Ph.D. AI Chips Developer@Intel Corp. Exploring Tomorrow’s Intelligence, Today!
United States
Joined January 2024
- Tweets403
- Following68
- Followers36
- Likes887
Microsoft integrates OpenAI’s Sora in Bing app to create AI videos for free. Check it out!
blogs.bing.com/search/June-2…
Anas retweeted
HUGE. A Chinese group has released one of the world's most powerful AI for robots as Open-Source.
Model Available on @huggingface
BAAI released RoboBrain 2.0 as an open-source embodied AI brain for humanoids and general-purpose robots. It advances spatial reasoning, multi-agent task planning, and closed-loop execution over its predecessor.
The model ingests multi-image, long video, and high-resolution inputs. A vision encoder and MLP projector merge visuals with task instructions into a unified token stream for the LLM decoder.
RoboBrain 2.0-7B is available on Hugging Face with a SimpleInference pipeline for general and embodied tasks. The pipeline supports up to 32768-token context for long-chain reasoning and structured plan output.
A 32B-parameter version is coming soon. On BLINK-Spatial, CV-Bench, EmbSpatial, and RefSpatial benchmarks it tops both open and closed-source rivals, achieving over 50% absolute gain on RefSpatial.
BAAI uses the FlagScale framework for distributed training and FlagEvalMM for multimodal evaluation. Both toolkits are open on GitHub and drive scalable development and thorough benchmarking.
RoboBrain 2.0 brings embodied AI into real-world robotics with open-source transparency and cutting-edge performance.
Anas retweeted
Welcome the newest addition to the @Xiaomi MIMO family
Remarkable multimodal reasoning power packed into a 7B model!
My take: The 7B scale is strategically important. Expect native execution on smartphones soon, shifting your device's role from information repository to proactive loyal intelligent agent!
PS: the MiMo-VL-7B series maintains full compatibility with the Qwen2_5_VLForConditionalGeneration architecture for seamless deployment and inference. Best part? You can deploy it right now!
Who will be the first @huggingface Inference Endpoint provider that supports this model?
Anas retweeted
🇺🇸 ANTHROPIC JUST OPEN-SOURCED A TOOL TO READ THE MINDS OF AI
Yes, for real.
They dropped a circuit tracer that lets you peek inside LLMs and see how the machine actually reasons through a problem.
Like watching neurons gossip in real time.
Using it, you can track how models like Llama-3 or Gemma-2b "decide" that the capital of Texas is Austin.
This could be a big deal for AI safety.
Or at least a handy way to catch your chatbot hallucinating before it claims Napoleon invented the iPhone.
Source: InfoQ
Model context protocol: The "USB-C" for Al Apps!
Ever wish connecting Al tools was as easy as plugging in a USB-C cable? That's Model Context Protocol (MCP)!
The Problem:
Without MCP, linking Al apps (like ChatGPT or Claude) to tools (weather APls, docs, etc.) requires custom integrations for every single combo. It's messy, slow, and expensive!
The MCP Solution:
MCP acts like a universal adapter:
1 standard protocol = fewer custom connections.
Hosts (Al apps) connect once → Servers (tools/data) connect once.
Solves the "MxN problem" → Saves time + money!
Why it matters:
Simpler experiences for users
Faster innovation for developers
Plug-and-play tools for everyone
What can MCP connect?
Tools (do tasks)
Resources (share data)
Prompts (reusable guides)
Sampling (self-improving Al)
MCP = Less chaos, more Al magic!
If you’ve ever used ChatGPT, watched a video on your phone, or used facial recognition—you’re benefiting from AI chips made with an invisible helper: photoresist.
Photoresist is like sunscreen for silicon.
When we shine special light through a stencil (called a mask), photoresist reacts and helps us “print” tiny patterns onto a wafer. Those patterns later become transistors—the building blocks of computer chips.
Without photoresist, we couldn’t build the super-small, super-fast chips powering AI.
And guess what? We’re not stopping there.
Innovators like Inpria are developing next-gen photoresists made from tin-based materials for next generation chips, helping us pack even more power into even smaller chips.
The future of AI depends on the chemistry behind the curtain.
What’s Stopping Sustainable AI Chips from Scaling?
As demand for AI accelerates, so does the need for powerful, efficient chips — but scaling them sustainably is still a major challenge.
The semiconductor industry is wrestling with:
•Material constraints (e.g., rare metals, ultra-pure gases, clean water)
•Geopolitical bottlenecks in chip supply chains
•Energy-intensive fab processes
McKinsey reports that while capacity investments are booming, “90% of new fabs are still concentrated in just 3 regions.”
That means more risk, more emissions, and less resilience.
To build sustainable AI infrastructure, we need to:
1.Rethink materials sourcing
2.Build localized clean fabs
3.Apply AI to optimize energy and yield
The future of AI depends on solving these material and manufacturing problems — not just model size.
#AI #Semiconductors #Sustainability #MaterialsScience #CleanTech #ChipManufacturing
AI chips demand atomic-level precision. That’s why materials like argon, nitrogen, and silane aren’t just support players — they’re core enablers.
From deposition to etching, these gases help shape the microscopic layers that give GPUs and AI accelerators their power.
A single impurity in these gases?
It can ruin an entire wafer batch.
As AI accelerates, so does the demand for clean, controlled chemistries in fabs. The next time you hear about an AI breakthrough — remember the invisible materials that made it possible.
What’s the most overlooked material in AI hardware?
Water.
Not in the data center.
In the fab.
Every wafer that ends up in an AI chip passes through hundreds of cleaning, etching, and developing steps—all using ultrapure water (UPW).
And a single fab can use up to 9 million gallons per day.
Yes, per day.
Here’s why it matters:
•UPW quality directly affects yield.
•Contaminants = defects = dead transistors.
•As AI chip demand grows, so does the need for ultrapure water systems and materials that can handle them.
Everyone’s chasing faster AI chips.
But most people don’t realize:
•Training a model uses millions of gallons of ultrapure water.
•Photopolymers can make or break yield at the most advanced chip technology nodes.
•The helium shortage? It’s not just for party balloons—it slows chip production too.
Anas retweeted
Grok 3 seems very strong. Great proof that the scaling laws are not, in fact, over. Very bullish for the future of AI.
Anas retweeted
40+ Must Know AI Tools;
1. Ideas
- YOU
- Claude
- ChatGPT
- Perplexity
- Bing Chat
2. Presentation
- Prezi
- Pitch
- PopAi
- Slides AI
- Slidebean
3. Website
- Dora
- Wegic
- 10Web
- Framer
- Durable
4. Writing
- Rytr
- Jasper
- Copy AI
- Textblaze
- Sudowrite
- Writesoni
Nike’s latest ad blends AI and traditional visual effects (VFX), marking a breakthrough in creative workflows!
🔹 5,000+ images generated via MidJourney
🔹 A hybrid approach showcases AI’s role in pro video production
🔹 Ad agencies are now building AI-focused teams!
Shopping hack I swear by: I've been using AI chat bot as my pocket nutritionist! 🛒
Just chatting with it about ingredients while shopping has helped me spot hidden sugars and processed stuff I used to miss. Saved me from so many "healthy" products that weren't actually healthy 😅
No more getting fooled by marketing! Here's how it works ⬇️
