The era of AI creation is coming, writing, speaking, drawing, and making videos are all possible——the workplace is “windy”, where do humans sit?

At the end of 2022, the most discussed topic on the Internet is a chat robot – ” ChatGPT “. Published by the American artificial intelligence research institution “OpenAI”, it has attracted millions of people to use it in just two weeks. Experts believe that the emergence of “ChatGPT”, like the emergence of smart phones and the Internet, will profoundly change human work and life.

The “Generative AI” technology of “ChatGPT” allows anyone to use simple natural dialogue (rather than programming language) to direct AI to create various content.

What is “generative AI”? How to become popular? What are the current applications? And how can humans “find another job” in the era of AI creation?

Innovation point: AI has been able to create text, pictures, audio and video, and design based on simple text instructions. What about the people who did these things in the first place?

The first picture of this article was completed by me using the AI ​​​​drawing tool “Stable Diffusion”, and the text was completed by me and “ChatGPT”. Can you guess which ones were written by AI? Which ones were written by real people? The answer is revealed at the end of the article.

The development history of “Generative AI”

Generative AI (Generative AI) refers to letting the “machine learning model” study the data of similar works, and then create a brand new work, which can be text, images, audio files, videos, codes, or even architectural designs.

In the past seven years, technology giants such as Google and Meta, as well as OpenAI invested by Microsoft, have been building the “Language Model” of Generative AI. These three pioneers used a lot of computing power and data to “train” these “language models” so that they could create their own content. The training process can be divided into three stages:

A. Emerging period (before 2015):

Only small models. These models perform well in “data analysis”, such as predicting delivery arrival time, classifying fraudulent messages, etc. However, it fails in the “content creation” that imitates human language, and cannot write, write programs, or draw pictures like real people.

B. Breakthrough period (2015-2022):

In 2017, Google Research published the ” Transformer Model “, which is a milestone in the field of Natural Language Processing (NLP). The Transformer model needs less time to be trained than before, but the quality of the output is greatly improved, and it is easy to customize for the needs of various application fields. (For details, please refer to the introduction of Wikipedia )

For example, Google developed BERT and LaMDA. OPT-175B, BlenderBot developed by Meta. OpenAI invested by Microsoft has developed GPT-3 (for text), DALL-E2 (for drawing), and Whisper (for speech recognition).

Why are they all big companies that are as rich as the enemy? Because training these models is expensive.

For example, GPT-3 was initially trained on 45TB of data, using up to 175 billion parameters to predict results, and a single training session cost $12 million. China’s Enlightenment pre-training model uses 1.75 trillion parameters for training, mobilizing resources from Tsinghua University, Peking University, Chinese Academy of Sciences and other institutions.

Between 2015 and 2020, the amount of computation required to train these models has increased by six orders of magnitude, enabling these models to perform tasks at a level close to, or even surpass, that of humans.

It’s just that at this stage, these models have not entered the public eye. Because they require huge resources to operate, and the cost has not been reduced enough to be used by the public in the cloud.

C. Industry landing period (2022~):

As the cost of computer computing decreased and new technologies such as the diffusion model (Diffusion Model) emerged, training and battalion operations gradually decreased. Companies such as Google have made these models public, allowing developers to try them out.

After the core generative model is trained, it can be customized and adjusted according to various fields without a large amount of data. Therefore, the BERT developed by Google has a model BioBERT focusing on the biomedical field and a model Legal-BERT in the legal field . This has led many professionals to try it out as well.

Then in 2022, generative AI becomes popular among the general public.

AI can do better than humans in various content such as text, video, and video

In 2022, the killer app of generative AI will be available to the general public. Such as Midjourney, Stable Diffusion, DALL·E2 in the field of drawing, and the chat robot ChatGPT.

At the same time, pioneers open their language models (such as OpenAI open GPT-3), allowing new companies to save money and time-consuming training phase, and directly launch application products in various professional fields.

The scope of these new creations is not only to generate text, pictures, sounds, programs, music, images, 3D, NFT through text, but also to reverse operations, using voice to generate text (such as verbatim script applications), and images to generate pictures ( Such as fast reciting, image editing), link generation text (such as quickly extracting url web content to generate scripts), video generation video (such as quickly generating blog articles, tweets, highlights… ….etc). Here are various types of applications of generative AI sorted out by netizens .

Taking the marketing industry with huge business opportunities as an example, the commercial application potential of generative AI is quite amazing.

From blog article writing, SEO optimization, video editing, social marketing copywriting, creative generation, etc., AI can do everything for you. The tracks are packed with contenders:

(only some are listed below)

From generating ideas, to actually writing content, taking pictures, and shooting videos, all kinds of marketing materials can now be generated by sitting in front of the computer with one click. There is no need to scratch your head, exhaust your gut, or run in and out to light up, NG do it over and over again.

Many companies in the United States have begun to adopt such AI tools. For example, the content marketers of VMWare, a major cloud computing company, use Japser to produce marketing email content, advertising copy, and community content. When AI can already perform most of the writing tasks, writers can focus on finding better writing topics, researching content directions, and formulating content strategy directions. This kind of phenomenon of playing “the wind blows and changes seats” with AI has already happened in large companies.

In addition, banks such as Morgan Stanley also use GPT-3 to create customized financial content for customers. A real person “prompts” the AI ​​with specific words, and then the AI ​​creates the first draft, which is then edited and reviewed by a real person.

The field of marketing is just the tip of the iceberg of generative AI applications. According to the analysis of Sequoia Capital, a well-known American venture capital company , generative AI can also be used to automatically create programming languages, artworks, games, product designs, etc.

AI is blooming, where will human beings go? After the wind blows, is there still room for us?

In the era of AI creation, how do humans “change positions”?

When generative AI output can create high-quality content in large quantities, quickly, and customized, what is the value of human beings?

Several trends are doomed to “cannot go back”:

1. AI will be the mainstay of writing most “functional” content: this type of content has a clear structure, high repetition, the purpose is to convey information, provide clear and necessary explanation

No need to convey personal style, brand, image. No sentimentality or interpretation required. For example, promotional information, discount code instructions, product brochures, community posts, educational training videos, advertising plans, research reports, white papers, business plans, brochures,

These contents can be drafted quickly through AI tools, and can be sent after fine-tuning. In other words, marketing, sales, operations, and customer service can all save a lot of manpower through AI. In other words, in this field, AI can most easily replace human values.

2. Treat AI as your hand and human as your brain

The value of content talents lies not in “writing”, but in capturing market trends, current events, and customer mentality, and formulating content strategies that best suit the current time and space background. Although the current AI can efficiently produce content, it is still unable to analyze current affairs issues and market trends of concern, and find out content strategies that can be cut into.

In addition, AI is still unable to empathize and simulate the psychology of customers, nor can it speculate on the articles that customers may want to read. AI is also unable to integrate the company’s internal database and dig out valuable topics to produce content. Entering the era of AI creation, the biggest win for human beings is to change their positions to be the brain and give up all the work to AI.

3. Only by understanding the emotions and needs of viewers can we kill the blood in the melee of new content

Generative AI accelerates content production and makes content more flooding. In Chaos Army, the content that can attract attention may be that the content is more closely related to current events, more catering to the psychological state of the viewers, more close to the various emotions pervasive in society, or more unique in style and characteristics.

Generative AI is still developing with a ferocious momentum-it will become popular in 2022, and it is expected to bloom more in 2023. The “arms supplier” of generative AI – OpenAI, became popular overnight because of ChatGPT. It is rumored that Microsoft will receive a new round of financing of up to 10 billion US dollars, making the company’s valuation as high as 29 billion US dollars. Microsoft is optimistic that countless third-party companies will be able to create new application services through OpenAI’s API and use AI to change business operations.

A new era led by AI creation is at hand. Perhaps our fear is that this is only the beginning.