Three Major Things in the AI Artificial Intelligence Industry

2023-10-09 13:20
优赛诺

On the occasion of National Day in 2023, the iAITalk official account was grandly launched, focusing on various topics in the field of AI artificial intelligence, especially the construction of AI applications such as AI intelligent question and answer applications, ChatGPT, and ChatGPT-like ones.

The most important thing is: iAITalk will observe and share AI from the perspective of ordinary users , uncover the mystery of AI applications, and leave the technological complexity of AI at the bottom to professional programmers.

This article is the opening article, focusing on sharing the following views of i AITalk:

1) AI artificial intelligence is not a new thing

AI artificial intelligence has actually appeared as early as around the 1960s, so AI is not a new thing. If anyone says that AI is the latest scientific and technological achievement, they are lying and do not understand it.

For example, the once famous IBM Watson robot is actually a typical representative of AI technology. It’s just that IBM Wats on cannot find application scenarios suitable for ordinary users (the general public), and there are no other AI products or technologies that can be used in daily applications after IBM Wats on. As a result, artificial intelligence seems to have become cold, profound, and advanced. synonymous with!

This is not the case! AI technology has actually been promoted and used, simple ones, such as customer service phone questions and answers that sometimes feel annoying, online customer service questions and answers after the Internet became popular, intelligent recommendations in the online shopping era, personal behavior recognition in the big data era, etc. , are all typical applications of AI technology. However, there is still a big problem, that is, ordinary users cannot directly use and enjoy the convenience brought by AI technology!

2) OpenAI’s ChatGPT opens the shackles of AI applications

It was not until the end of December 2022, more than nine months ago, that OpenAI, an AI artificial intelligence company from the United States, opened a new chapter in AI artificial intelligence by launching an AI application that truly benefits ordinary users - ChatGPT, and built a The super-large industry surrounding the AI artificial intelligence large language model LLM represented by ChatGPT : AI artificial intelligence industry!

Why is it said that ChatGPT opened the shackles of AI applications? How is it done?

First, ChatGPT uses the popular Web App, which is an application that can be accessed as long as there is an Internet connection (in the form of a web interface Webpage or website, as well as an App built on the basis of LLM that is suitable for smart terminal access) , opening up the channel for everyone to personally experience and use AI artificial intelligence! It has truly brought AI artificial intelligence into thousands of households and into the hands of ordinary people. It is no exaggeration to say that anyone who can speak has the opportunity to use AI artificial intelligence technology! This is what iAITalk believes is the most important contribution of ChatGPT!

Second, ChatGPT provides a better user experience. If providing everyone with an open channel to use AI artificial intelligence technology is the key to OpenAI opening the door to a new world of AI artificial intelligence , then the user experience provided by ChatGPT so far is still generally at the forefront, which is why OpenAI captures the majority of users. Fangxin’s trump card! I believe that any user who has used ChatGPT can't help but admire that ChatGPT understands them very well. Conversations with ChatGPT are comparable to conversations with real humans.

Third, after ChatGPT, OpenAI further opened up and provided APIs that can access its LLM large language model. Based on ChatGPT, it further expanded the ecosystem of the AI artificial intelligence industry and consolidated its leadership position . Most of the better-quality AI artificial intelligence applications that can be seen now are basically built based on OpenAI's LLM API (although the open source LLM community - the most typical representative: Huggingface - is also gradually forming and growing, but due to the artificial intelligence of AI The threshold in the field of intelligence is relatively high. Generally speaking, the resources currently provided by the Open Source AI Community are inferior to OpenAI by an order of magnitude. Only a few come from major manufacturers or projects with sufficient resource support, the latest such as Meta AI's Llama2 and TII's Falcon series can compete with OpenAI in some aspects).

3) Domestic AI artificial intelligence urgently needs to follow up quickly, especially to expand ordinary user scenarios

The domestic AI artificial intelligence industry has actually invested a large amount of public resources (funds and resource investment from the government, such as the AI artificial intelligence fund and AI artificial intelligence research institutions established under the leadership of the government) and private resources (from the private sector). capital and resource investment (such as venture capital, corporate investment in R&D, etc. ), there have been some achievements worthy of recognition (such as Wisdom's BGE), but overall, the gap with foreign countries (especially the United States) is still large. If LLM is regarded as the infrastructure in the field of AI artificial intelligence, then domestic investment in this field can be said to be huge with little effect ( Note: iAITalk focuses on the perspective of ordinary user application scenarios).

To expand a bit, applying AI artificial intelligence to the field of ordinary users basically requires three major items :

3.1 LLM large language model

According to the specific App application scenario, two major types of LLM are needed. One is the text generation LLM that allows ChatGPT to be applied on a large scale, and other generation LLMs (such as Midjourney's image generation LLM). These LLMs are also used by most ordinary users for AI applications . When directly contacted and used, iAITalk believes that this type of LLM can be called front-end LLM (Front-end LLM).

But in fact, there is another type of LLM, which is actually equally important (or even more important), and that is Embedding LLM (vectorized large language model). It is precisely because of the basic support provided by Embedding LLM that AI artificial intelligence Only then can we realize the function of "thinking like a human being".

It can be said that LLM is the core of AI artificial intelligence and the largest piece of AI artificial intelligence ! If you want to succeed in the field of AI and lead the AI industry, LLM must focus on development, develop practically, and develop effectively!

3.2 Software and hardware computing resources

If LLM provides the principles, mechanisms and tools for realizing AI artificial intelligence, then the software and hardware computing resources that allow LLM to actually run constitute the second largest piece of the AI artificial intelligence industry ! The operating efficiency and final output quality of LLM rely heavily on software and hardware computing resources, and in this field, the United States still occupies a dominant position. Especially in terms of GPU computing power, the current situation is basically dominated by NVIDIA, followed by AMD, and Intel follows suit. Domestic players have little influence. Since software and hardware computing resources have such an important impact on the AI industry, if you want to have a place or even lead in the field of AI, you must also pay enough attention and investment!

3.3 App builds front-end and back-end platforms

The third major component that allows AI Apps to reach ordinary users is the front-end and back-end platforms for AI Apps, because for users of AI Apps, they must interact with AI applications through some kind of interface. , Therefore, the platform used to develop and deploy AI App is the most critical at this stage.

As far as iAITalk knows, currently, it is suitable for ordinary users (note: ordinary users here refer to developers who develop AI Apps. Don’t be scared by the word “developer”. In fact, due to computer technology Especially with the development of software programming technology, software development is no longer the exclusive domain of so-called professional programmers. As long as you have certain computer programming knowledge, you can develop your own AI App. In the future, iAITalk will launch a series of articles for It is a user group of non-professional programmers, helping everyone to build their own AI applications) The front-end and back-end platforms or tools for developing AI Apps are basically from abroad (especially the United States), such as for front-end development Streamlit, Gradio, PyW ebIO, WebNiceGUI, etc., the programming languages Python and JavaScript used to actually write AI Apps, and Langchain, LlamaIndex, etc., which can be regarded as middleware, all come from abroad (the United States) without exception. What’s more, these platforms or tools are available to anyone! ! !

Looking back at China, although we have the largest group of programmers in the world, there seems to be no benchmark that can compete with the above platforms or tools. It has to be said to be a great regret, and it also inevitably reminds us that we must strengthen research in these fields. , investment and continuous progress, otherwise, if you want to truly master and own AI artificial intelligence , the foundation will be unstable!

Conclusion: iAITalk believes that, similar to the application of Internet technology, China’s ability to apply new technologies and improve new technologies (from the perspective of improving user experience) will also be the same or even better in the field of AI . Performance, but the three major items listed above are something we must not turn a blind eye to! Otherwise, it is just hiding one’s ears and stealing the bell!

But iAITalk also believes that, like in many other fields, China has never lacked successful cases of becoming a rising star! Although we are latecomers in the field of AI, as long as we face up to the problem and take practical actions, China will surely be able to gradually follow up and gain its own unique advantages in the core key areas of AI artificial intelligence (especially the above three major areas ), thus Truly establish and form an independent and autonomous AI artificial intelligence industry! Let the AI artificial intelligence industry serve the country’s all-round development and enhance people’s well-being!

So, let us work together, each from his or her own perspective, to jointly contribute to the development and progress of China’s AI field !

Come on, Chinese AI!

Finally, but most importantly, on this special day, I wish my motherland a happy Mother's Day!

Life Enchancing With AI!

aichat101@foxmail.com