Llama 3: The Next Big Thing in AI?
A report in the Wall Street Journal suggests that Meta is developing Llama 3. This language model is being designed to compete directly with OpenAI's GPT-4. The news was inadvertently revealed by Jason Wei, an OpenAI engineer formerly with Google Brain, during a social event hosted by Meta.
Wei's subsequent tweet hinted at Meta's vast computational resources, suggesting the company is not only working on Llama 3 but also its successor, Llama 4. Interestingly, even if Llama 3 matches GPT-4's performance, it will remain open-source. This commitment was confirmed by a statement from a Meta representative, which read, “Yeah we will. Sorry alignment people.“
The AI Race: A Historical Perspective
Historical data reveals a five-month gap between the launch of Llama 1 in February 2023 and Llama 2 in July 2023. While Llama 2, with the help of the open-source community, has achieved performance levels similar to GPT-3.5 in certain applications, it still lags behind closed-source models like GPT-4 and Google's PaLM-2.
Despite this, Meta's open-source strategy has garnered attention and praise from the developer community. However, some skepticism remains, especially since the decision to open-source LLaMA was made post its leak. A study from Radboud University in Nijmegen, Netherlands, suggests that companies, including Meta, aren't entirely transparent about their AI models. The study calls for greater honesty and openness in the AI community.
Despite Meta's efforts, OpenAI remains a dominant player in the large language model market. OpenAI's CEO, Sam Altman, stated in June 2023 that the release of GPT-5 isn't on the immediate horizon. Meanwhile, Google is gearing up to launch Gemini, its next-gen multimodal LLM, either later this year or early next year.
Concerns and Implications
The move to open-source powerful AI models has raised eyebrows in the tech community. OpenAI's Sam Altman has humorously mentioned having a “kill switch” for OpenAI's systems, emphasizing the potential risks of AI systems going rogue. With Meta possibly open-sourcing a GPT-5 level model, concerns about the absence of a “kill switch” for such models have emerged, raising fears about potential misuse by malicious entities.
In conclusion, the AI landscape is rapidly evolving, with major tech companies like Meta and OpenAI at the forefront. As these giants continue to innovate and push boundaries, the broader tech community eagerly awaits the next big breakthrough in AI technology.
A Race of Profit, Not Just Innovation
While tech companies are competing on a technological front, they are also competing for a market that is already worth billion dollars. Despite the relatively early days of generative AI, we can already see the economic potential for companies.
Recent analyses have unveiled that Nvidia is securing nearly a 1,000% profit on every H100 Tensor Core GPU it sells. Financial insights from Raymond James, a reputable financial services firm, shared on Barron's, have estimated the production cost of one such GPU to be around $3,320. In stark contrast, Nvidia's selling price for these GPUs fluctuates between $25,000 and $30,000, contingent on the order volume.
This year, OpenAI has become a prominent leader in the AI industry. After the launch of ChatGPT, the company quickly rocketed into the top 50 most viewed websites. Last month, OpenAI's monthly revenue grew to $80 million, surpassing expectations.
However, there is also more scpetic outlooks for the future of OpenAI. Despite the popularity of ChatGPT, the chatbot could be a threat to OpenAI‘s financial security. As the company pushes to be the leading generative AI developer in the world, it is losing around $700,000 a day maintaining ChatGPT.
In a report published by Analytics India Magazine, it was revealed that OpenAI's losses have doubled since it started developing ChatGPT. The company is now losing over $500 million per year, and it could go bankrupt by the end of 2024 if it doesn't find a way to turn a profit.