OpenAI, the company which produced ChatGPT, has hired hundreds of remote contractors to teach the artificial intelligence system how to write basic code.
The language processing tool has earned worldwide recognition as knowledge workers use the system to complete tasks such as writing emails and reports in a matter of seconds. OpenAI, which recently announced another series of multibillion-dollar investments from Microsoft, has temporarily hired approximately 400 computer programmers who are creating data for models to learn basic software engineering tasks, according to a report from Semafor.
The datasets include both lines of code and human explanations for the code, according to people interviewed by the outlet, implying that the new tool will involve dialogue between the artificial intelligence and the human seeking to build or implement a computer program. OpenAI previously trained models with content pulled from GitHub, an online forum owned by Microsoft where developers troubleshoot their code and ask for advice.
Artificial intelligence systems are trained with large datasets to make decisions and produce desirable outcomes. Another 600 contractors are meanwhile creating datasets filled with images, audio clips, and other information that can be leveraged to train other artificial intelligence tools, such as autonomous vehicles. The contractors are from Latin America, Eastern Europe, and other parts of the world where low-level engineering talent is more affordable for American companies.
Some 27% of employees at prominent consulting, technology, and financial services companies have already used ChatGPT in various capacities, according to a survey from Fishbowl. One lawyer from Amazon said in an internal message to employees that “your inputs may be used as training data for a further iteration of ChatGPT, and we wouldn’t want its output to include or resemble our confidential information,” according to a report from Business Insider.
OpenAI currently offers a chatbot called Codex, which is “proficient in more than a dozen programming languages” and able to “interpret simple commands in natural language and execute them” on behalf of the user.
“Our models displayed strong performance on a dataset of human-written problems with difficulty level comparable to easy interview problems,” researchers from OpenAI said in a paper about the system published two years ago. “Model performance could be improved by training on a distribution more similar to the evaluation set, and also by producing multiple samples from a model.”
Conversations surrounding technological unemployment over the past several decades have centered around blue-collar workers losing their jobs to automated robotics solutions; the widespread adoption of ChatGPT has led some to conclude that many white-collar professions could soon be rendered obsolete. The system performed “at or near the passing threshold” for all three components of the United States Medical Licensing Exam and earned passing scores on the multiple choice section of the Bar Exam.
New York Times columnist and City University of New York economics professor Paul Krugman recently wrote that artificial intelligence “may be able to perform certain knowledge-based tasks more efficiently than humans, potentially reducing the need for some knowledge workers.” Virginia Tech economist Jadrian Wooten meanwhile predicted that artificial intelligence will create entirely new occupations and has “historically targeted routine tasks that are easy to replicate,” meaning that workers can reduce the time spent on tedious parts of their jobs.