The Path of an AI Creator News: Mistral AI announces "Magistral", a multilingual inference model! Useful for research and strategy. #MistralAI #inferencemodel #AItechnology
Video explanation
A new star in the AI world! "Magistral" has arrived! What exactly is this French-born "thinking AI"?
Hello and welcome to the world of AI technology! I'm John, your AI guide. Recently, new news has been coming out almost every day in the world of AI, but today I would like to focus on an AI model called "Magistral" I would like to explain Magistral in an easy-to-understand manner for beginners. Magistral was developed by a French AI startup company,Mistral AIThe name is kind of futuristic and cool!
Mistral AI is a rising star in the AI industry, announcing a series of innovative technologies. What kind of AI can Magistral, which they have finally announced, do? Let's take a closer look at its secrets!
What kind of AI is Magistral? ~We are experts in "thinking AI"~
Describe Magistral in one word:An expert in "thinking AI"That's about it. In AI terminology, it's called a "reasoning model." "Reasoning model" sounds a little difficult, but don't worry. In essence, it works like a human solving a puzzle, or a great detective deducing the culprit from clues in a complex case.Follow logical steps to arrive at the answerIt's an AI that's good at that.
Magistral is said to excel in three areas:
- Specialized areas of expertise: For example, in fields that require specialized knowledge, such as medicine, law, and finance, you can use your in-depth knowledge to make accurate judgments and analyses, just like an expert in that field.
- Transparent thinking: This is one of the major features of Magistral. The AI shows the thought process of "why did it reach this answer?", that is, the path it took to understand it. This is a very important point when it comes to trusting the judgment of AI.
- Multilingual support: He can "think" with high accuracy in a great many languages, including not only English, but also European languages such as French, Spanish, German, and Italian, as well as Arabic, Russian, and Simplified Chinese. He's just like a simultaneous interpreter at an international conference!
Some previous AI models could give a vague answer, but when you asked "Why did you reach that conclusion?" it was difficult to give a clear explanation. However, Magistral aims to be like a reliable consultant who can clearly answer that "why?" question.
How can Magistral help us in our work and lives? Let's take a look at some specific use cases.
So, in what specific ways will this "thinking AI" Magistral be useful to us? According to Mistral AI, it is expected to be used in a wide range of fields, including:
- Assistance with corporate strategy meetings: We provide support in analyzing risks from multiple angles when starting a new project, and in formulating optimal business strategies based on complex market data. It's like having an experienced management consultant by your side!
- Greatly improve business efficiency: For example, it helps logistics companies solve difficult problems such as, "How can we deliver packages in the most efficient and on-time manner?" It calculates the best delivery route and time by taking into account many conditions, such as weather, traffic conditions, and delivery destination restrictions.
- Make smarter decisions based on data: We are surrounded by a wealth of information, but it helps us to pinpoint the truly important points and make better decisions.
- A powerful supporter of programming and development work: It is expected to play a major role in software development as well! It will provide a wide range of support, from planning the entire project, to designing the back-end of the system (called back-end architecture), designing the screen that users see (front-end design), and how to handle important data (data engineering). It looks like it will be more deeply involved in the development process than previous AIs.
What I want to pay particular attention to is the point I mentioned earlier."Transparency of thinking"Instead of AI simply telling us the conclusion, "This is the answer!", it carefully explains the thought process, saying, "For these reasons, we thought through these steps in order, and came to this conclusion." This way, we can accept the AI's proposal with confidence, and even if there is something about the AI's decision that makes us think, "Hmm?", it will be easier to find the cause of the misunderstanding.
Furthermore, Magistral supports many languages, so it can overcome language barriers and demonstrate its capabilities in international business situations and in research that involves documents written in various languages.
What types of Magistral are there? - Two types to choose from depending on your purpose -
The newly announced Magistral comes in two versions so that users can choose according to their purpose and situation. Let's take a look at the features of each version.
- Magistral Small: This isOpen Source EditionIt will be provided as a service. "Open source" means that something like an AI blueprint is published on the Internet, and anyone can freely download it, use it, and improve it for their own purposes. Magistral Small has a scale of 24B parameters (the number of parameters indicates the complexity and intelligence of the AI brain, and generally the higher the number, the higher the performance), making it perfect for AI researchers and engineers who want to develop new services to freely try out ideas and explore the possibilities of AI. This model can be obtained from "Hugging Face," a popular platform for researchers and developers that has many AI models and datasets from around the world.
- Magistral Medium: This is a more powerful and high performanceEnterprise versionIt is designed to handle more complex and large-scale challenges that are faced in the business world. It is currently available as a preview version through Mistral AI's AI chat service "Le Chat" and API "La Plateforme" (application programming interface, a kind of connection port for different software to exchange information). It will also be available on Amazon's cloud-based AI development platform "Amazon SageMaker".
In other words, there are two options: "Small" for individuals and small teams who want to deeply study AI technology or develop new applications, and "Medium" for companies that want to fully utilize the power of AI in business situations. This will allow more people to benefit from Magistral.
Mistral AI's challenge continues ~ Beyond "smart AI", research also made public! ~
Mistral AI has not only released a fantastic AI model called Magistral, but has also published a detailed report summarizing how they evaluated this AI model, how they trained it (the learning process to make the AI smarter!), and their unique ideas and algorithms (calculation procedures and methods) for more effectively training the "thinking AI" = inference model.Research paperIt is also available to the public.
This is not just an announcement of the results of the project, saying, "We've created an amazing AI!", but also a reflection of the company's willingness to share with the world the technical background and the company's thinking, saying, "We're making AI smarter by applying scientific approaches and applying our ingenuity in this way." This kind of open information sharing is very important for the healthy and rapid development of AI technology.
We can't wait to see how "thinking AI" like Magistral will evolve and what new possibilities and changes it will bring to our society, work, and daily lives!
The evolution of AI is truly remarkable! I personally find the emergence of an AI like Magistral, which can explain "why it thinks a certain way" in a way that humans can understand, very exciting news. We may be one step closer to a future where AI is not just a tool, but is integrated into our lives as a closer, more trustworthy partner.
This article is based on the following original articles and is summarized from the author's perspective:
Mistral AI unveils Magistral reasoning model