top of page

AI 2.0: ChatGPT 4 Is Moving to Commercialization

Artificial Intelligence (AI) has come a long way since its inception, and the advancements in this field never cease to amaze us. Chatbots are one of the most significant results of AI technology, and they have changed the way businesses interact with their customers. One of the latest innovations in chatbot technology is AI 2.0, and ChatGPT is moving towards commercializing this game-changing technology.

What is AI 2.0?

If the past AI is regarded as the 1.0 stage, it can be concluded that its implementation requires five prerequisites, namely massive data; accurate and automatic labeling of data; a single clear field; large-scale computing power; and top AI scientists.

AI 2.0 refers to the next generation of artificial intelligence technologies that employ more sophisticated algorithms and larger datasets to generate greater accuracy and better results. This new generation of AI is more adaptable, creative, and even able to mimic human traits like intuition, empathy, and decision-making. AI 2.0 has the potential to help solve a range of challenges across many industries, from healthcare and finance to manufacturing and transportation.

ChatGPT's AI 2.0 technology is based on a transformer-based model that offers a higher level of accuracy and versatility than traditional chatbot technology. The AI 2.0 model does not require a large amount of training data, making it more flexible and adaptable than other models.

From ChatGPT-1 to ChatGPT-4

In 2015, the artificial intelligence research laboratory OpenAI was founded in San Francisco, California, USA. In 2018, OpenAI launched GPT-1 (Generative Pre-trained Transformer-1). GPT-1 is a generative pre-training Transform model: a deep learning model using a self-attention mechanism, and the entire GPT series has since followed, executed, and optimized this model. The GPT-1 method includes two stages of pre-training and fine-tuning. The pre-training follows the language model target, and the fine-tuning process follows the target logic of the text generation task.

In 2019, OpenAI launched GPT-2. GPT-2 has the same language model structure as GPT-1, but thanks to higher data quality and larger data scale, GPT-2 has amazing language generation capabilities. However, GPT-2's performance did not meet ideal expectations when performing more professional and sensitive tasks such as music interpretation, story sharing, and emotional analysis. In 2020, OpenAI launched GPT-3. GPT-3 can construct a dialogue between two artificial intelligences, can complete difficult tasks such as answer analysis, excerpt text, language translation, and code generation, and talk about how to become a human being. Its birth is regarded as a milestone in the history of artificial intelligence development event.

At the same time, GPT-3 elevates the GPT model to a new level of technical practice and application. Its training parameters are more than 10 times that of GPT-2. In the technical route, the fine-tuning steps of the first generation of GPT are removed, and natural language is directly input as instructions. After GPT training has read the corresponding sentence, it has the ability to continue to answer questions.

In 2022, OpenAI will evolve GPT-3 into ChatGPT. ChatGPT has demonstrated a series of amazing capabilities in various tests in multiple scenarios, industries, and fields, such as writing code, writing news, question correction, factual question and answer, entity extraction, and other NLP tasks. Practitioners in specialized positions such as programmers, journalists, editors, and scientific researchers have all felt new opportunities and challenges. The birth of ChatGPT opened another window for artificial intelligence technology to have a profound impact on human society.

Giant Embedding GPT-4

Recently, companies have disclosed their commercialization progress, from the upstream provision of training data and other infrastructure for the AI industry, to the developers and downstream applications of ChatGPT competing products, all are trying to "take a share" in different ways.

On March 16th, just two days after OpenAI released GPT-4, Microsoft announced that it would embed it into the "family bucket" of Microsoft Office software. At the same time, A-share listed companies on the Shanghai Stock Exchange are also competing to participate in the trend of ChatGPT products.

Microsoft announced at the press conference that Microsoft 365 Copilot with embedded GPT-4 will be unveiled soon. Microsoft CEO Satya Nadella said: "Today marks an important step in the evolution of how we interact with computing."

Microsoft released the Copilot at this AI-themed conference, supported by the GPT-4 platform. According to Microsoft, Copilot will be embedded in Word, Excel, PowerPoint, Outlook, Teams, and other office software. In Word, Copilot can summarize files and provide editing suggestions; in Excel, users can use natural language to call various complex functions; in PowerPoint, Copilot can enable the one-click conversion between PPT and Word.

Microsoft is also introducing a new experience — Business Chat. This service combines data from multiple Microsoft applications, including documents, slides, emails, calendars, notes, and contacts. The service can use large language model technology to help users summarize chat content, compose emails, search Key dates, and develop plans based on other project documents.

In addition to Microsoft, OpenAI also disclosed more application scenarios of GPT-4, and its customers involve finance, language, education, video, consulting, and other fields. From the perspective of specific applications, GPT-4 applications can be divided into two types: one is To B, which is provided to large companies to improve internal work efficiency; the other is To C, which is embedded in application products and provides consumers with Value-added services.

Morgan Stanley started last year to explore how to use GPT's embedding and retrieval functions to efficiently use its knowledge base system so that employees can better call the company's investment database, which is currently used by at least 200 employees daily.

A-share Companies Are Actively Exploring Commercialization in SSE

From the perspective of commercialization, the business of A-share ChatGPT concept companies can be generally divided into three routes: one is the company that wants to create ChatGPT products; the other is the upstream "water seller", such as Speechocean; The third is To C-oriented business, which directly calls ChatGPT products.

As an AI training data provider, Speechocean previously disclosed that Microsoft has been one of its customers. Because of this, the market regards it as one of the "ChatGPT concept stocks". In the survey minutes released this time, the list of institutions reached hundreds, which shows that the capital market favors it.

In addition, FII also stated that the high-efficiency computing HPC system supplied to customers by the company has been applied to ChatGPT, ChatGPT Plus, and other AI-related applications in the future. Still, the relevant revenue currently accounts for a small proportion.

In terms of To C, some listed companies are trying to embed ChatGPT-like models into their main business, but the specific direction of commercialization is still being explored. Take Jinke Culture as an example. On March 14, the company's "Talking Tom Cat" started an internal test for Android users. Recently, the stock price of Tom Cat has risen significantly. On March 17, it rose by 15.6% and has risen by 112% since February.

How to cash out with "Talking Tom Cat"? A person from the securities department of Jinke Culture said that there are certain uncertainties in the current business monetization model. The company once said in an agency survey that the company will combine the existing revenue model to expand more new monetization forms, such as traffic monetization models and recharge charging models, in addition to the advertising charging model. The company also deeply empowers emerging technologies such as artificial intelligence pre-training models in offline physical products or offline application scenarios such as the company's IP-authorized derivatives, parent-child playgrounds, and vehicle-mounted equipment.


Disclaimer: All the information on this website is provided on an “as is” and “as available” basis, and you agree to use such information entirely at your own risk. Monisight gives no warranty and accepts no responsibility or liability for the accuracy or completeness of the information and materials contained in this website. Under no circumstances will Monisight be held responsible or liable in any way for any claims, damages, losses, expenses, costs, or liabilities whatsoever (including, without limitation, any direct or indirect damages for loss of profits, business interruption, or loss of information) resulting or arising directly or indirectly from your use of or inability to use this website or any websites linked to it, or from your reliance on the information and material on this website, even if the Monisight has been advised of the possibility of such damages in advance.


Rated 0 out of 5 stars.
No ratings yet

Add a rating
bottom of page