Amazon, which was publicly dissed by Gates: the "underachiever" of generative AI

"The 'AI assistant' will become the standard of mass life. It will disrupt the way of consumption and production, and people will never use search engines or use Amazon again.”

Like Chat GPT

On May 22, Microsoft founder Bill Gates published his prediction for the future of society at a campaign to discuss artificial intelligence. Gates believes that the fierce competition in the tech industry is accelerating the pace of AI assistants into consumer life, and he hopes Microsoft will go the farthest in the AI war.

 Mr.Gates certainly has the confidence. After all, Microsoft, which is heavily tied to OpenAI for just $1 billion, has a advantage in a competition dominated by generative AI.

Amazon is not uncomfortable. According to foreign media reports, an Amazon spokesman declined to respond to Gates's comments.

 To some extent, Amazon should not have lost its voice in the "iPhone moment" of AI. On the one hand, the construction and training of big models need to be based on computing power, so the mainstream players in the market are basically large cloud service providers or their Allies, while Amazon AWS is the global cloud service provider; on the other hand, from a broader perspective, Amazon is the company that productized AI technology earlier.

 Back in 2014, the  generation Echo device was supported by AWS'Alexa system.

Alexa was able to perform tasks at consumer orders and start software and hardware services, when the interactive experience was evident in the consumer market, creating a boom in smart speakers. By the end of 2021, a quarter of US homes had at least one Alexa device, according to data provided by Amazon.

Alexa is the beginning of many consumers' exposure to the AI assistant concept and the AI robot dialogue mode. At the time, who would have thought that Amazon would have become an underachiever in AI.

01

 Amazon founder Jeff Bezos was directly involved in the testing and development of Alexa, even designing the look and language himself.

 Bezos had two wishes in his heart.

 Alexa wants to be everywhere in consumers' life, not only as a fixed home and office space, but also to accompany consumers at any time, just like a smart phone to become a necessity of life. Alexa has been actively integrated with third-party devices in recent years. Amazon says its partners have created more than 140,000 Alexa-compatible products and control more than 300 million smart devices.

 The second desire is that Alexa can adversely motivate consumers to shop through dialogue to return to Amazon's core business and create greater business value. Currently, more than 50 percent of Alexa users use it for shopping, but official statistics don't show how much sales Alexa drives. It seems that  results are just a way to log on to the platform.

For a long time, Alexa has been an important part of Amazon's strategic layout. Amazon sold its hardware almost at cost, targeting the value added to smart services. However, Alexa's profitability situation has not improved.

Alexa uses a command and control system that can understand a limited list of questions and requests, but once handling complex conversations, it goes from AI to "artificial retarded."

 In 2019, Alexa used a conversation with users to quote a maliciously modified Wikipedia page to commit suicide. In order to improve the experience, Amazon introduced real employees to recognize consumers' voice clips and help AI process data, which also triggered market fears that voice assistants are encroachment on private space.

 The conversation between users and Alexa is basically focused on simple conversations such as switching music and asking about the weather. Amazon has tried to expand, but Alexa's system is far less flexible than generative AI. Adding simple conversations leads to a lot of bugs, and small snowflakes bring a snowball workload.

 The awkward user experience led to Alexa's user churn rate, and Amazon wasted its stature in the AI space. In media interviews, insiders directly described Alexa as a "great failure of imagination" and "a wasted opportunity".

 Alexa's global digital services segment, where it is based, lost $10 billion in 2022. By the first quarter of 2023, Amazon's Worldwide Digital division (including everything from Echo smart speakers and Alexa voice technology to Prime Video streaming services) had lost more than $3 billion in its operations.

 As a result, at the end of 2022, Amazon made historically * -scale layoffs, and Alexa's division was hit hard by the layoffs.

 In fact, the smart speaker dilemma is not only for Amazon, but the signs are that without innovating the technical core of the product, it will be seen as the last product of the global era.

 Sales of smart speakers in China have also fallen for two years in a row, according to data from Loto Technology.In the first quarter of this year, it fell by more than 20% year-on year.

 Just when the voice assistant was considered, ChatGPT changed everything. Generated AI has got rid of the embarrassment of "artificial mental retardation", and made AI robots become an industry tide with hundreds of millions of monthly active users in just three months.

Amazon wants to find a better place of power in the volatility.

According to an Amazon employee: " The ChatGPT has made a splash. Some executives gave instructions for the team to brainstorm and discuss how to make Alexa smarter.”

02

"Given the current challenging macroeconomic conditions, AWS faces short-term headwinds."Adam Selipsky, as the CEO of AWS, said in a letter to his employees in early 2023.

 In 2022, Amazon posted a net loss of $2.7 billion, the worst performance in the company's history, and net profit fell far more than market expectations.

More importantly, AWS cloud services have narrowed their operating margins for four consecutive quarters. Although AWS only occupies 16% of Amazon's business volume in terms of * size, it carries the outside world's expectation for Amazon's imagination space and development potential.

 In the quarter of 2023, the AWS growth rate was only 16% year on year, which is the * value of AWS since the establishment of it. Olsavsky, Amazon's chief financial officer, admitted: " In the face of severe macroeconomic conditions, companies large and small are evaluating how to optimize their cloud computing spending.”

By contrast, Microsoft's cloud business Azure and Google's cloud computing units grew 27% and 28%, respectively in the latest quarter. It seems not appropriate to attribute the decline in growth to corporate costs. In the explosion of generative AI, budget-conscious companies prefer companies in both AI and big models.

If AWS continues to decline, it will damage market confidence in Amazon. Amazon naturally needs to reverse the pessimism of the outside world —— is a transition, pain is inevitable.

 Amazon CEO Andy Jasi believes that large language models and generative AI are the core that will allow Amazon to innovate in every area of its business in the coming decades. He cited the well-performing advertising business as an example. " The strong growth is largely due to our continued investment in machine learning. The addition of AI and large models provides technical support for more precise marketing. With unusually strong sales results for the brand.”

While Gates expects AI to replace Amazon's online platform, Amazon is also thinking about taking advantage of the Echo's penetration rate to find its own generative AI landing site, integrating AI into the physical environment of real life, and depicting a more imaginative future.

 In the view of the industry, the fusion of generative artificial intelligence and voice assistant is only a matter of time. For example, in April, with the release of Ali Tongyi big model, Tmall Spirit announced the full access to "Tongyi Thousand Questions".

During the earnings call, Jasi stressed that Amazon has sold more than 100 million Alexa-powered devices worldwide, and that the company is building a "larger, more versatile and more powerful" language model as the basis for Alexa.

"I think this will greatly accelerate our vision of becoming the world, and there is a big business model behind it.”

It was revealed that Jasse plans to transform Alexa with generative AI like ChatGPT to make it more like thinking about users' questions than extracting information directly from the database.

 Sources revealed that Amazon is integrating resources and evaluating benefits for "strengthening generative AI and large language model development," and even shifting " some of its spending on its retail business to cloud services."Currently, Amazon's plans for Alexa focuses on entertainment convenience, including smoother video search, personalized recommendations, and more interactive communication and information refinement.

 Amazon's goal * is for more than just entertainment.

According to leaks, Amazon is secretly developing an upgraded version of its home robot, Astro, according to leaked internal documents. Burnham The robot will introduce "situational understanding" technology based on large language models (LLM) and other advanced AI models. In the vision, it can keenly observe the surrounding, intelligently receive and understand what is heard, and have a dialogue with the user, and take appropriate action on this basis.

If an old man slips, Burnham can check that he's safe, automatically call the police and call someone else to help.

03

From an objective point of view, Amazon's 34% market share in cloud computing is still distant. But under the AI frenzy, Amazon has become an "underachiever". It is an inevitable choice for Amazon to catch up and further intelligent AWS along the technical route of generative AI.

In April, AWS announced the launch of a generative AI service called Bedrock to specific users.

 Jasi said: " Most companies want to use large language models, but really good language models take billions of dollars and years to train. But not all companies have this condition, and people expect to upgrade from an already very large basic model and customize it for their own purposes.”

 Bedrock The goal is to lower the barriers to entry for generative AI.

This is a service to build and expand generative AI applications that consists of two parts: AWS's own model, Titan, and a basic model from startups AI21 Labs, Anthropic and Stability AI. AWS enterprise customers can use their own data to customize their base models, and design and build generative AI applications.

Amazon hopes Bedrock to become a basic tool for small and medium-sized enterprises to generate AI, allowing it to earn additional benefits far beyond the consumer side.

 Coda, a collaborative documentation software company, was among the first users, and Coda's chief executive said in an interview that he gave Bedrock an "unfinished" rating. He believes the tool is still in the early stages of development, and that Amazon is "developing and repackaging it on top of its existing service," but he believes it will be competitive in the long term.

 From the current feedback, Bedrock's performance is not yet able to work, and Amazon has been delaying expanding the scope of the Bedrock trial, appearing to be adjusting on the sidelines.

 After years of explosive and rapid development, Amazon has returned to the preparation period of accumulating potential energy.

 Using Bezos's "flywheel theory," the generative AI race has to make a change to build a new flywheel. Although it takes a lot of effort to make the flywheel turn again, the potential energy of the flywheel will eventually be part of the driving force.

The wave of generative AI is just beginning, and Amazon, which has started a little slower, has endless possibilities.