Sora's Official Music Video

Happy short week everyone! Recently, I made a video about how AI will replace programmers and got a lot of appreciation from the community. Because I mentioned that it is all about perspective on how you look at it. If you look at it in a positive way it can accelerate the development process. So, when a new AI tool hits the market don’t treat it as a threat and rather try it out and see if you need it.

Here are the 4 interesting AI things that I learned and enjoyed this week.

4 AI Things

AI Research
Qdrant introduced FastLLM (FLLM), a breakthrough in AI for generating and retrieving content, designed especially for Retrieval Augmented Generation (RAG) cases. It's unique because it can handle a massive 1 billion tokens, setting it apart with its optimized architecture perfect for RAG applications. By pairing FastLLM with Qdrant, developers can process huge data amounts more efficiently. The model underwent training on 300,000 NVIDIA H100s and showcases unparalleled capabilities, like achieving 100% accuracy in identifying specific text within 1 billion tokens. Despite its immense power and a total of 1 trillion parameters, the real applications are still being explored, humorously suggesting its capacity might be best suited for yet another RAG chatbot. This step represents Qdrant's ambition to consolidate AI tools, providing a versatile resource for developers and researchers to push AI's boundaries further.

AI News
In our 7th edition, we introduced Sora. OpenAI dropped a music video created by Sora, showing what an artist imagined while composing a song. Only a few have tried Sora, which can turn text into videos. The artist, August Kamp, said it removes creative limits. The video, without much detail on its creation, showcases longer, consistent clips compared to other platforms. Set for public release this year, it's potential is huge but awaits safety checks to avoid misuse, especially with the upcoming global elections.

AI Concept

Last edition, we saw what a self-RAG is. Today we are going to talk about The Corrective RAG. CRAG method enhances language model precision by smartly integrating data from documents it retrieves, using a system to evaluate and decide the utility of these documents. It enriches its knowledge base through web searches, accessing a broader and more current spectrum of data. By selectively processing and refining the information from these sources, it focuses on the essentials, discarding irrelevant details. This approach allows for improved accuracy and relevance in generated content. Additionally, CRAG can seamlessly enhance the performance of existing RAG models, offering significant improvements without necessitating substantial modifications.

Exclusive Content

AI big thing

I spoke about quantum computing on being the next big thing. Microsoft and Quantinuum made a big leap in quantum computing, making these super-fast computers more reliable. They tackled the problem of "qubits," the building blocks of quantum computers, being very sensitive and prone to errors. By using an error-correction method, they turned 30 shaky qubits into four solid ones, hitting record reliability. They did a ton of tests without messing up once and plan to share this tech with their cloud computing users soon. They're aiming for 100 solid qubits to outdo regular supercomputers and think they've sped up the process by at least two years.

If you liked today’s edition

⏭️ Stay curious, keep questioning.