- 4 AI Things
- Posts
- Elon Musk's xAI Eyes $18 Billion Valuation
Happy Spring, everyone! AI is nothing but matrix multiplication. You multiply a bunch of numbers, add the products and pass the result through an activation function. A decision-maker that helps the system learn complex patterns. That's three basic operations for a single neuron in the neural network. The process of matrix multiplication allows neural networks to learn patterns and make predictions based on input data. This process of matrix multiplication is computationally intensive. Company which has more GPUs will be able to solve complex matrix multiplication problems and hence advance rapidly.
Here are the 4 interesting AI things that I learned and enjoyed this week.
4 AI Things
AI Research
Apple's developing a smart tool called ReALM for Siri to understand and respond better in conversations, even considering what's on your screen. This new tech is already doing better than OpenAI's GPT-4, despite being smaller in size. It's not confirmed when Apple will release it, but it might come with the next iPhone update or new phones. ReALM stands for Reference Resolution as Language Modeling. It's aimed at enhancing Siri's ability to understand context in conversations and process onscreen content.
Elon Musk's AI company, xAI, is looking at an $18 billion valuation with investors possibly putting in $3 billion. The company had aimed to raise $1 billion, getting $135 million so far. Despite Musk denying fundraising reports earlier, the focus remains on boosting xAI's value. This move reflects the broader AI sector's trend of significant investments, like Microsoft's over $10 billion in OpenAI. The industry's push for funding highlights a rush for development despite concerns about the feasibility and reliability of AI technologies.
In this 3 part series, we saw about Self-RAG and CRAG. Today, you guessed it we gonna talk about technique that combines both in LLM. RAG-Fusion starts by using a smart program to come up with a bunch of related questions to better understand what you're asking. Then, it searches the web (vector search) to find information that answers all these questions, pulling together a wide variety of info. Next, it sorts this info, picking out the most relevant bits to use. All this selected info is then mixed together into one big, useful pile of knowledge. In the final step, an algorithm (Reciprocal Rank Fusion) looks at this pile along with all the questions to come up with a single, and detailed answer. Basically, RAG-Fusion takes a deep dive into your question to give you back the best answer possible.
Exclusive Content
With Synchron speeding past Neuralink in the brain chip race, industry insiders are half-expecting Elon to pivot, announcing a new Tesla model that runs on pure ego and can autopilot directly to Mars.
Generative AI’s second wave will be video. Text to video tools are already getting lot of traction. Sora’s recent teasers and music video gave goosebumps to Hollywood technicians. Runway ML, a hard contender, has setup an annual AI Film Festival that showcases experimental movies created via AI tools. Studios like Paramount and Disney are considering the use of Gen AI tools for their production pipeline.
If you liked today’s edition
⏭️ Stay curious, keep questioning.