Since the beginning of this year, there has been a significant focus on artificial intelligence. As we now find ourselves in the last four months of 2023, an increasing number of companies are discussing AI and actively working to harness this emerging technology to their advantage. While OpenAI, Google, and Microsoft have taken the lead in the AI race, Apple and Meta have maintained a relatively low profile in this regard. Recent reports, however, indicate that both Apple and Meta are making substantial investments in the field of AI.
Apple, in particular, has been reported to be investing millions of dollars daily in AI projects and is actively involved in multiple AI initiatives. On the other hand, Meta has entered the scene with ambitions to develop an AI model that can compete with OpenAI’s ChatGPT-4.
According to a report originally attributed to The Wall Street Journal and covered in The Verge, Meta’s upcoming AI system appears poised to surpass OpenAI’s most advanced model, the GPT-4. However, Meta is also taking steps to reduce its reliance on Microsoft for AI training.
Meta’s forthcoming AI system is set to commence training in early 2024. With this new large language model (LLM), Meta aims to introduce an AI system even more powerful than OpenAI’s GPT-4. Currently, GPT-4 powers ChatGPT’s premium version, showcasing capabilities such as coding, poetry composition, problem-solving, and more.
Reports indicate that Meta has acquired additional Nvidia H100 AI-training chips and is actively improving its infrastructure to lessen dependence on Microsoft’s Azure cloud platform for AI training. Earlier this year, a dedicated team was assembled to expedite the development of AI tools capable of replicating human expressions.
The specific features and level of capability of Meta’s new AI chatbot in comparison to ChatGPT-4 remain a matter of anticipation. Only time will reveal the extent of its capabilities.
Turning our attention to ChatGPT-4, it was unveiled in March of this year as “a large multimodal model capable of processing both image and text inputs while generating textual outputs.” While it may not match human abilities across all real-world scenarios, it exhibits human-level performance in numerous professional and academic benchmarks.
At its launch, GPT-4, the successor to GPT-3.5, demonstrated its ability to handle more extensive and intricate queries compared to its predecessor. OpenAI also disclosed a list of exams that GPT-4 had successfully passed at the time of its introduction, along with its respective scores. Impressively, it achieved an 88th percentile in the LSAT, 89th percentile in SAT Math, 80th percentile in GRE Quantitative, and 99th percentile in GRE Verbal, among other notable scores.