ChatGPT: Are enterprises ready for the next generation of natural language processing technology?

Network analysis data released by

Digital-adoption.com found that after the release of the latest version of ChatGPT, the number of visits to its developer OpenAI increased by 3572%, from 18.3 million to 672 million.

In January this year, Microsoft announced its third phase investment plan for OpenAI, with a total amount of billions of dollars. Now, Microsoft has launched an AI-supported version of the Bing search engine, which uses ChatGPT. Company CEO Satya Nadella believes that AI will fundamentally change various software categories, and Microsoft hopes that this technology can support Bing to answer 5 billion queries every day. Currently, the ChatGPT function is still in the preview stage, but users can sign up and wait in line for access.

Just one day before Microsoft’s announcement, Google and Alphabet also released a competitor called Bard. This competitor is based on Google's Lambda technology and is said to be able to support more users through expansion. A blog post discussing Bard reads, “Soon, users will see AI-powered features in search. These features will distill complex information and multiple perspectives into easy-to-understand formats, allowing users to quickly understand the big picture and Grasp more information from it."

People also saw the opportunity to use generative AI within the enterprise.

Carolyn Prior, head of data, AI and applications practice at Kyndryl UK and Ireland, said, “Based on our observation, we found that those who can be data-driven and truly integrate AI Organizations that operate systems, as well as those that actively explore the latest/best emerging technologies and modernize their data management architecture, can better navigate the current market competition. And their embrace and adoption of AI and other emerging technologies , will become the key to maintaining competitive advantage."

Knowledge platform provider eGain believes that software developers are increasingly incorporating generative AI into their product and service development. The company has integrated ChatGPT into its Instant Answers product. Company CEO Ashu Roy emphasized, “Generative AI such as ChatGPT opens up exciting automation possibilities in knowledge management and conversational participation. Instant Answers is very popular among our invited customers, who like the value it creates based on speed. ”

Knowledge management also helps software developers work more efficiently.

Romy Hughes, director of Brightman Business Solutions, said, "ChatGPT can help software developers complete a very challenging piece of code." For example, developers can ask ChatGPT how to optimize existing code. Looking forward, she believes “the technology will democratize programming by giving non-coders the possibility to develop apps — essentially the same benefits low-code promises, just with far more functionality.” The wave of democratization will allow organizations to establish new innovation processes that do not involve the IT department and maximize the ideas of other employees. ChatGPT will make this easier."

Arthur D Little AI company expert Albert Meige and Gregory Renard analyzed the risk of this technology lowering the overall technical level of programmers, and warned that the generated code is likely to contain algorithmic biases. When bias exists in the data used to train an algorithm, it can lead to inaccurate or unfair results. According to Meige and Renard, despite efforts to eliminate such biases, they still exist and are quite troubling.

They also cited an example of analysis by ChatGPT, which focused on North Korea, Syria and Iran when asked to write a Python program to determine whether someone should be imprisoned based on their origins. . Moreover, this criterion is hard-coded into the Python code, and this bias is likely derived from extensive analysis of the content of conversations on the Internet.

Although this example itself is very simple, after all, there are certainly other factors to consider when judging whether imprisonment should be implemented, but it also shows that using ChatGPT to generate code does have risks, and its model can easily be biased/biased in the training data misled.

Another problem faced by enterprises is the huge cost of generative AI training, especially the high risk that potential biases in training data may bring. "Harvard Business Review" recently published a report stating that training generative AI is still a patent of a few technology giants because the training process consumes a large amount of data and computing power. The author of the article cited data from the GPT-3 model on which ChatGPT is based, saying that 45 TB of data material was used for initial training and 175 billion parameters were used to make predictions. Therefore, the single training cost of GPT-3 is as high as 12 million US dollars.

The authors warn that “most enterprises do not have the corresponding data center or cloud computing capacity to train their own similar models from scratch.”

From this perspective, it is the first time to It makes sense that Microsoft and Google have introduced generative AI into their own search engines. As for other organizations that do not have such giant-level computing facilities and data reserves, GPT technology may not be able to bring practical help in the short term.

The above is the detailed content of ChatGPT: Are enterprises ready for the next generation of natural language processing technology?. For more information, please follow other related articles on the PHP Chinese website!