OpenAI debuts commercial product for natural language processing

Artificial intelligence research lab OpenAI is taking its first step into the corporate world.

The buzzy organization, which has both a nonprofit and for-profit arm, on Thursday debuted its first commercial product: technology that helps developers create apps that understand human language.

OpenAI was founded in 2015 with investments from tech luminaries including Tesla CEO Elon Musk and LinkedIn cofounder Reid Hoffman. It originally focused on developing cutting-edge A.I. technologies “to benefit humanity as a whole, unconstrained by a need to generate financial return,” but more recently it has said it would create software to sell.

Last year, OpenAI formed a for-profit arm intended to help it raise more money, which the firm’s CEO Sam Altman, a former president of startup incubator Y Combinator, said was needed in order to pay the enormous computing bills required for big A.I. projects. Around the same time, Microsoft invested $1 billion in the organization and became its “preferred partner for commercializing new A.I. technologies.”

For now, Microsoft is not helping OpenAI sell or market the new tool, which is being tested by a legal technology company Casetext, Reddit, Middlebury College, and video-game maker Latitude, which makes the niche game AI Dungeon. But Altman said he hopes that will change and that Microsoft, with its big sales staff, will eventually partner with OpenAI on the new product.

OpenAI’s new service is based on its previous research into deep learning and natural language processing, a subset of A.I. that enables computers to respond to human language.

One such tool the organization has developed, GPT-3, can answer trivia questions and perform basic arithmetic based on queries typed by humans. The GPT-3 language model has 175 billion parameters, making it 18 times as large as the Blender chatbot from Facebook that also mimics human dialogue.

Greg Brockman, an OpenAI cofounder and chief technology officer, said that OpenAI’s GPT-3 language model is one of the core technologies powering the new commercial product. But while developers using GPT-3 must have proficiency using complex A.I. tools, those using the new product need “no machine learning expertise,” Brockman said.

Similar to other natural language processing services sold by companies like Amazon and Google, OpenAI’s service is available through an application programming interface, or API. In response to fears about the use of A.I. for nefarious purposes, Brockman said OpenAI will not sell its service to organizations to use in ways that violate OpenAI’s ethics code, which prohibits generating spam messages and harassment.

OpenAI faces many challenges in commercializing its language technology. Tech giants like Google, IBM, and Microsoft already sell language processing services to developers. Additionally, many smaller companies already sell language services that are tailored to individual industries like finance and health care, which have their own lingo, explained Forrester analyst Boris Evelson.

While A.I. researchers may be aware of OpenAI or the complex language models being developed by companies like Facebook, he said that universe is still very small. For OpenAI to stand out, its technology must be superior to what’s offered by rivals and be easy enough for non-tech-whizzes to use.

By selling enterprise products, OpenAI also risks damaging its reputation among A.I. researchers, who could frown upon the idea of “selling out” to commercial interests. OpenAI has already come under fire from A.I. critics who accuse the firm of overhyping its technology.

But Altman said that OpenAI is still pursuing its original goal of creating A.I. that benefits humanity, and that selling commercial software will be a way the organization can better understand the usefulness of its technology. Merely beating academic benchmarks or tests to prove the capabilities of its technology “was never what we cared about,” Altman said.

He mentioned a Walt Disney quote, “We don’t make movies to make money, we make money to make more movies,” as a sort of OpenAI axiom.

“I don’t think we could succeed in our mission without doing things like this along the way,” Altman said.