One Man Conglomerate, #02. So, the generative AI market will grow rapidly [PART I]

The following is English version of Chapter 2 of ‘One Man Conglomerate’, published in August 2023 by Do-Jeon, Marcus, Jeong.

 

2nd Chat: How will our lives changes?

 

#02. So, the generative AI market will grow rapidly.

 

[01] AI chatbot services equipped with ChatGPT will flood the market.

 

In the future, various AI chatbots utilizing ChatGPT’s API will be poured out like a flood. ❶ First of all, the number of AI chatbots on websites will increase significantly. Of course, there are already AI chatbots on the websites of most companies (i.e. business entities). They are mainly intended to improve the quality of service to their customers (old and new) and reduce labor costs. However, the quality of chatbot service was not very satisfactory. But now, these chatbots will be upgraded to AI chatbots with improved performance, utilizing ChatGPT’s algorithms. And of course, the AI agents who communicate with customers by voice will also become smarter.

In addition, ❷ differentiated AI chatbots equipped with a large language model (LLM) such as ChatGPT, which can be recognized as a single independent service (or product or business) Note that these services must be differentiated from ChatGPT and other AI chatbots. Otherwise, they will not be able to meet consumers’ expectations.

A representative example of a differentiated AI chatbot service is ① Upstage’s AI chatbot AskUp (which is called Ah-sook-up in Korean pronunciation). AskUp is an AI chatbot service launched on Kakao Talk by adding GPT-4 and Upstage’s Optical Character Reader (OCR) technology.

AskUp is not just a conversation service like ChatGPT. Specifically, Asukup is a chatbot that uses GPT-4 to search for the latest internet information, and it also uses OCR technology to recognize letters in images as text and then summarize or translate them. If you upload a photo of food, it provides various information such as ingredients and calories. And it can also generate a text-to-image. It has become so popular that it has been installed on LINE since May 5, 2023.

In addition, many differentiated models will emerge to overcome the limitations of general-purpose LLMs such as ChatGPT, which are good at answering ordinary, everyday questions like humans, but fall short of expectations when it comes to specific (or specialized) areas. These are called ② “Domain-specific AI (or GPT or LLM) models”.

In this context, domain can be interpreted as “field”. So this is a way to create AI models that are optimized for a specific field, i.e., with added expertise. For example, AI chatbots can be optimized (or specialized) to provide services in fields such as finance, law, health, sports, gaming, education, foreign languages, IT, etc. Of course, the number of market players offering such services is growing rapidly. But it’s not enough yet, because the market is still in the process of being created.

In addition, there will be a significant increase in the number of ③’AI agent models’ that are slightly different from these. A typical example of this is ‘AutoGPT’. AutoGPT is a GPT-4-based AI chatbot released on GitHub, a developer’s code sharing platform, on March 30, 2023, by a developer named Significant Gravitas (a British indie game developer).

However, unlike other AI chatbots, when a human sets a goal, AutoGPT performs tasks on its own, without human intervention, to complete the goal. Currently, AutoGPT is open-sourced and anyone can run it on their personal computer. Of course, it’s still in the early stages of development, so it’s not perfect. However, many people have praised it, saying that it shows promise for AGI (the second stage of AI development). For reference, there’s also an AI chatbot called AgentGPT that does the same thing.

The important point here is that it does not require huge capital to create AI chatbots that can be recognized as one independent service. Therefore, it is very likely that individuals will also actively participate in this competition in the future. This means that individuals with LLMs (including GPTs) can compete with existing companies by offering various AI chatbot services. Therefore, companies (not only AI model developers but also general companies) will be in a difficult situation.

We are moving away from an era where scale was a competitive advantage. If the weapons are the same, the market can be dominated by the competitiveness of the product alone, because AI models that meet the needs of consumers will quickly spread through the Internet community (including the global community). The above-mentioned AutoGPT and ChatGPT have reached this stage through word of mouth alone.

 

Photo: Developed by Midjourney

 

 

[01-1] AI chatbots will become a necessity for businesses.

 

In the future, LLM-based AI chatbots will become essential assets for business. Because, if utilized properly, ❶ they can improve business efficiency in the medium and long term, as well as ❷ explore new business opportunities. And, the most important reason (especially for executives) is that ❸ competitors will use it very aggressively.

However, as mentioned earlier, funds on an astronomical scale are needed to make LLM such as ChatGPT directly. And even if these funds are invested, there is no guarantee that it will be successful. Therefore, many companies will continue to use AI chatbot’s API with proven performance like ChatGPT.

Companies will use AI chatbots in the following ways. ❶ First, they will use AI chatbots to improve service quality. This is the most common approach. For example, ① AI chatbots are used for customer service. This includes (a) keyboard-based AI chatbots that support a variety of languages, including Korean; (b) multimodal AI agents that can communicate via voice (including foreign languages), sign language, and gestures; and (c) AI chatbots that can assist agents in responding to customer requests based on information about the customer (past history + future predictions), where the AI first generates an answer and the agent uses it to respond to the customer. ② The second is an AI salesman that automatically recommends the most suitable product (or product or service) for the customer, explains the reason for the recommendation, the differences between each product, and the details of payment and delivery (interest-free installment, delivery date, etc.), and then connects it to the sale. ③ And finally, AI VOC (Voice of Customer) chatbots that listen to the voices of various customers, monitor (including measurement and evaluation) the handling of customer complaints and suggestions, and analyze various information learned in the process and reflect it in future product development.

For example, a company that distributes women’s clothing is adding AI chatbots to its online shopping malls. Of course, this AI chatbot must be an optimized model for the domain, which has undergone prompt engineering or fine tuning. TThis AI chatbot comprehensively analyzes the search history of shopping mall users and recommends the most suitable product for the customer (at this stage, if the customer’s photo is entered, it shows the video of the customer wearing the clothes), and explains the characteristics of the product, how to use it, interest-free installment plans, delivery methods, etc., if necessary. Of course, the ultimate goal is to make an order. In this process, feedback from customers is analyzed and reflected in the product development process. Of course, it will respond more quickly to various requests from customers, such as refunds, in various languages 24/7, year-round, 365 days a year. Then, of course, the company’s service quality will improve and customer satisfaction will increase.

In addition, ❷ AI chatbots specialized for specific tasks or organizations will be used. In other words, companies will actively use AI chatbots optimized for specific tasks or individual organizations within the company to improve efficiency. Of course, most will use ready-made products (such as ChatGPT or Bard), but some companies will look a little further ahead and develop their own AI chatbot(including LLM). This is because, then, the concern of leakage of corporate information disappears. For example, it is possible to develop an AI chatbot optimized for ‘corporate strategy and budgeting tasks’. This is because, in general, the task of establishing strategies and budgets is important enough to determine the future of the company, but the difficulty of the work is quite high, so even those with sufficient experience and knowledge are very difficult. And it takes quite a bit of work time. In other words, it is difficult to perform properly in the 52-hour workweek. So, if AI chatbots are used for this task, companies will be able to (a) complete their strategy and budgeting tasks much more efficiently than in the past, and (b) human error and risk will be significantly reduced in this process. In addition, (c) the progress of the project will be monitored meticulously during the implementation of the strategy, and (d) appropriate alternatives (e.g., modifying the strategy or budget) will be available in a timely manner even if unexpected business issues arise.

For the above reasons, companies will actively use AI chatbots specialized for each domain (i.e., business field). For example, there will be a number of domain AI chatbots optimized for tasks such as sales, marketing, CRM, risk management, production, purchase, human resources, and finance. And companies that apply Agent GPT, such as AutoGPT, will also increase significantly.

 

[ Attention ]

The above article excludes footnotes and details, unlike the actual book.
Please read the following post for details of “#02”

위 글은 정도전 작가가 2023년 8월에 출간한 ‘1인 대기업(One Man Conglomerate)’의 챕터2 영문 버전입니다.