Martin Signoux, a public policy expert at Meta France, recently shared his perspectives on the future of AI models in a series of tweets. His insights focused on developments expected in 2024 received considerable attention. of Blue predictions cover a range of topics, 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 expects that LMMs will soon dominate the AI conversation, citing their role as a stepping stone to more generalized AI assistants. While he doesn’t expect major breakthroughs, he predicts that iterative improvements in various AI models will improve their robustness and utility for multiple tasks. These improvements, including advances in Retriever-Augmented Generation (RAG), data processing, fine-tuning, and quantization, will drive adoption across industries.
Another key point that Signoux raises is the growing importance of small language models (SLM). He suggests that cost-effectiveness and sustainability considerations will accelerate the trend towards SLMs. In addition, he foresees significant advances in quantization that will facilitate device integration for consumer services.
Regarding the open versus closed model debate, Signoux predicts that open models will soon surpass the performance of models such as GPT-4. It recognizes the contributions of the open source community to the development of AI and envisions a future in which open models coexist with proprietary ones.
Signoux also highlights the challenges of benchmarking AI models. He believes that no single benchmark or assessment tool will emerge as the definitive standard in 2024, especially in multimodal assessments. Instead, there will be various improvements and new initiatives.
The public debate, according to Signo, will shift from existential risks to more immediate AI concerns. These concerns include issues of bias, fake news, consumer safety, and election integrity.
The replies to Signoux’s thread show differing opinions. John Smith, for example, expected LMMs to have less reasoning capacity than token-based LLMs. David Clinch suggests that the LLM and LMM license access to valuable journalism and media, emphasizing the importance of proper context and rights management.
Image source: Shutterstock