Big models are not easy to make money with AIGC, and Wal-Mart uses it as a shopping guide

Article source: Yiou.com

Text | Fantan 123

Image source: Generated by Unbounded AI

Walmart uses AIGC as a shopping guide application, which not only embraces the spiritual inheritance of cutting-edge technology, but also opens up new landing application scenarios of AIGC.

Walmart, the world’s largest supermarket chain, announced on its official website that it will trial launch three new AIGC applications on e-commerce platforms to help users improve their shopping experience and efficiency. Walmart created these three new applications through its own massive data and fine-tuning of third-party LLM.

According to the description of Walmart’s official website and the information currently available, this is a ChatGPT-like product that automatically generates shopping suggestions, search suggestions, and review summaries based on text prompts. For example, “Do you have any good advice for buying diapers and milk powder for a 1-year-old baby?”

Previous survey data has shown that American families spend an average of about 6 hours a week on shopping planning and selection, and this process is extremely cumbersome and tedious, which is where AI can come into play, and Walmart hopes to simplify this process through AIGC.

Coincidentally, another e-commerce giant, Amazon, has also recently experimented with the same type of function, which shows that ChatGPT, an AIGC product, has a broad space for scene landing in the e-commerce field and has been recognized by industry leaders. **

01 Walmart and High-Tech

In 1962, Sam Walton opened the first Wal-Mart department store in Rogers City, Arkansas. Wal-Mart Department Store Co., Ltd. was officially incorporated in 1969 and gradually developed into the world’s largest retailer and the world’s largest enterprise.

At present, Wal-Mart has more than 2 million employees around the world, has opened tens of thousands of stores in dozens of countries and regions, and has been ranked first on the Fortune Global 500 list for more than a decade. Walmart is also very active in technological innovation, especially in e-commerce, big data and AI, and its online shopping platform (which has become one of the largest e-commerce sites in the United States.

In many people’s impressions, Wal-Mart does not seem to have much to do with high technology. However, how can companies that have been able to dominate the “Fortune Global 500” for more than a decade rely only on traditional retailers? **This combination with AIGC is not the first time Walmart has embraced cutting-edge technology, and last year it proposed the combination of digital shopping guide and virtual shopping, which can achieve an experience similar to real shopping without leaving home. Earlier, drones and self-driving cars were used to deliver goods, adding automated micro-delivery capabilities to hundreds of stores.

Just a month ago, Walmart announced that it would equip its employees with AIGC assistants in batches, and the first batch will be trialed on tens of thousands of employees to help them draft emails, summarize content, generate creative content, etc., to improve work efficiency and save time. Walmart announced it would test a Text to Shop tool that integrates AIGC apps with a common search and checkout process. Walmart’s Sam’s Club is also developing a new version of Scan & Go technology and testing a feature called Scan & Ship, which allows shoppers to choose what they want in-store and deliver it to their home.

Walmart has revealed several times before that it is exploring some of the most disruptive technologies in the coming years, while also maintaining a positive attitude towards advanced technologies from the outside. For example, language assistants and blockchain technology, as well as VR and AR, are currently used in furniture and outdoor goods, as well as virtual fitting functions for clothing.

"You can take a piece of furniture and simulate it in your living room. Let’s say you want to buy a tent, but you don’t really have to buy it and move home to find out it’s not for you. **We will mainly look at technology in terms of, does it reduce customer friction? Did it make shopping easier for them? Does having the backend give us better and faster insights and more effective reasoning on data? **”

As technology continues to evolve, it will eventually make the shopping experience easier and more enjoyable.

Although AIGC is the wave of the times for nearly a year, the use of it as a shopping guide is not an accident for Walmart. **Walmart has been using machine learning and AI for years before to help save money, create personalized experiences at scale, and increase employee productivity. **For example, the “prediction basket” feature, which uses machine learning and AI to understand shoppers’ history, preferences and what they typically buy, saving costs and creating additional experiences. Walmart is also building its omnichannel assortment planning capabilities to determine which stores need specific products and how to introduce them in stages.

Machine learning and AI allow Walmart to aggregate all of this information, synthesize it and gain insights into its essence for more dynamic management.

02 Large models and AIGC difficulties

** Last Monday, the news attracted a lot of attention, big models and AIGC huge losses, Microsoft has not yet made money, the SD of the Wensheng diagram is slightly better, but not optimistic. ** In the past year, the global technology circle’s investment in large models and AIGCs has been a top one, and great breakthroughs have been made in technological change, but in terms of profitability, it can be described as “miserable”, and each company is basically losing money and making money.

**The reason for this is mainly in two aspects: cost and value creation. **

**The costs in these two aspects should be more obvious, such as the hardware required to operate, maintenance and other technical costs. **For example, GitHub Copilot, Microsoft’s AI programming tool with 1.5 million users, plasters an average of $20 per user per month, with a maximum of $80. In other words, since Microsoft made Copilot, it has lost as much as $30 million a month alone, and you must know that Copilot cannot be prostituted in vain, and members have to pay $10 a month in “royalties”. After all, Microsoft’s products still have a membership fee system, and there is no large model of fees and AIGC enterprises and products, so “pouring money” should have to go up an order of magnitude. Another well-known is OpenAI, which previously broke the news that the daily cost of running ChatGPT alone could be as high as $700,000.

**This is the cost of technology, as well as the cost of copyright used for promotion. **In order to avoid the high litigation costs caused by AI infringement, some companies have begun to spend a lot of money to “buy” celebrity images and voice rights. In the past two days, a very hot news shows that Meta is paying for the celebrity portraits behind its AI Chatbot. Celebrities only need to work with Meta for 6 hours, record portraits and other information for “chatting with the public”, and Meta’s single payment is as high as millions of dollars. It is worth mentioning that Meta has paid dozens of celebrities for this.

But such a high cost does not seem to bring the imaginary “gold-absorbing power” to AI and the companies that use it. On the one hand, the “incremental effect” of AIGC and the big model on the company’s other products is not as violent as imagined. **Microsoft has been emphasizing that “OpenAI uses Microsoft’s intelligent cloud services”, but at present, the publicity effect of this move is not obvious. **On the other hand, the landing scenarios and application value of large models and AIGC itself still need to be tapped. **

Not long ago, Sequoia Capital’s article Generative AI’s Act Two mentioned that Whether it is the user retention rate in the first month or the daily use of current users, compared to other products, the current large models and AIGC usage is not optimistic. Compared with the highest retention and daily/monthly activity rate of ordinary products, the rate of users who are willing to continue to use AIGC products after experiencing them, or open them every day, is not high, only 56% and 41%. A survey of 1,600 scientists in Nature found that only 4 percent of those scientists who used AI in their research thought AI tools were now “essential.”

** After such a comparison, it is not difficult to find that Walmart’s ChatGPT products such as shopping assistants, search assistants, and comment assistants are not an active practice to promote the landing of large models and AIGC? **Wouldn’t it be a pity to have good technology but not actively promote its application? This is the dilemma of the current big model and AIGC, and related products and services have not been really recognized by most users, and more are still watching the excitement. Historical experience has shown more than once that there is no landing application, especially transformed into products that ordinary people can reach, no matter how advanced the technology is, it is no different from a castle in the air. Although ChatGPT can speak well today, Microsoft’s GitHub Copilot has millions of users, but there are many problems. At the same time, Wal-Mart is also a company with a global market, and compared with ChatGPT’s eloquent but often serious nonsense, it may still be more grounded in supermarket shopping. **