Article reprinted from: Kyle
Image source: Generated by Unbounded AI
OpenAI is asking Microsoft for money again.
In an interview with the Financial Times, Altman said that their relationship with Microsoft "works well" and hoped to get more financial support from Microsoft and other investors in the future to support the difficult costs of building more complex AI models.
Microsoft invested $10 billion (some say it’s now $12 billion) in OpenAI earlier this year as part of a “multi-year” deal that valued the San Francisco-based company at $29 billion at the time, a figure that has since tripled to nearly $90 billion, according to people familiar with the matter.
Asked if Microsoft will continue to invest, Altman said, “I hope so. It’s a long way from now to AGI and it requires a lot of computing power… The training of the models is very expensive.”
OpenAI is still not profitable and needs external funding
Although OpenAI announced earlier that the company had achieved a milestone of $1.3 billion in revenue, this still cannot make up for the huge expenses of OpenAI. Altman said that "revenue growth this year is good", and although he did not disclose the latest figures, they are still not profitable due to training costs. He said that the Microsoft partnership will ensure that "we can both profit from each other's success and everyone is happy."
Although OpenAI did not release the next-generation large model GPT-5 at last week's developer conference, they have not given up, and Altman said it is still under development. This new generation of large models will definitely bring more powerful capabilities, and even Altman said that it is technically difficult to accurately predict what new features and skills the model may have. This means that GPT-5 requires more training data and costs more than GPT-4. On the other hand, as major global technology giants scramble for high-end AI chips needed for large model training, OpenAI may face chip shortages or higher training costs.
In order to achieve the AGI goal, OpenAI is also recruiting a lot of talents, including recently hiring executives such as Brad Lightcap as chief operating officer, who has worked at Dropbox and startup accelerator Y Combinator. This also requires huge capital expenditures.
But one thing is certain, OpenAI still cannot do without Microsoft, its big sponsor. In order to achieve its general artificial intelligence (AGI), they need more funds. The profitability of OpenAI alone is far from enough, which will at least delay their development speed.
Altman's two main priorities: building superintelligence + enhancing computing power
Altman said he currently divides his time between two areas: studying "how to build superintelligence" and methods to enhance computing power.
Speaking about the recently launched custom version of GPT, he said OpenAI is working to build more autonomous agents that can perform tasks and actions, such as executing code, making payments, sending emails, and more.
“We’re going to make these agents more and more powerful. . . From here, the behavior of these agents is going to become more and more sophisticated. I think the business value of being able to do that in every category is pretty good.”
Is Microsoft's ban on ChatGPT a simple "misunderstanding"?
How good is the relationship between Microsoft and OpenAI? In other words, is it really good or just a show? In short, their relationship is complicated. Although OpenAI CEO Sam Altman said at the developer conference last week that they have the best partnership with Microsoft. Microsoft CEO Satya Nadella also praised the efforts of OpenAI personnel on the spot.
Maybe this is just a polite remark between them.
Shortly after last week's developer conference, Microsoft announced that it would ban employees from using ChatGPT for work.
"Due to security and data concerns, many AI tools are no longer available to employees," Microsoft said in an update on its internal website. CNBC also viewed a screenshot showing that ChatGPT was inaccessible on corporate devices.
Microsoft said: "While Microsoft does invest in OpenAI, and ChatGPT has built-in protections to prevent inappropriate use, the site is still a third-party external service. This means that you must use it with caution due to privacy and security risks. This also applies to any other external AI service, such as Midjourney or Replika."
It may be reasonable to ban other third-party external services on the grounds of privacy, which is what many large companies do, but it is a bit unreasonable to ban ChatGPT. Musk once said that as one of the conditions for investing in OpenAI, Microsoft has exclusive access to the entire OpenAI code base, which is obviously not justifiable with a "third party".
Shortly after the news of Microsoft disabling ChatGPT services came out, Microsoft once again stated that the ban has been lifted and these services can be used normally. Is Microsoft testing OpenAI? In short, the relationship between the two companies may be in a honeymoon period at present. Maybe Microsoft can continue to pay for it now, but facing the long-term high costs, is Microsoft willing to continue to pay?
References:
https://www.ft.com/content/dd9ba2f6-f509-42f0-8e97-4271c7b84ded
https://www.cnbc.com/2023/11/09/microsoft-restricts-employee-access-to-openais-chatgpt.html