The field of artificial intelligence (AI) is witnessing significant advancements and shifts, particularly in the realm of language learning models (LLMs). Recent developments indicate that the performance gap between closed-source and open-source LLMs is narrowing, with a current lag of only 6 to 10 months compared to the years-long gap observed during the release of GPT-4. Furthermore, the first open LLM, WizardLM 2, has surpassed OpenAI's GPT-4 in the MT-Bench benchmark, highlighting the rapid progress in open-source AI technologies. This model, based on the Mixtral 8x22B architecture, is licensed under Apache 2.0 and represents a significant milestone in AI development. Additionally, discussions in the AI community focus on whether GPT-5 will mark another significant leap forward, the potential convergence of open-source and closed-source models, and the overall trajectory of AI progress. It is also noted that Supervised Learning will continue to play a dominant role in AI's development, and the current Auto-Regressive LLMs may not achieve human intelligence surpassing in the near future.
We can do it! 🙌 First open LLM outperforms @OpenAI GPT-4 (March) on MT-Bench. WizardLM 2 is a fine-tuned and preferences-trained Mixtral 8x22B! 🤯 TL;DR; 🧮 Mixtral 8x22B based (141B-A40 MoE) 🔓 Apache 2.0 license 🤖 First > 9.00 on MT-Bench with an open LLM 🧬 Used multi-step… https://t.co/XcixP226Cz
There is no question that AI will eventually reach and surpass human intelligence in all domains. But it won't happen next year. And it won't happen with the kind of Auto-Regressive LLMs currently in fashion (although they may constitute a component of it).…
This chart from shows the 3 big questions of AI right now (even given that the limits of MMLU as a test): 1) OpenAI was very far ahead, now it isn’t. Will GPT-5 be another leap? 2) Will open source models converge with closed source? 3) Will rapid progress continue or hit a wall? https://t.co/Halw8dvagK