Martin Signoux of Meta predicts AI model development in 2024

Martin Signoux, a public policy expert at Meta France, recently shared his views on the future of AI models in a series of tweets. His insights, focusing on developments expected in 2024, have received significant attention. sister-in-law prediction Topics range from the emergence of Large Multimodal Models (LMMs) to the ongoing debate between open and proprietary AI models.

Signoux begins by discussing the transition from large language models (LLMs) to LMMs. He predicts that LMMs will soon dominate AI conversations, citing their role as a stepping stone toward more generalized AI assistants. He doesn’t expect major breakthroughs, but predicts that iterative improvements to various AI models will improve their robustness and usability for multiple tasks. These improvements, including advances in Retriever-Augmented Generation (RAG), data curation, fine-tuning, and quantization, will drive adoption across a variety of industries.

Another key point raised by Signoux is the growing importance of Small Language Models (SLMs). He suggests that cost-effectiveness and sustainability considerations will accelerate the trend towards SLM. He also anticipates significant advances in quantization, which will drive integration within devices for consumer services.

Regarding the open versus closed model debate, Signoux predicts that open models will soon outperform models like GPT-4. He acknowledges the contributions of the open source community to AI development and foresees a future where open models coexist with proprietary models.

Signoux also highlights the challenges of benchmarking AI models. He believes that by 2024, no single benchmark or assessment tool will emerge as the final standard, especially for multimodal assessments. Instead, there will be various improvements and new plans.

According to Signoux, the public debate will shift from existential risks to more immediate concerns related to AI. These concerns include issues of bias, fake news, user safety, and election integrity.

Responses to Signoux’s thread show a variety of opinions. For example, John Smith predicts that LMMs will have lower reasoning power than LLMs on a token-by-token basis. David Clinch emphasizes the importance of appropriate context and rights management, suggesting that LLMs and LMMs should be granted access to valuable journalism and media.

Image source: Shutterstock