In the digital age we live in, content control has become one of the most pressing challenges for online platforms. With the evolution of artificial intelligence, OpenAI strongly believes in GPT-4's ability to address and potentially solve this problem. On its official blog, the company most in the spotlight of the period talks about a great way to change direction in the near future.
Topics of this article:
OpenAI's idea of content control is anything but stupid
OpenAI, the organization behind revolutionary software such as ChatGPT and DALL-E, is convinced that GPT-4 can be la key to address one of the most complex problems of the technology sector: the control of content at scale. The idea is that GPT-4 could replace a vast number of human moderators, ensuring almost comparable accuracy and greater consistency in decisions.
Recently, OpenAI revealed have already experienced the use of GPT-4 to develop and refine their content policies. This not only pertains to content labeling, but also making crucial decisions about what should and should not be allowed. Lilian Weng, head of security systems at OpenAI, underlined the importance of this evolution, stating that it represents a significant step in the way AI can be used to tackle real problems, offering solutions that can benefit the whole of society.
Read also: This company proves that AI won't steal our jobs…on the contrary
The three reasons why OpenAI's operation is revolutionary
- Consistency in decisions: unlike humans, who can interpret policies in different ways, machines are consistent in their judgments. Guidelines, which can be as long as a book and change frequently, can be implemented instantly by large language models, whereas humans would require a lot of training to learn and adapt.
- Speed in creating new policies: GPT-4 has the ability to develop a new policy in hours, revolutionizing a process that would traditionally take weeks or months. This includes drafting, labelling, gathering feedback and refining policies.
- Workers' welfare: OpenAI places emphasis on the well-being of workers who are continually exposed to harmful content, such as child abuse or torture videos. Automating this process could reduce human exposure to such traumatic content.
However, there is still a long way to go towards fully automated moderation. While AI has the potential to improve moderation, there are still many challenges ahead. For example, the distinction between deceptive, incorrect and aggressive content which are not necessarily illegal presents a major challenge for automated systems. Even human experts struggle to label such posts, and machines often make mistakes.
Danger to human work?
As usual, we always talk about this: will AI steal our jobs? For most companies that use content moderation, it's worth erring on the side of caution. Although artificial intelligence has advanced significantly in recent years, it won't take away human jobs any time soon. Human oversight remains a critical component to accurately monitor online content. So the answer is no, it won't steal the work but it will improve it.
However, it is important to be aware that such an evolution in technology must also correspond to an increase in the level of specialization of workers in the technological field. If this is missing, replacement is inevitable.
"At best, what we have and will continue to have is a hybrid. But in the last few years, all I've seen is an increase in hiring, not a decrease" said UCLA professor Sarah T. Roberts a The Verge several years ago. A successful content moderation strategy must combine the best of next generation technology and human agents.
