Artificial intelligence, once considered the stuff of science fiction, has drifted into the mainstream in recent years with Siri, Alexa and other voice assistants. AI is also driving tech innovation with algorithms, autonomous robots and software applications.
In the past month, though, AI has seen a resurgence of interest among techies and non-techies alike due to generative AI technologies like DALL-E, which can transform the most off-the-wall text descriptions into an AI-generated image within seconds, and ChatGPT, an AI chatbot that can answer users’ questions with an almost eerie humanness.
Both DALL-E and ChatGPT were developed by OpenAI, a San Francisco-based AI research company founded in 2015 by former Y Combinator President Sam Altman, Tesla and SpaceX CEO Elon Musk and other investors that collectively committed $1 billion to the project.
The company’s mission is to “ensure that artificial general intelligence benefits all of humanity,” according to its charter. Artificial general intelligence, or AGI, refers to the yet-to-be-realized ability for a machine to learn and understand anything a human can. In its charter, OpenAI defines AGI as “highly autonomous systems that outperform humans at most economically valuable work.”
OpenAI originally launched as a nonprofit, but in 2019 the company created a “capped-profit” company called OpenAI LP, which now employs most of the organization’s employees. Several months after shifting to the capped-profit model, OpenAI announced that it received a $1 billion investment from Microsoft and that it would run its services on Microsoft Azure. The two companies also agreed to work together to build new Microsoft Azure AI supercomputing technologies.
Microsoft and OpenAI said in a statement that they envision AGI playing a role in solving global challenges like climate change, healthcare and education.
“The creation of AGI will be the most important technological development in human history, with the potential to shape the trajectory of humanity,” Altman, OpenAI’s CEO, said in the release. “Our mission is to ensure that AGI technology benefits all of humanity, and we’re working with Microsoft to build the supercomputing foundation on which we’ll build AGI.”
ChatGPT, an AI chatbot powered by OpenAI’s GPT-3.5 language learning model, is OpenAI’s latest product. GPT-3.5 is a slightly more advanced version of GPT-3, the company’s third iteration of a deep learning model released in 2020. Essentially, the model is informed by vast amounts of data and trained to predict the next word of a given sentence.
OpenAI launched ChatGPT on November 30, and within five days, Altman said it had reached 1 million users.
In just a short amount of time, people have giddily tested the capabilities of the platform. Users have shared stories of ChatGPT detecting bugs in code, answering technical questions and writing essays that help students skirt anti-plagiarism software.
Musk, who left the OpenAI board in 2018, tweeted that “ChatGPT is scary good,” and that, “We are not far from dangerously strong AI.”
While it may have captured the public’s imagination, Altman recently tweeted that ChatGPT was “obviously not close” to AGI, adding that “the field has a long way to go, and big ideas yet to discover.”
OpenAI is upfront about ChatGPT’s limitations. The website acknowledges that the chatbot often provides inaccurate information and that it is “often excessively verbose and overuses certain phrases, such as restating that it’s a language model trained by OpenAI.”
While GPT-3 uses its knowledge of language to generate text, OpenAI’s Image GPT model is able to generate pixels in sequence to create images for OpenAI’s text-to-image tools, DALL-E and DALL-E 2.
DALL-E was first released in 2021, but it didn’t gain mainstream popularity until an updated version called DALL-E 2 was made available for public use without a waitlist on September 28.
While DALL-E could generate images from text descriptions, DALL-E 2 can generate more realistic, accurate images with a resolution that is four times better than its predecessor, according to OpenAI.
DALL-E 2 is intelligent enough to combine objects, concepts and styles. Within seconds, it can generate 10 different images of “an astronaut riding a horse in the style of Andy Warhol,” for example, or “teddy bears shopping for groceries in ancient Egypt.” DALL-E 2 also allows the user to add and remove objects from an existing image using text prompts.
OpenAI has developed several other projects that have generated enthusiasm in the tech community but do not get the same attention as DALL-E or ChatGPT. OpenAI Codex, for example, translates natural language to code and was used to develop GitHub Copilot, which autocompletes code for computer programmers.
OpenAI also recently unveiled Whisper, an automatic speech recognition system that can transcribe spoken language with greater accuracy than existing technology, according to the company. Trained on 680,000 hours of multilingual data, Whisper can transcribe speech with accents, background noise and technical language.
OpenAI currently has more than 100 employees. Most of the open jobs on its career page are based at the company’s San Francisco headquarters, but there are also some remote positions. The company is currently hiring for 36 positions on Built In. OpenAI also offers a salaried six-month residency program for researchers and engineers who don’t currently focus on AI.