A group of Silicon Valley gurus created the foundation-turned-company, with Microsoft as a major investor.
Paula Solanas Alfaro - Diari ARA
February 25th, 2023
BARCELONA "Our mission is to ensure that Artificial Intelligence benefits all of humanity": This is the messianic purpose with which OpenAI sums up why in 2015 a group of Silicon Valley figures decided to create the laboratory (initially non-profit) behind such viral tools as ChatGPT or DALL-E. As Wired magazine explained, it all started with a summer dinner.
One of the rooms of the luxurious Rosewood Hotel in Menlo Park, the same Californian city where Meta's headquarters are located, hosted a rather eclectic gathering. There was Tesla founder Elon Musk and an old friend, Sam Altman, then head of the startup accelerator Y Combinator, where companies such as the tourist rental platform Airbnb, the cloud-based document service Dropbox and the streaming application Twitch had already passed through. The South African tycoon was very interested in Artificial Intelligence. He knew that in the not too distant future it would be very relevant for some of his businesses, starting with autonomous vehicles, and the entrepreneur was one of the voices warning of the danger of machines becoming a threat overtaking the human mind.
That meeting, where a dozen experts in Artificial Intelligence were also invited, aimed to set up a laboratory dedicated to developing this technology in a "safe" way. In December 2015, other well-known names in the technology sector, such as the co-founder of PayPal, Peter Thiel, and one of the driving forces behind LinkedIn, Reid Hoffman, joined the cause and this initial group joined forces to contribute 1,000 million dollars to the initiative.
Thus was born OpenAI, then a non-profit organisation that set up its headquarters in a landmark building in San Francisco's Latin Quarter. They had managed to recruit some of the brightest minds in the field, who until then had worked for giants such as Google or Facebook, paying them less "astronomical" salaries.
"At the beginning they published many articles and the first products they launched were Gym and Universe, which were based on reinforcement learning (making a machine learn with a system of rewards so that it gets it right through trial and error)," explains Juan Ramón González, co-founder and technology director of the Barcelona-based Big Data and Artificial Intelligence company Mática Partners. The expert also points out, however, that OpenAI took a turn a couple of years later. They changed the focus to focus on a more general view of the technology.
Moreover, in 2018 Musk also announced that he was stepping down from the entity's board of directors to "eliminate a potential future conflict of interest" with Tesla's autonomous driving projects, although he remained involved as a donor. Later, the South African has launched some criticisms against the new drift of the project.
FROM FOUNDATION TO LIMITED COMPANY
At the time, OpenAI was already making waves, after explaining that it had developed a language model so good at generating fake news that they could not make it public. This algorithm was capable of translating text, answering questions and writing articles capable of misleading its readers. "2019 was an important moment because the lab became a limited profit corporation," recalls González.
OpenAI thus became OpenAI LP, a company that claimed it would limit the return to its investors by up to 100 times its initial contribution. "The original foundation still exists to control the ethical part of the algorithms, but they realised that more money was needed," says the expert. Not only to hire highly specialised staff, but also to buy the super-powered machines that are dedicated to training.
The original foundation still exists, but they realised they needed more money”, JUAN RAMÓN GONZÁLEZ (Mática Partners Cofounder)
But the real shake-up was yet to come. In December 2019, software giant Microsoft announced a $1 billion investment in OpenAI. This cash injection propelled all new product launches, such as the AI-based imager DALL-E and ChatGPT itself. "It costs a lot of money to maintain all this infrastructure and serve millions of users for free. It's a very smart move on Microsoft's part to differentiate itself from the competition and explore ways of monetisation," says González.
In January 2023, the multinational founded by Bill Gates went a step further and extended this alliance with another 10 billion dollars in funding and the integration of OpenAI into its Azure IT services platform. Just a few days earlier, Time magazine had exclusively revealed more details about what was behind the development of ChatGPT: in order to clean the tool of violent and sexually abusive content, OpenAI paid Kenyan workers less than 2 dollars an hour to filter out responses considered toxic. The articles and reports in this Diari ARA news item have been illustrated with images generated by two artificial intelligence applications: Midjourney and DALL-E, the latter from OpenAI, like ChatGPT. Both applications are still in the testing phase, but they are capable of generating unique images according to the user's requests and requirements.