Shulou(Shulou.com)11/24 Report--

Recently, aurora, a leading customer interaction and marketing technology service provider, announced the launch of the GPTBots platform, which supports one-click access to multiple mainstream models, while allowing developers to seamlessly connect LLM (large language model) with enterprise data and service capabilities, build their own AI Bot services, and open registration.

"artificial intelligence is the biggest opportunity for productivity change in the next 10 years, and every enterprise has to embrace it. Whether it can make good use of AI in the future is a very core competitiveness."

Luo Weidong, CEO of aurora, believes that generative AI is experiencing explosive growth, with applications involving well-known companies such as Microsoft, Microsoft, Salesforce, Meta and Adobe. This brings great development opportunities for enterprises, but at the same time, there is also a cognitive and technological gap.

In his view, when many enterprises use technologies such as GPT, they are often limited to generating articles or images, and fail to make full use of their internal data and knowledge. This leads to huge opportunities and potential loss of benefits. Enterprises need to integrate technologies such as GPT into their business processes to improve efficiency and create new opportunities.

At present, the domestic large model is booming, no matter, Ali, byte and other big manufacturers have launched their own large models. However, in terms of application, generative AI still has the problem of hallucination, and many general language models lack vertical domain knowledge, which leads to the inability to solve complex tasks.

For enterprises, the application of AI model involves many links, such as compliance, data, computing, engineering and algorithms, and quality problems in any link will affect the implementation of the application scenario.

For example, in a question-and-answer scenario, sensitive personal information, such as health insurance card accounts, medical records, and drug use, is usually required. If there is a data call error in the AI application system, it may adversely affect the subsequent diagnosis. In this case, the AI model specifically designed for the scene can play a key role in marking potential private or critical medical record information and alerting back-office doctors or customer service personnel for review, which will reduce the cost of consultation and reduce potential risks.

In Luo Weidong's view, these problems can be solved by creating a custom AI BOT (or AI Agent)

To solve the problem, combine the internal data and domain knowledge of the enterprise, and integrate into the process business through API. This method is not only suitable for large enterprises, but also can be easily applied by small companies that do not have a lot of R & D resources. Enterprises should actively explore how to use these technologies, regardless of whether they have a large amount of data or not.

He said that it was precisely to see the needs of enterprises that Aurora launched the GPTBots development platform based on the accumulation of developer services over the years. On the auroral GPTBots platform, developers only need to have generative AI and related domain knowledge to build their own AI Bot and connect it to enterprise services, including out-of-the-box functions such as LLMs, Plugins, long / short memory, knowledge base and flow to help enterprises tailor AI Bot applications.

In the future, generative AI will become a tool for enterprises to improve efficiency and save costs. However, Luo Weidong believes that to maximize the use of generative artificial intelligence, the gap between enterprises will not be narrowed, but will be widened by the landing of AI applications.

From the birth of each revolutionary technology to the landing application, it needs a certain acceptance time, and at the same time, it is also faced with user privacy and data security, followed by the optimization of the algorithm and the iteration of the model. As far as enterprises are concerned, they need to study how to combine generative AI technology with their own business to the maximum extent in order to avoid being eliminated by the times.

The following is a transcript of the interview (edited on the basis of the original text):

After the popularity of Q:ChatGPT, many senior executives will have a great sense of anxiety. How do you feel?

Aurora CEO Luo Weidong: I don't think it's an exaggeration to compare it to the iPhone moment. I'm a continuous entrepreneur, and I clearly remember the moment when iPhone came out, and the two feelings were very similar. However, I also believe that it still needs time to achieve the maturity of various applications, just like after the advent of iPhone, the application market did not appear until a year later, and the application ecology did not emerge until 2 or 3 years later.

Q: aurora has released GPTBots. What opportunities do you see?

Aurora CEO Luo Weidong: when many enterprises use technologies such as GPT, they are often limited to generating articles or images, and fail to make full use of their internal data and knowledge. This leads to huge opportunities and potential loss of benefits. Enterprises need to integrate technologies such as GPT into their business processes to improve efficiency and create new opportunities. This can be achieved by creating a custom AI BOT that combines internal data and domain knowledge into the process business through API.

This method is not only suitable for large enterprises, but also can be easily applied by small companies that do not have a lot of R & D resources. Enterprises should actively explore how to use these technologies, regardless of whether they have a large amount of data or not. Based on the troubles of the enterprise, we launched the GPTBots development platform.

Q: can you elaborate on the plug-in features of GPTBots?

Aurora CEO Luo Weidong: first of all, we need to understand what is the plug-in of LLMs: the LLMs plug-in is different from the traditional plug-in, it is independent, flexible, free and powerful. The core of the big language model plug-in is Web API, so the big language model plug-in completely embraces the Internet, at the same time, there is no development language "discrimination", no matter the developer's language stack is Python, Java, Go, PHP, etc., as long as you can develop HTTP protocol interfaces and follow RESTful rules You can build a large language model plug-in. GPTBots plug-in integrates ChatGPT plug-in creation specification (general OpenAPI specification) and function call function, so that developers only need to focus on their own functional interface development, while GPTBots will automatically be compatible with the mainstream LLM that supports plug-in capabilities in the market.

For example, in the case of a mail service, if a customer wants to send a letter from Singapore to the United States, they may ask about the cost of the service. This price may be constantly changing and stored in a structured database in which case we can provide a price query plug-in through which we can call and get this dynamic data.

Developers can create plug-in customizations according to their needs, which are usually used to deal with structured data services within the enterprise. For example, the HR department of an enterprise could develop a plug-in to query employees' holiday balances.

Q: what are the problems for enterprises to apply large models at present? How does GPTBots help enterprises solve these problems?

Aurora CEO Luo Weidong: if an enterprise wants to develop an AI Bot, developers need to have certain technical skills and related domain knowledge. From academic theory to commercial products, it involves professional knowledge of machine learning tools, programming languages, frameworks, front and rear engineers, operation and maintenance engineers and so on. This is a challenge for enterprises.

On our platform, developers only need to have generative AI and related domain knowledge to build their own AI Bot and connect it to enterprise services. We provide a simple, efficient and secure LLMOps platform, including out-of-the-box functions such as LLMs, Plugins, long / short memory, knowledge base and flow to help developers tailor AI Bot applications for their business.

Q: what do you think of the relationship between open source and closed source models? Will GPTBots support all of them?

Aurora CEO Luo Weidong: the development of open source in the field of AI is really strong, and the open source community is constantly innovating and iterating to catch up with some leading native technologies. Open source is likely to become mainstream in the future because it meets the needs and needs of enterprises of different sizes.

For small and medium-sized enterprises, they can use OpenAI or Ali's cloud services to meet their needs while reducing technology tariffs. For large enterprises and government departments, they may prefer to use privatized deployment of open source models to ensure data privacy and security. Therefore, GPTBots also provides a solution that supports privatization deployment combined with the open source model, which can meet the needs of customers of different sizes and nature.

Q: what can we do about the data security that large enterprises are generally worried about?

Aurora CEO Luo Weidong: we have provided some solutions to the security concerns of enterprises. If enterprises use a public model, they can take some measures to ensure the security of the data, such as extracting the user's sensitive information during transmission. If the user asks questions about personal physical symptoms during the Q & A, the enterprise can desensitize the sensitive information, such as name and card number, and then submit the desensitized data to the GPT model. GPT then returns the answer and replaces the sensitive information in the answer to protect the user's privacy.

Another solution is to filter data within the enterprise to ensure that sensitive information is not submitted to the model. Enterprises need to deal with this process on their own to avoid risks. Enterprises can also choose to deploy the open source model locally, but this requires higher costs, including hardware and GPU.

What is the profit model of Q:GPTBots?

Aurora CEO Luo Weidong: for the public version, we use a prepaid subscription and price it according to usage and features. We provide the standard SAAS pattern, which we have been using before. If the enterprise chooses to deploy locally, we provide the pricing of the privatized version and related services. The price of a privatized version usually includes the cost of hardware and services.

Q: is there anything else you'd like to share?

Aurora CEO Luo Weidong: artificial intelligence is the biggest productivity change opportunity in the next 10 years. Every enterprise has to embrace it. Whether it can make good use of AI in the future is a very core competitiveness. The emergence of AI-BOT development platform not only provides opportunities for enterprises to apply practice, but also brings new impetus to the development of the whole AI industry.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.