開発スタッフがアプリ関連の最新技術やニュースをご紹介します。

ONNX is a machine learning framework which acts like a medium to convert between different machine learning frameworks.

ONNX is designed to enable framework interoperability. There are many great machine learning libraries in multiple languages — PyTorch, TensorFlow, MXNet, and Caffe are just a few of the most popular ones in recent years, but there are many others.
The idea is that you can use one stack of tools to train your model and use another for inference and prediction to expand your model.

For example, once you’ve trained your model, you need to deploy it to a new iOS app so that anyone with a previously trained model can see the safety of their food. I first trained my model with PyTorch, but iOS expects to use CoreML for use within the app. ONNX is an intermediate representation of the model that allows you to easily move from one environment to another.

With tools such as ONNX-CoreML, you can now easily convert pre-trained models to files, import them into XCode, and integrate them seamlessly with your application.

This is just an overview of what ONNX is about. I will talk more detail in future.

Yuuma

Tweet

Previously , I talked about general knowledge of AI and it’s types , ANI, AGI, ASI which are in Type I group. Today I will talk about Type II AI group. With further research on AI, this form can be further divided into two types, called strong AI and weak AI.

The main reason was to distinguish between the production levels of different types of AI machines.

Strong AI

• This is a wide range of widespread applications.
• This form of application has incredible intelligence at the human level.
• It uses combined action of gathering data to process the information.
• Example: Advanced robotics

Weak AI

• This is a small application with a finite scope.
• This type of AI is useful for some limited tasks.
• Process information using supervised and unsupervised surveys
• Example: Siri, Alexa.

Why AI

The reason for artificial intelligence is to support human capabilities and data resolution with widespread consequences. That’s the right answer from a professional point of view. From a philosophical point of view, artificial intelligence can help people lead a more profitable life without working hard.

Of course, there are consequences with the development of AI, but we are still far from such consequences. If you are more interested in artificial intelligence, I suggest that you take an artificial intelligence course to understand the delicacy and skills of today’s artificial consciousness.

Tweet

Artificial intelligence can be illustrated as a machine. This technology has spread to become very famous in the modern world. It is the intelligent process of machines that is taught to learn, understand, and copy human behaviors. These machines can learn new things and act tasks like human without further instruction.

Technologies such as AI and ML are growing at a very high rate and have a great impact on our lives. In fact, everyone wants to connect to this technology anyway. We can say AI is a set of patterns and algorithms that can generate all the solutions without being explicitly instructed to do the job.

AI Types

AI cannot be categorized into a single category. Different types of AI are created to perform different tasks, and this is the difference. AI can be divided into two types based on how it works. I will talk about the type 1 first today.

AI Type I

1. Artificial Narrow Intelligence (ANI)
2. Artificial General Intelligence (AGI)
3. Artificial Super Intelligence (ASI)

ANI

It is one of the most common forms of AI accessible on the market today. These artificial intelligence systems are programmed to resolve a single complication. In addition, the tasks you perform are performed efficiently enough. As the narrow name suggests, it has some limited features such as an mail spam filter system and production recommendations. This is the most demanding and popular form of AI available in the world today.

AGI

A form of AI with human-level cognitive tasks in various domains such as in-game bots like chess, computing functions, and systems. We all have to wait a long time to build the AGI ecosystem. This system can be created using a combination of thousands of ANI systems that communicate with each other to copy human inference and operate sequentially. AGI can complete all human tasks more efficiently in less time.

ASI

ASI may be an inference developed by AGI. Once the ASI is created, it will be possible to exceed the capabilities of all human beings. This includes decision making and more. It’s also like making better logical thinkings and better emotional connections. With the development of artificial general intelligence, AI systems will be able to quickly upgrade skills that no one has thought of. Also, the gap between AGI and ASI is relatively narrow.

That’s all for this week. Interesting , right ? I will talk about type II in coming week.

Yuuma

Tweet

“Chimera Painter“を紹介しているGoogle AIのブログが
今月17日に公開されていました。

Tweet

Pixel 3が発売となりましたね

これに合わせてGoogleレンズアプリが日本語対応となりました。

続きを読む

Tweet