Open AI president Brockman: "We're a strange company"

"The president of Open AI speaks in Austin about his nervousness before the launch of ChatGPT, the unusual corporate structure - and criticism from Elon Musk.

When Open AI was preparing to release its ChatGPT language model last November, Greg Brockman had a message for his team: The most important thing was not to have to shut down the software again after three days. That's what the company's president and co-founder said at the South by Southwest digital festival in Austin. It was a big moment for Open AI, as for the first time the company made one of its artificial intelligence (AI) language models available to the general public. "We were definitely nervous," Brockman said.

His fears have not come true, the software has been a phenomenal success, even though it has revealed weaknesses from the start. Brockman said he still ponders why ChatGPT caused such a hype, given that the underlying technology was a year old.

 He explained it by saying that ChatGPT has been made available to the general public with "the simplest user interface we could think of".

The success has made Open AI the company of the hour and a sought-after partner in the technology industry. Accordingly, Brockman's appearance in Austin was also eagerly awaited, he spoke to a packed hall. He is not the most prominent representative of Open AI, the company is primarily identified with Sam Altman, who is CEO and also one of the co-founders. According to the organizers of the festival, Altman did not accept an invitation.

Open AI was not initially a for-profit organization

Open AI was created in 2015, and at his appearance, Brockman told of the initial founders' meeting in a luxury hotel in California's Silicon Valley, that Elon Musk was there in addition to him and Altman. The CEO of the electric car manufacturer Tesla no longer plays a role in the company. Since its inception, Open AI has been committed to developing AI that benefits all of humanity. The idea behind this is that such technologies also entail great dangers.

Originally, Open AI was a not-for-profit organization, and Brockman now called that a "good place to start," but that wasn't sustainable. 

Open AI has found that enormous financial resources are required to be able to develop powerful AI technologies, for example in the form of expensive computer capacities. 

However, as a not-for-profit organization, it has been difficult to raise money from venture capital firms through traditional channels. Therefore, Open AI 2019 had a different structure, with a new, profit-oriented company below the non-profit-oriented unit. This is how Open AI works to this day. "We have a weird looking company," Brockman said. There are critics who accuse Open AI of having thrown its former principles overboard with this structure.

In parallel with the change in structure, Open AI also formed an alliance with Microsoft. The software group invested billions in the company and was given privileged access to its AI technologies. The partnership was recently intensified with another billion-euro investment. Microsoft has used the cooperation in recent months to make some high-profile announcements. For example, the presentation of a new edition of the Bing search engine, which works with an advanced version of ChatGPT.

ChatGPT too "woke"?

Open AI itself has pointed out weaknesses in its AI technology from the start. Sam Altman has said that ChatGPT is not a commercial product yet, but a research project. Microsoft's AI variant of Bing provided incorrect or questionable answers. A journalist from The New York Times caused a stir when he recounted a conversation with Bing's chat function in which the software told him about dark fantasies and described his marriage as unhappy. Brockman said he believes it is important to uncover such cases as early as possible so that corrective action can be taken before an AI system becomes even more convincing and powerful. "For me, that's the process and the pain of learning."

Brockman was self-critical with a look at other concerns. Critics from the right political camp have complained that ChatGPT is biased or "woke". Co-founder Elon Musk also shared this criticism. A much-cited example was when ChatGPT refused to write a positive poem about former US President Donald Trump , but did so without any problems with the current incumbent Joe Biden. Musk described the case as a cause for "serious concern." There were even media reports that Musk was planning a competing product for ChatGPT. Brockman admitted Open AI didn't respond quickly enough to curb bias in its technology. "People were right in their criticism." With this in mind, Open AI has made "improvements" in the past few weeks and will soon be able to say more concrete things about it.

Little could be elicited from Brockman in Austin about future Open AI products. In the industry, the company is expected to soon release GPT-4, a much improved version of ChatGPT. Brockman simply said Open AI is working hard on the performance and security of its products and is making "significant progress".