Google takes a quantum leap
More topics: xAI integrates a new image generator in Grok | How does Meta’s Llama 3.3 model perform?
Hi AI Enthusiasts,
Welcome to this week’s Magic AI News, where we present you the most exciting AI news of the week. Today, we are talking about Google’s advancements in quantum computing and the new image generator of xAI’s Grok.
In addition, we are looking at the new state-of-the-art 70B model from Meta, which offers performance similar to the Llama 3.1 405B model. Stay curious! 😎
This week’s Magic AI tool is a 100 % local and free alternative to paid chatbots like ChatGPT or Perplexity. Ideal for environments that require strong data protection!
Let’s explore this week’s AI news together. 👇🏽
Top AI news of the week
🤔 Google takes a quantum leap
Google (Quantum AI) has announced a new quantum chip called Willow. The chip has state-of-the-art performance and enables two major achievements. Let’s take a closer look.
The details:
- The chip reduces errors exponentially if it uses more qubits. This solves a major challenge in quantum error correction, which the field has worked on for almost 30 years.
- Willow completed a standard benchmark computation in less than five minutes, a task that would take one of today’s fastest supercomputers 10^25 years. This number exceeds the age of the Universe.
- The published results show that using more qubits in Willow reduces errors and makes the system more quantum-like. The researchers speak of an exponential reduction in the error rate. This is a historic achievement among experts.
The Quantum AI researcher has a long-term roadmap:
Our thoughts
This is a big milestone because Quantum AI researchers have shown for the first time that it is possible to reduce errors by increasing the number of qubits. Quantum error correction is necessary to run useful applications. A qubit is a basic unit of quantum information that can exist in more complex states beyond just 0 and 1.
In the future, quantum computers could lead to technological progress that we cannot yet imagine - a fascinating topic.
More information: 🔗 Google Blog | Quantum AI
🌉 xAI integrates a new image generator in Grok
The company xAI has integrated a new image generation model called Aurora into Grok. Previously, Grok used the Flux.1 model from Germany.
The details:
- It is an autoregressive mixture-of-experts network. xAI trained the model using billions of examples from the internet.
- According to xAI, the new model can generate photorealistic images and follow text instructions precisely.
- The model supports multimodal inputs so users can edit provided images.
- The Grok chatbot is available for free in certain regions (with limits). You need a free X account to use it.
Below you can see an example of the new image generator:
Our thoughts
We tested the new image generation model from Grok and are impressed. We were especially impressed by how accurately the model follows instructions. Like the previous Flux.1 model, the new model has no restrictions. For example, it is possible to generate photorealistic images of famous people.
It’s difficult to determine if the new model is better than the previous Flux.1 model, but we think it is similarly powerful.
More information: 🔗 xAI Blog
🤖 How does Meta’s Llama 3.3 model perform?
Meta has released Llama 3.3 70B, which offers similar performance to the Llama 3.1 405B model.
The details:
- It is a pre-trained and instruction-tuned multilingual large language model (text in/text out). The model supports the following languages: English, German, French, Italian, Portuguese, Hindi, Spanish, and Thai.
- Llama 3.3 outperforms many of the other open-source and closed-source chat models on common benchmarks.
- It is available on HuggingFace and Ollama.
Our thoughts
Meta is the leader in open-source AI. It’s impressive how powerful the Meta models are. The benchmarks show that the new 70B Llama has a similar performance to OpenAI’s GPT-4o. 🤯
Imagine you always have your own chatbot with you. Whether you have internet access or not, the model runs 100% locally. You can try today’s Magic AI Tool to use the latest LLMs on your local machine.
More information: 🔗 Meta Llama | Ollama Llama3.3
Magic AI tool of the week
Are you looking for a user-friendly tool that can run state-of-the-art LLMs on your local machine? Just like ChatGPT, but 100 % local and private!
Then, the tool AnythingLLM might be just what you need! It is a desktop application with a suite of tools. You can run state-of-art LLMs locally without an internet connection. It supports custom AI agents and multi-modal models.
Just upload PDF or TXT files and chat with your documents. It’s available for Mac (Apple Silicon and Intel), Windows and Linux. There is also a Docker version with multi-user support.
Articles of the week
- Our Journey to Become a Data Scientist
- Using Miniconda x86_64 & Apple Silicon Side-by-Side on Your Mac
- Understand and Implement an Artificial Neural Network from Scratch
- Struggling with Linear Regression? - Theory and Practice Clearly explained
💡 Do you enjoy our content and want to read super-detailed articles about AI? If so, subscribe to our blog and get our popular data science cheat sheets for FREE.
Thanks for reading, and see you next time.
- Tinz Twins
Leave a comment