Thank you so much for sharing this post and for summarizing the great achievements of 2023; there has been significant progress.
I am Mohamed Elhoseiny, faculty at KAUST and the PI of the Vision CAIR group. I truly appreciate your featuring of MiniGPT v2, which was developed by my team as an advancement over our earlier version, MiniGPT-4 v1, released in April 2023, which in fact is MiniGPT-4's key milestone. MiniGPT-4 v1 was launched on April 16, 2023, just days before LLaVA was announced (announcement: https://twitter.com/tikgiau/status/1647767975804452864, example third party tech Youtubers discussing it mostly around 18-19 April 2023; https://www.youtube.com/watch?v=hNAFuuXYL58
Of-course being days apart, LLaVa and MiniGPT-4 are indeed contemporary works, and they have been recognized as such in most follow-up papers in the Vision LLM space.
I was initially hesitant to reach out with this comment, but I felt it is helpful to clarify that MiniGPT v1 (16 April 2023) represents a main MiniGPT milestone and key findings for promising/emerging Vision LLM skills. As you know, these matters can be of importance in our field. The difference between it and MiniGPT v2, is that v2 extends it to be massively multitasking and to be capable of grounding objects as it speaks about them.
Thanks a lot for your efforts writing this amazing post. I do appreciate it and I believe it is much appreciated by the vast majority of the Generative AI community.
Great sources for insights on AI papers also from Elvis Saravia, Sebastian Raschka, PhD, Davis Blalock, Ahsen Khaliq, Santiago Valdarrama and so many other great ML scientists like Nathan Lambert. Find them on LinkedIn, X and Threads.
Thank you so much for this valuable piece. For people in the edtech world like myself, the 2023 AI advancements, exemplified by InstructPix2Pix and Mistral 7B, offer game-changing potential. These innovations pave the way for personalized learning experiences, dynamic content creation, and inclusive education through features like text-to-audio models. The prospect of leveraging AI's reasoning abilities for assessments holds promise in understanding students' problem-solving skills. But this is a watershed moment. We must treat it with care.
Thanks, Sahar, for your tireless efforts. The online tech and AI community is smarter and can respond more quickly to changes in the industry as a consequence of your research. Keep up the amazing work!
Dear Michael and Sahar,
Thank you so much for sharing this post and for summarizing the great achievements of 2023; there has been significant progress.
I am Mohamed Elhoseiny, faculty at KAUST and the PI of the Vision CAIR group. I truly appreciate your featuring of MiniGPT v2, which was developed by my team as an advancement over our earlier version, MiniGPT-4 v1, released in April 2023, which in fact is MiniGPT-4's key milestone. MiniGPT-4 v1 was launched on April 16, 2023, just days before LLaVA was announced (announcement: https://twitter.com/tikgiau/status/1647767975804452864, example third party tech Youtubers discussing it mostly around 18-19 April 2023; https://www.youtube.com/watch?v=hNAFuuXYL58
https://www.youtube.com/watch?v=SAjrpYjx0ps ).
Of-course being days apart, LLaVa and MiniGPT-4 are indeed contemporary works, and they have been recognized as such in most follow-up papers in the Vision LLM space.
I was initially hesitant to reach out with this comment, but I felt it is helpful to clarify that MiniGPT v1 (16 April 2023) represents a main MiniGPT milestone and key findings for promising/emerging Vision LLM skills. As you know, these matters can be of importance in our field. The difference between it and MiniGPT v2, is that v2 extends it to be massively multitasking and to be capable of grounding objects as it speaks about them.
Thanks a lot for your efforts writing this amazing post. I do appreciate it and I believe it is much appreciated by the vast majority of the Generative AI community.
Thanks for clarifying Mohamed and appreciate the great work you guys do at KAUST!
We've been alerted to another writer who writes on AI papers: https://recsys.substack.com/
Great sources for insights on AI papers also from Elvis Saravia, Sebastian Raschka, PhD, Davis Blalock, Ahsen Khaliq, Santiago Valdarrama and so many other great ML scientists like Nathan Lambert. Find them on LinkedIn, X and Threads.
Thank you so much for this valuable piece. For people in the edtech world like myself, the 2023 AI advancements, exemplified by InstructPix2Pix and Mistral 7B, offer game-changing potential. These innovations pave the way for personalized learning experiences, dynamic content creation, and inclusive education through features like text-to-audio models. The prospect of leveraging AI's reasoning abilities for assessments holds promise in understanding students' problem-solving skills. But this is a watershed moment. We must treat it with care.
Thanks, Sahar, for your tireless efforts. The online tech and AI community is smarter and can respond more quickly to changes in the industry as a consequence of your research. Keep up the amazing work!
Thanks for the kind words, Nick. Same to you and looking forward to reading more from Educating AI in 2024.
Thanks for sharing the list in timeline order.
This helps a lot for someone who want to understand what's happening in AI and LLM world.
Some Newsletters truly are rather educational, here are 40 I really like: https://aisupremacy.substack.com/p/top-ai-newsletters-of-2024-on-substack