China’s livestream shopping industry is booming, with over 500 million viewers and $4.6 trillion in sales in 2023. But behind the scenes, a new technology is fueling this growth: artificial intelligence (AI). AI is enabling brands to create virtual hosts that can speak any language, interact with customers, and sell products 24/7. These AI hosts are not only cheaper and more efficient than human streamers, but also more adaptable and customizable to different markets and audiences.
The rise of AI hosts in China
Livestream shopping, or live commerce, is a form of online shopping where hosts showcase and sell products in real-time through video platforms. It is a popular and lucrative business model in China, where consumers can watch, chat, and buy from their favorite streamers. Livestream shopping combines entertainment, social interaction, and convenience, creating a unique and engaging shopping experience.
However, livestream shopping also has its challenges. It requires a large pool of talented and charismatic hosts who can attract and retain viewers, as well as a lot of time and resources to produce and manage live shows. Moreover, livestream shopping is limited by language and cultural barriers, making it hard for Chinese brands to expand to overseas markets.
This is where AI comes in. AI can help solve these problems by creating virtual hosts that can mimic human streamers, but with more advantages. AI hosts can:
- Speak any language and dialect, making it easier for Chinese brands to reach global customers
- Work around the clock, without fatigue or boredom, increasing the frequency and duration of live shows
- Adjust their appearance, voice, and personality according to the product, audience, and occasion, enhancing the diversity and appeal of live shows
- Learn from data and feedback, improving their performance and interaction over time
The technology behind AI hosts
AI hosts are powered by generative AI, a branch of AI that can create new content from existing data. Generative AI uses deep learning, a technique that trains neural networks to learn from large amounts of data and generate outputs that resemble the data.
One of the most common and powerful generative AI models is called generative adversarial network (GAN). GAN consists of two neural networks: a generator and a discriminator. The generator tries to create realistic outputs, such as images or videos, while the discriminator tries to distinguish between real and fake outputs. The two networks compete and learn from each other, improving the quality and diversity of the outputs.
GAN can be used to create realistic and expressive faces, bodies, and voices for AI hosts. For example, one of the leading AI host providers in China, Midjourney, uses GAN to generate lifelike avatars that can speak, move, and emote. Midjourney claims that it only needs one minute of video to train an AI host, and charges about $1,000 for each avatar.
Another generative AI model that is widely used for AI hosts is called neural machine translation (NMT). NMT is a technique that uses neural networks to translate text or speech from one language to another. NMT can learn from large corpora of parallel texts or speeches, and produce fluent and natural translations.
NMT can be used to enable AI hosts to speak any language and dialect, as well as to understand and respond to customer queries and comments. For example, another AI host provider in China, Zhiyin, uses NMT to translate Chinese speech to English, Spanish, French, and other languages. Zhiyin also uses natural language processing (NLP), a technique that analyzes and generates natural language, to create interactive and personalized dialogues for AI hosts.
The future of AI hosts and livestream shopping
AI hosts are not only disrupting China’s livestream shopping industry, but also opening up new opportunities and possibilities for e-commerce and entertainment. AI hosts can help Chinese brands to reach new markets and customers, as well as to differentiate themselves from competitors. AI hosts can also provide more choices and fun for consumers, who can enjoy watching and interacting with various AI personalities and styles.
However, AI hosts also pose some challenges and risks, such as ethical, legal, and social issues. For example, how can consumers trust and verify the authenticity and quality of the products and information presented by AI hosts? How can AI hosts protect the privacy and security of the customers and the data they collect? How can AI hosts avoid infringing the intellectual property and personality rights of the human streamers they imitate? How can AI hosts ensure that they do not harm or offend the customers and the society with their actions and words?
These are some of the questions that need to be addressed and regulated as AI hosts become more prevalent and influential in the livestream shopping industry. AI hosts have the potential to revolutionize e-commerce and entertainment, but they also need to be responsible and ethical.
