OpenAI's strategy has changed? Reports suggest that Altman wants OpenAI to become a "super assistant" and directly compete with Microsoft and other downstream software.

OpenAI and Microsoft have had a "crack" in their relationship for some time, and now the competition between the two is becoming more intense, with a "love-hate" drama unfolding.

On June 27th, according to The Information, in just half a year, ChatGPT has become one of the most popular internet brands worldwide. Now, OpenAI is developing even bigger plans for ChatGPT, which could mean taking business away from partners like Microsoft and Salesforce.

OpenAI CEO Sam Altman privately told developers that the company hopes to turn ChatGPT into a "super-intelligent personal work assistant" that can perform a variety of tasks based on personal and work needs, such as drafting emails or documents in the user's style and providing the latest information on relevant business.

It is worth noting that Microsoft already wants to turn Windows Copilot into a real "personal assistant", and if OpenAI wants to develop ChatGPT into a future entry-level platform, it will undoubtedly conflict with Microsoft's interests, and competition will intensify. Salesforce is also planning to integrate ChatGPT into its various software, and if ChatGPT becomes a "super assistant", it may also face positive conflicts with Salesforce.

According to Altman and two other sources familiar with the matter, building new ChatGPT features will be a key focus of OpenAI's commercial efforts.

Therefore, in the short term, both Microsoft and OpenAI can provide technical services to B2B customers who need to build AI capabilities, and there is direct business conflict between the two. In the long term, if OpenAI accelerates its layout of software for individuals and enterprises, ChatGPT is likely to reshape the C2C application ecology, and perhaps the "breakup" of the relationship between the two is a matter of time.

ChatGPT is to become a "super assistant"

According to minutes of a closed-door meeting leaked in early June, Altman mentioned that ChatGPT's mission is to attract more enterprises to access its API, rather than to take away customer traffic.

Altman has "reassured" customers that OpenAI will avoid direct competition with them and will not release more products than ChatGPT in the future:

Many developers are concerned that when they use OpenAI's API to build products, they are worried that OpenAI will eventually release their competitors. But OpenAI will not release more products than ChatGPT in the future.

Historically, every great platform company has a "killer app", and ChatGPT's role is to help customers make their products better. ChatGPT's vision is to become a super-intelligent work assistant and will not touch on other fields, so developers need not worry too much.

Altman's "reassurance" seems to have had no effect, and analysts believe that the application of only ChatGPT, this "super assistant", is enough to make OpenAI's major clients "tremble with fear".

According to insiders cited by The Information, in addition to helping clients draft emails based on their personal style, developers have also been using various clients of ChatGPT to create their own artificial intelligence agent programs:

When the agent program sets a goal, such as increasing the sales of software products, ChatGPT will propose specific tasks that can be performed, such as creating an application or a promotional email template for potential customers.

Some analysts believe that earlier this month, Altman may have started recruiting related to the "super assistant", and the move to recruit talent shows OpenAI's emphasis on this work. OpenAI appointed Peter Deng, a well-known product manager at Facebook, as Vice President of Consumer Products, responsible for ChatGPT's product, design, and engineering teams.

In a post announcing his new position on LinkedIn, Deng mentioned that OpenAI hopes to develop "auxiliary capabilities that enhance people's productivity and creativity".

The commercialization of AI is still ongoing

From the source, OpenAI has always emphasized the goal of achieving AGI (general artificial intelligence) as soon as possible. The way to achieve this goal is to use a large amount of computing power to continuously train, optimize, and iterate existing large models. According to relevant reports, "AI large model training costs at least hundreds of thousands of dollars per training session", which means that OpenAI faces high computing and R&D costs in the process of achieving AGI.

MIT Research Review once reported that "OpenAI's commercial pressure is increasing, and founder Altman believes that research needs to be continued through making money."

Therefore, even though OpenAI was initially a purely non-profit organization with a beautiful vision of never allowing AI technology to be monopolized by commercial companies, in 2019, OpenAI underwent structural reforms and added a for-profit entity, OpenAI LP, which has a "profit cap" (return on investment controlled within 100 times).

According to Fortune, there are reports that OpenAI's revenue in 2022 is only about $35 million, and it still loses more than $500 million. OpenAI hopes to achieve revenues of $200 million and $1 billion in 2023 and 2024, respectively. However, for OpenAI, which was still losing money until last year, achieving these two figures is not easy. Therefore, OpenAI must accelerate the commercialization of ChatGPT, such as incorporating more and more core functions of applications into the GPT model, and becoming a super APP with API services.

Just a few days ago, OpenAI made a big move late at night. Altman said that he would create an epic LLM "application store". Connect all ChatGPT applications and completely seize the AI model ecosystem.

According to two informed sources who spoke to The Information, through the large model store, enterprises can obtain various finely-tuned AI software, such as AI that solves online retail financial fraud, and AI that has the latest data and can answer specific market information.

It is currently unclear whether OpenAI will charge commissions for these transactions or obtain revenue from the large model store in other ways.

In May, Microsoft just released the most powerful plug-in system in history. OpenAI's LLM application store can be said to directly confront Microsoft and forcibly seize the ecological market.

Wall Street News previously analyzed that OpenAI's LLM App store will have a good response, and the continuously hot user base of ChatGPT is a good guarantee, but it is not enough to have a very serious impact on Microsoft's ecology, after all, Microsoft's strength on the application side cannot be underestimated.

Before this "large model store" was rumored to be launched, OpenAI also explored other profit channels. Earlier this year, OpenAI launched GPT plug-ins, which allow users to quickly book or purchase products through AI queries of OpenTable, Shopify, or Expedia websites without visiting these websites.

However, these plug-ins did not ignite the market like ChatGPT. According to The Information, at a developer conference last month, Sam Altman directly pointed out that the GPT plug-in did not find the right market.

"Open Relationship" between Microsoft and OpenAI

The reason why the two companies that are closely cooperating on the surface have a tense confrontation behind the scenes is inseparable from the special cooperation relationship between OpenAI and Microsoft.

Since 2019, Microsoft has invested more than 10 billion U.S. dollars in OpenAI and allows the latter to use the computing power provided by Azure for free. The condition is that before Microsoft's investment can be recovered, Microsoft enjoys 75% of OpenAI's profits and obtains 49% of subsequent profits within a certain upper limit. From a paper perspective, this investment has brought Microsoft a huge return: in the AI boom of the first half of the year, Microsoft's market value skyrocketed from $1.79 trillion at the end of 2022 to nearly $700 billion, to today's $2.48 trillion.

However, in reality, OpenAI's technology does not belong to Microsoft, and Microsoft does not have direct control over OpenAI.

According to the Wall Street Journal, in order to avoid antitrust scrutiny, Microsoft only holds 49% of OpenAI's shares. OpenAI operates almost completely independently. Only a few teams within Microsoft can access the core information such as the code library and model weights of the GPT series large models. For most Microsoft employees, OpenAI is no different from other external suppliers.

This situation is very rare in the investment of large technology companies in startups. Generally speaking, after a large company acquires or invests in a startup, it will either completely integrate the two teams or completely absorb the technology of the latter. There are few companies like OpenAI that compete with their investors in this way.

Behind the cooperation between Microsoft and OpenAI, there are subtle cracks and fissures.