Yann LeCun - GenAI has a shelf life of three years
More topics: Mistral releases a speedy AI assistant | InvestAI: Europe plans to invest €200 billion € in AI
Hi AI Enthusiasts,
Welcome to this week’s Magic AI News, where we present you the most exciting AI and tech news of the week. Today, we’re talking about the future of generative AI.
In addition, we’re looking at Mistral’s speedy new AI assistant and Europa’s massive investment in AI infrastructure. Stay curious! 😎
This week’s Magic AI tool reveals what AI knows about you when you post a personal photo online. It’s very impressive but also a bit creepy.
Let’s explore this week’s AI news together. 👇🏽
Top AI news of the week
🤔 Yann LeCun: GenAI has a shelf life of three years
Meta’s head of AI, Yann LeCun doesn’t believe in the future of generative AI. He also doesn’t like the term Artificial General Intelligence (AGI). Instead, he prefers the term Advanced Machine Intelligence (AMI).
The details:
- According to LeCun, governments should support open-source AI. Meta offers its Llama models as open-source. However, they don’t publish the training data.
- LeCun thinks AI models need an understanding of the physical world. So we can develop more intelligent AI models.
- He believes that scaling generative AI models is the wrong approach, unlike Sam Altman from OpenAI. In addition, LeCun also thinks that the future of AI will not be generative AI.
- In an interview with John Werner on Forbes, LeCun said that generative models will be replaced by another model called Joint-Embedding Predictive Architecture (JEPA). He also added that GenAI has a shelf life of three years.
Our thoughts
We agree with LeCun on this topic. It will probably not be possible to achieve AGI by scaling generative AI models. There are several reasons for this.
First, everyone has a different definition of AGI, making it impossible to know when we’ve achieved it. Second, the models rely on predictions, which is not the same as truly understanding the world.
In our opinion, GenAI models are excellent for specific tasks. We also believe that the world doesn’t need AGI at all. We need specific AI models that can perform specific tasks.
More information: 🔗 Forbes | Heise Online | JEPA arXiv
💬 Mistral releases a speedy AI assistant
The French AI startup Mistral AI has released a new version of its chatbot, Le Chat. In addition, Mistral unveiled a new app for iOS and Android.
The details:
- The new mobile app has features like web search, code interpreter, document processing, and image generation powered by Black Forest Labs Flux Ultra.
- Mistral has also launched Flash Answers, which processes responses over 10 times faster than competitors like Claude or ChatGPT. Take a look at the video below.
- In addition, there are new subscription plans: a free plan, Pro, and Team. There is also an enterprise plan for custom deployments.
Our thoughts
Mistral AI is the only European AI company that can keep up with the top AI players in the world. As Europeans, we are happy about that. We often use Mistral’s open-source models in the development of AI-powered apps. The advantage of Mistral’s models is their speed.
More information: 🔗 Mistral AI
🤖 InvestAI: Europe plans to invest €200 billion in AI
At the AI Action Summit in Paris, the EU announced a €200 billion investment for building AI infrastructure as part of the “InvestAI” project.
The details:
- The InvestAI project aims to be the largest public-private partnership for trusted AI worldwide, with €150 billion contributed by industry and investors.
- The approach from Europe focuses on openness and cooperation. The project follows the CERN model.
- Commission President Ursula von der Leyen said: “AI will improve our healthcare, spur our research and innovation and boost our competitiveness. We want AI to be a force for good and for growth. (…)”
Our thoughts
This initiative aims to turn Europe into an AI continent. The European approach stands for openness and cooperation. This approach contrasts with closed-source models like those of OpenAI. As LeCun has pointed out, future models should be open-source. We believe this is the only way to build trustworthy AI models.
More information: 🔗 EU Commission | Heise Online
Magic AI tool of the week
Do you post photos of yourself online? Then, you can upload one of the photos to the “they see your photos” website. You can also use the sample images. The website analyzes the image using the Google Vision model.
The model can make several conclusions from the image. For example, age, political orientation, and mood. There are even recommendations for companies on how to advertise to you most effectively.
👉🏽 Try it out and see what the AI knows about you from just one photo!
Articles of the week
- Our Journey to Become a Data Scientist
- Build a Local RAG App to Chat with Earnings Reports
- Understand and Implement an Artificial Neural Network from Scratch
- Build a financial multi-agent system with AG2 (formerly AutoGen) and Ollama
💡 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
