The entertainment industry is witnessing a major shift as artificial intelligence (AI) advances rapidly in content creation.
Tyler Perry, the renowned US film and TV mogul, on Friday, February 23 made headlines when he halted an $800 million (Approximately Rwf1.02 billion) expansion of his Atlanta studio complex in response to the capabilities demonstrated by OpenAI’s video generator, Sora.
Perry’s decision underscores a growing concern within the industry that AI advancements could lead to the displacement of traditional roles and reshape the content creation process as we know it.
ALSO READ: There’s more to ChatGPT than just academic dishonesty
Recent years have witnessed advancements in Generative AI tools, such as DALL-E and Midjourney. These tools have rapidly progressed to the point where they produce images nearly indistinguishable from photographs. This progress has also presented challenges, including the proliferation of disinformation online and concerns among digital artists who find it increasingly difficult to secure work.
The landscape of Generative AI is competitive, OpenAI and Google are considered as frontrunners, continuously innovating to redefine content creation.
“Gen AI” refers to Generative Artificial Intelligence, a subset of AI technologies designed to generate content autonomously, often mimicking human creativity and imagination. This branch of AI includes systems capable of producing text, images, audio, and even videos without direct human input. Generative AI models learn from vast datasets and are trained to create new content based on patterns and structures present in the data.
OpenAI, in particular, unveiled its latest breakthrough: the Sora video-generation model. Sora represents a leap in Gen AI technology, enabling the generation of photorealistic videos based on text prompts. With the ability to create complex scenes, depict vivid emotions through characters, and even fill in missing frames in existing videos, Sora heralds a new era of content creation.
Despite the capabilities of Sora, OpenAI remains cautious, acknowledging the need to comprehend the potential risks associated with the technology. As part of this effort, the company collaborates with a select group of academics and researchers to assess the system’s vulnerabilities, a process known as “red teaming.”
To address concerns about the potential misuse of Sora-generated content, OpenAI is implementing watermarks to identify AI-generated videos. However, the company recognizes the challenges in identifying such content, as watermarks can be removed and detection can be complex.
Google’s AI evolution seems chaotic
On February 8, Google announced its AI model ‘Bard’, which has undergone a transformation into Gemini and was introduced alongside a new mobile app. Shortly after its release, Gemini was temporarily halted from generating images on February 21.
Google assured users of their plan to reintroduce an improved version soon, prompting speculation about the company’s AI strategies and why they released what some consider an “incompetent model.”
Google faced criticism for what it acknowledged as “inaccuracies in some historical image generation depictions” produced by the Gemini AI tool. Concerns arose when specific figures, such as the US Founding Fathers, and groups like Nazi-era German soldiers, were portrayed with inaccurate racial characteristics.
This led to speculation that Google’s efforts to promote racial and gender diversity through its AI may have inadvertently resulted in inaccurate depictions.
Despite Google’s efforts, social media discussions are still questioning the platform’s ability to produce historically accurate results, with some critics suggesting: “a deliberate avoidance of depicting white individuals.” This has sparked debates concerning the extent to which AI may perpetuate stereotypes and biases and Google refrained from explicitly identifying specific images it deemed erroneous in its communications.
It is conceivable that Gemini aimed to enhance diversity overall due to the persistent lack of diversity in Generative AI. Image generators are trained on vast datasets of images and text descriptions to generate the most suitable output for a given prompt, making them susceptible to reinforcing stereotypes. A 2023 Washington Post investigation revealed that prompts such as “a productive person” predominantly yielded images of white and male figures, while prompts related to “a person at social services” tended to depict individuals of color, indicating a continuation of patterns observed in search engines and other software systems.