as misleading information around COVID-19 and hostile rhetoric related to bullying, to harassment, violence and incitement, the spokesperson added.
According to early testing, the company said, the new AI system was able to correctly detect posts that traditional systems may miss and helped reduce the prevalence of harmful content. Meta is currently working on additional tests to improve the system.
“We have seen that in combination with existing classifiers as well as efforts to reduce harmful content, continuous improvements in our technology and changes we have made to reduce problematic content in the News Feed, FSL has contributed reduce the prevalence of other harmful content such as hate speech. Said the spokesperson.
The FSL model shows how much Facebook, now Meta, relies on AI. In its latest earnings call, the company said it was forecasting capital spending of $ 29 billion to $ 34 billion next year, up from $ 19 billion this year.
David Wehner, CFO of Meta, said a major factor driving this increase is “an investment in our artificial intelligence and machine learning capabilities, which should benefit our ranking and recommendation efforts. for experiences on our products, including in feed and video, as well as improving ad performance and relevance.
Although FSL is in its early stages, “these early production results are an important milestone that signals a move towards smarter, generalized AI systems that can quickly learn multiple tasks at once,” Meta said in the post. blogger, adding that the long-term vision is to achieve “human-like flexibility and learning efficiency.”