According to that report, Meta intends for its new AI model to be several times more powerful than its current model, Llama 2. The company also expects that the new model will have similar capabilities to GPT-4, OpenAI’s competing model.
GPT-4 went live in March 2023 and was extended for general use in July 2023. Facebook’s competitor, however, will not begin training until 2024.
Current reports suggest that it will be possible to use Facebook’s upcoming model to create services that produce sophisticated text, analysis, and outputs.
Meta’s upcoming model will likely be released as an open-source product available at no cost, in line with the company’s current AI releases and stated goals.
The company is also building new data centers and acquiring chips required to perform training tasks. According to WSJ‘s reporting, Meta is buying NVIDIA H100 Tensor Core GPUs—the company’s most advanced model for AI training—which are in very high demand.
Meta currently offers Llama 2
Though Meta’s new AI release will be a significant step forward, it is not the company’s only release to date. After partial releases and source code leaks in early 2023, Meta fully released Llama 2 in July 2023 for research and commercial purposes.
Meta has numerous other AI efforts underway, many of which are aimed at the public. The company launched a developer assistant tool in August and announced AI-powered ad services in May. The company’s chatbots are also expected to go live soon.
Incidentally, Meta CEO Mark Zuckerberg and other tech leaders are set to discuss artificial intelligence with the U.S. senators in a series of upcoming forum events, according to CNBC. Other tech leaders set to attend include Tesla and Twitter/X executive Elon Musk, Microsoft co-founder Bill Gates, and OpenAI CEO Sam Altman. Those forum dates will begin in September and run on an ongoing basis.
The post Meta reportedly building a new AI model to compete against OpenAI’s ChatGPT appeared first on CryptoSlate.