Alibaba Is Mastering Human Video Generation
Plus: Apple optimises LLMs for Edge, Nvidia's Chinese competitor unveils AI GPUs.
Hello Engineering Leaders and AI Enthusiasts!
Welcome to the 173rd edition of The AI Edge newsletter. This edition brings you Alibaba’s DreaMoving that produces HQ customized human videos.
And a huge shoutout to our incredible readers. We appreciate you😊
In today’s edition:
🎥 Alibaba’s DreaMoving produces HQ customized human videos
💻Apple optimises LLMs for Edge use cases
🚀 Nvidia's biggest Chinese competitor unveils cutting-edge AI GPUs
🧠 Knowledge Nugget: Explaining ChatGPT to Anyone in <20 Minutes by
Let’s go!
Alibaba’s DreaMoving produces HQ customized human videos
Alibaba’s Animate Anyone saga continues, now with the release of DreaMoving by its research. DreaMoving is a diffusion-based, controllable video generation framework to produce high-quality customized human videos.
It can generate high-quality and high-fidelity videos given guidance sequence and simple content description, e.g., text and reference image, as input. Specifically, DreaMoving demonstrates proficiency in identity control through a face reference image, precise motion manipulation via a pose sequence, and comprehensive video appearance control prompted by a specified text prompt. It also exhibits robust generalization capabilities on unseen domains.
Why does this matter?
DreaMoving sets a new standard in the field after Animate Anyone, facilitating the creation of realistic human videos/animations. With video content ruling social and digital landscapes, such frameworks will play a pivotal role in shaping the future of content creation and consumption. Instagram and Titok reels can explode with this since anyone can create short-form videos, potentially threatening influencers.
Apple optimises LLMs for Edge use cases
Apple has published a paper, ‘LLM in a flash: Efficient Large Language Model Inference with Limited Memory’, outlining a method for running LLMs on devices that surpass the available DRAM capacity. This involves storing the model parameters on flash memory and bringing thn-feature-via-suno-integration/em on demand to DRAM.
The methods here collectively enable running models up to twice the size of the available DRAM, with a 4-5x and 20-25x increase in inference speed compared to naive loading approaches in CPU and GPU, respectively.
Why does this matter?
This research is significant as it paves the way for effective inference of LLMs on devices with limited memory. And also because Apple plans to integrate GenAI capabilities into iOS 18.
Apart from Apple, Samsung recently introduced Gauss, its own on-device LLM. Google announced its on-device LLM, Gemini Nano, which is set to be introduced in the upcoming Google Pixel 8 phones. It is evident that on-device LLMs are becoming a focal point of AI innovation.
Nvidia's biggest Chinese competitor unveils cutting-edge AI GPUs
Chinese GPU manufacturer Moore Threads announced the MTT S4000, its latest graphics card for AI and data center compute workloads. It’s brand-new flagship will feature in the KUAE Intelligent Computing Center, a data center containing clusters of 1,000 S4000 GPUs each.
Moore Threads is also partnering with many other Chinese companies, including Lenovo, to get its KUAE hardware and software ecosystem off the ground.
Why does this matter?
Moore Threads claims KUAE supports mainstream LLMs like GPT and frameworks like (Microsoft) DeepSpeed. Although Moore Threads isn’t positioned to compete with the likes of Nvidia, AMD, or Intel any time soon, this might not be a critical requirement for China. Given the U.S. chip restrictions, Moore Threads might save China from having to reinvent the wheel.
We need your help!
We are working on a Gen AI survey and would love your input.
It takes just 2 minutes.
The survey insights will help us both.
And hey, you might also win a $100 Amazon gift card!
Every response counts. Thanks in advance!
Knowledge Nugget: Explaining ChatGPT to Anyone in <20 Minutes
Recently, generative LLMs like ChatGPT have rapidly evolved, garnering attention from researchers and the public. It's crucial that developers of this tech effectively communicate its nuances to prevent public skepticism or overly restrictive regulations (like past experiences with nuclear energy research).
So, in this overview,
proposes a three-part framework for understanding and explaining generative LLMs. For a clear and simple manner, he identifies the key ideas and technologies that underlie them:Transformer architecture: the neural network architecture used by LLMs.
Language model pretraining: the (initial) training process used by all LLMs.
The alignment process: how we teach LLMs to behave to our liking.
Why does this matter?
This helps develop (and democratize) a working understanding of these key ideas and how they combine together to create a powerful LLM. Since AI is now moving from research to popular mainstream culture, effective communication about AI's evolution, challenges, and future is now more crucial than ever for all.
What Else Is Happening❗
📥ChatGPT now lets you archive chats.
Archive removes chats from your sidebar without deleting them. You can see your archived chats in Settings. The feature is currently available on the Web and iOS and is coming soon on Android. (Link)
📰Runway ML is Introducing TELESCOPE MAGAZINE.
An exploration of art, technology, and human creativity. It is designed and developed in-house and will be available for purchase in early January 2024. (Link)
💰Anthropic to raise $750 million in Menlo Ventures-led deal.
Anthropic is in talks to raise $750 million in a venture round led by Menlo Ventures that values the two-year-old AI startup at $15 billion (not including the investment), more than three times its valuation this spring. The round hasn't finalized. The final price could top $18 billion. (Link)
🤝LTIMindtree collaborates with Microsoft for AI-powered applications.
It will use Microsoft Azure OpenAI Service and Azure Cognitive Search to enable AI-led capabilities, including content summarisation, graph-led knowledge structuring, and an innovative copilot. (Link)
🌐EU to expand support for AI startups to tap its supercomputers for model training.
The plan is for “centers of excellence” to be set up to support the development of dedicated AI algorithms that can run on the EU’s supercomputers. An “AI support center” is also on the way to have “a special track” for SMEs and startups to get help to get the most out of the EU’s supercomputing resources. (Link)
That's all for now!
Subscribe to The AI Edge and join the impressive list of readers that includes professionals from Moody’s, Vonage, Voya, WEHI, Cox, INSEAD, and other reputable organizations.
Thanks for reading, and see you tomorrow. 😊
Nice summary! Thanks for sharing my work