It’s time for artificial intelligence (AI)! After Google and Snapchat, it is now Meta’s turn (Facebook, Instagram, Whatsapp) to get involved with generative artificial intelligence. Since the emergence of ChatGPT, this technology has evolved to the point that tech giants now consider it crucial to their development strategy. How Meta’s AI will stand out from the competition remains to be seen…
Given the success of tools like ChatGPT, Dall-E or Midjourney, generative artificial intelligence (AI) is more than indispensable for large companies that want to assert their position in the tech scene.
The global AI market is currently valued at over $433 billion, which is why Meta, the parent company of Facebook, Instagram and Whatsapp, has decided to start developing its own generative AI.
They are nom: CALL.
Mark Zuckerberg, CEO of Meta, announced it himself on his Facebook account:
“We will create a new high-level product group at Meta, based on Generative AI, to increase our efforts in this area. There’s still a lot to do before we get to some truly futuristic experiences, but I’m excited for all the new things we’ll build.”.
Behind their AI, the idea of this platform, which brings together 77% of users, is simple: develop creation tools and create “AI characters” that will be used to help people in “different ways”. This likely means that the project coincides with another corporate project that hasn’t had the success it expected: the Metaverse. LLaMA is currently only available to scientists.
Tools to be corrected focused on text, image and video (the digital factory).
For now, a new team formed by the Mark Zuckerberg company will work on:
- Text-centric, chatbot-like experiences (for WhatsApp or Messenger);
- images with Instagram filters;
- Creative advertising formats, also video or multimedia formats.
According to American website Axios, this new team will be led by Meta’s vice president in charge of AI and machine learning, Ahmad Al-Dahle.
Specifically, the California-based company can rely on Meta AI or Facebook AI Research (FAIR), its research unit specialized in artificial intelligence, to carry out its project.
Researchers in this field have already published several experiments using generative AI techniques in the last two years (the last two were Galactica and BlenderBot). However, these were not integrated into the company’s products. The team dedicated to integrating these tools with Facebook, Instagram and WhatsApp should be able to solve this problem.
Artificial intelligence is an error-prone field. Meta is aware of the risks they run when they want to bring their AI to market, whether it’s about factual errors or data security.
In the tech sector, many giants have already lost a lot by getting involved in this field. For example, we can cite the case of Google Bard, whose error cost the MountainView company $100 billion.
And remember, technical tests performed by Bing, Microsoft’s search engine, using OpenAI’s GPT-4 resulted in inconsistent responses from the AI.
For its part, Meta claims that additional research is essential to analyze and anticipate the risks of prejudice, toxic comments and hallucinations. For the Mark Zuckerberg company, considerable resources, which primarily concern computing power, are necessary in order to train the new language models as well as possible.
Therefore, Meta has preferred to present several versions of LLaMA that require more or less resources for the time being. They also made the decision to share the process of developing this AI so that researchers have the opportunity to effectively test new approaches to solving recurring problems such as bias, toxicity and the possibility of generating false information.
