Meta working on new AI model more powerful than ChatGPT 4, does not want to rely on Microsoft for training – India Today
Listen to Story
Since the beginning of this year, we have been talking a lot about artificial intelligence. And as we are now in the last four months of 2023, more and more companies are talking about AI and are working towards using the emerging tech for their benefit. While OpenAI, Google and Microsoft already seem to be leading the AI race, Apple and Meta are relatively quiet about it. It was recently reported that Apple is investing millions of dollars in the AI space every day and is working towards multiple AI projects of its own. And now, as per recent reports, Meta is also working to develop an AI model aimed at giving stiff competition to OpenAI’s ChatGPT-4.
According to a report in The Verge originally attributed to The Wall Street Journal, Meta’s upcoming AI system might be more powerful than OpenAI’s most advanced LLM currently- the GPT-4. And that’s not it, the company is also taking measures to ensure that it doesn’t have to rely on Microsoft for training its AI system.
Meta’s upcoming AI system
The report says that Meta plans on beginning training its upcoming AI model in early 2024. WIth this upcoming LLM (large language model), Meta is hoping to roll out an AI system that is even more powerful than OpenAI’s GPT-4. As of now, GPT-4, powering ChatGPT’s premium version, is the most sophisticated generative AI chatbot that can write code, compose poetry, tackle complex problems, and a lot more.
The report also reveals that Meta has bought more of Nvidia’s H100 AI-training chips and is also working on improving its infrastructure so that it doesn’t have to rely on Microsoft’s Azure cloud platform to train the upcoming chatbot. A group of people was reportedly put together earlier this year, to speed up creation of AI tools copying human expressions.
So what exactly will Meta’s new AI chatbot do and how powerful will it be when compared to ChatGPT 4? Time shall tell.
About ChatGPT 4
Coming to ChatGPT 4, it was introduced in March this year as “a large multimodal model (accepting image and text inputs, emitting text outputs) that, while less capable than humans in many real-world scenarios, exhibits human-level performance on various professional and academic benchmarks”.
At the time of its launch, GPT-4, successor to GPT 3.5, was said to handle larger and more complex queries when compared to its predecessor. OpenAI had also released a list of exams that GPT-4 had passed at the time of its launch and shared its scores as well. And the scores were pretty good. For instance, the language model passed LSAT with 88 percentile and SAT Math with 89 percentile. It also passed GRE Quantitative exam with 80th percentile and GRE verbal and writing exams with 99th and 54th percentile.
