(Yicai) Jan. 4 -- The artificial intelligence industry is still facing many challenges, including practical applications and data monopolies, according to one of the founding members of Alibaba Group Holding’s cloud business unit.

There is a missing intermediary link in the AI-generated content industry’s chain, Wu Hanqing told Yicai today. To-business and to-customer AI applications both need an intermediary link that connects them with large language models’ CPUs, Wu noted, adding that personal AI computing servers can do that.

Wu was Alibaba’s second security engineer. He helped establish the security systems of Alibaba, Taobao, and Alipay. Wu left Alibaba Cloud in May 2022 to join hands with two other co-founders and establish KMind, a developer of personal AI computers.

Many companies have high expectations for LLMs’ capabilities but find it hard to create practical value in the actual production process upon implementation, Wu explained.

One of the main problems is the illusion of LLMs, according to Wu. Many people may lose confidence in AIGC products because LLMs sometimes provide users with wrong, inconsistent, or even fabricated answers.

LLMs can only generate one-time results, meaning that modifying one element can bring a completely different set of results, Wu noted.

Last July, KMind officially launched hikOS, a brand aiming to provide personal AI computing services to ordinary users. All of KMind’s products, including recommendation and retrieval algorithms, have been open-source since the establishment of the company.

The drawback of data monopolies is that users contribute data, while big companies retain all profits, leading to issues such as data taxation and inflated product prices, Wu added. Moreover, data monopolies make the internet territorial and reduce user efficiency and experience, thus undermining the freedom of the internet, he pointed out.

Editor: Futura Costaglione