In this blog post, I will explain the difference between the existing visual AI model editors and Deep Block.
When I talk about Deep Block, some people say that there are already many services similar to Deep Block.
However, if they use Deep Block and other services, they will think differently.
First, let's talk about the existing AI model editors.
Many existing AI tools can design AI models in flowchart format.
But if you use an AI model editor with this flowchart format, you eventually think that it is more comfortable typing code rather than using this type of editor because your hand will get tired by moving and clicking the mouse.
And there is a bigger problem. These model editors actually require users' AI knowledge.
Many AI model editors claim that their service is no-code, easy, and they can democratize AI.
However, in my opinion, these editors are not really “no-code” and are not “easy”.
To use these model editors, you need to understand each layer of deep learning models and understand how to combine them.
And to know this, you need to study linear algebra, computer science, calculus, and deep learning.
Do people without knowledge of AI or computer science really understand LSTM, convolution layer, various loss functions, etc., and can they combine these to implement an AI model?
Most editors connect each layer with a flow chart format, but joining in layers like this is already supported by Pytorch and Keras.
What's different from CODING? And is this really EASY?
Even a software engineer, who knows this, finds it difficult to make a GOOD AI model by combining the layers of AI model.
Even if you have a good AI model editor, it's useless unless you have the knowledge to use it.
I am not disparaging the existing AI model editors. However, to use these AI model editors, users must have knowledge of AI, and people with AI knowledge usually write code instead of using such editors.
In the past, I also thought of an AI model editor with a flow chart or block-coding concept.
But, in the end, I realized that there was such a problem, and in order to create a more usable tool, I made Deep Block.
In Deep Block, users do not need to design an AI model themselves.
And this means users don't need background knowledge of AI and they don't really need to code.
We have already implemented an AI model for each purpose, and users just need to choose the purpose and put the training data into the model.
Take a look at our demo video. Even kids can use this.
We are focusing on the field of computer vision, and users can create their own AI detectors with their own images.
Some will point out that our platform has limitations in terms of optimization and customization.
I agree. So, we are working hard to provide better AI models, and we are constantly improving our platform to provide a more customizable interface.