SpaceX founder Elon Musk attends a post-launch press conference after the SpaceX Falcon 9 rocket, transporting the Crew Dragon spacecraft, took off on a screwless test flight to the Kennedy Space Center International Space Station in Cape Canaveral , Florida, on March 2, 2019.
Mike Blake | Reuters
Avocado armchairs and baby daikon radishes wearing tutus are among the peculiar images created by new software from OpenAI, an artificial intelligence laboratory supported by Elon Musk in San Francisco.
OpenAI trained the software, known as Dall-E, to generate images from short captions. He specifically used a dataset of 12 billion images and their captions, which were found on the internet.
The lab said Dall-E – a suitcase by Spanish surrealist artist Salvador Dali and Wall-E, a small animated robot from the Pixar film of the same name – learned to create images for a wide range of concepts.
OpenAI showed some of the results in a blog post published on Tuesday. “We found that [Dall-E] has a diverse set of features, including creating anthropomorphized versions of animals and objects, combining unrelated concepts in plausible ways, rendering text and applying transformations to existing images, “wrote the company.
Dall-E is built on a neural network, which is a computer system loosely inspired by the human brain that can detect patterns and recognize relationships between large amounts of data.
Although neural networks have generated images and videos before, Dall-E is unusual because it depends on text input, while others do not.
Synthetic videos and images have become more sophisticated in recent years, to the point that it is difficult for humans to distinguish between what is real and what is generated by a computer. General adversary networks (GANs), which employ two neural networks, have been used to create fake videos of politicians, for example.
OpenAI recognized that Dall-E has “the potential for broad and significant social impacts”, adding that it plans to analyze how models such as Dall-E “relate to social issues, such as economic impact on certain work processes and professions, the potential bias in the model’s results and the long-term ethical challenges implicit in this technology. “
Successor GPT-3
Dall-E comes just a few months after OpenAI announced that it had built a text generator called GPT-3 (Generative Pre-training), which is also supported by a neural network.
The language generation tool is capable of producing human-like text on demand and became relatively famous for an AI program when people realized that it could write its own poetry, news articles and short stories.
“Dall-E is a Text2Image system based on GPT-3, but trained in text plus images,” Mark Riedl, associate professor at the Georgia Tech School of Interactive Computing, told CNBC.
“Text2image is not new, but the Dall-E demonstration is notable for producing illustrations that are much more coherent than other Text2Image systems I have seen in recent years.”
OpenAI has competed with companies like DeepMind and the Facebook group AI Research to build general-purpose algorithms that can perform a wide range of tasks at the human level and beyond.
The researchers built AIs that can play complex games like chess and the Chinese board game Go, translate one human language to another and detect tumors on a mammogram. But getting an AI system to show genuine “creativity” is a major challenge in the industry.
Riedl said Dall-E’s results show that she learned how to combine concepts coherently, adding that “the ability to combine concepts coherently is considered to be a fundamental form of creativity in humans”.
“From the point of view of creativity, this is a big step forward,” added Riedl. “While there is not much agreement on what it means for an AI system to ‘understand’ something, the ability to use concepts in new ways is an important part of creativity and intelligence.”
Neil Lawrence, the former machine learning director at Amazon Cambridge, told CNBC that Dall-E looks “very impressive”.
Lawrence, who is now a professor of machine learning at Cambridge University, described it as “an inspiring demonstration of the ability of these models to store information about our world and generalize in ways that humans find very natural.”
He said, “I hope there are all kinds of applications for this type of technology, I can’t even begin to imagine. But it’s also interesting in terms of being another incredible technology that is solving problems that we didn’t even know we really had.”
‘AI status does not advance’
However, not everyone is so impressed with Dall-E.
Gary Marcus, a businessman who sold a machine learning start-up to Uber in 2016 for an undisclosed sum, told CNBC that it is interesting, but “does not improve the state of AI”.
He also pointed out that the code has not been opened and the company has not yet published an academic article on the research.
Marcus had already questioned whether some of the research published by rival laboratory DeepMind in recent years should be classified as “discoveries”.
OpenAI was created as a non-profit organization with a $ 1 billion pledge from a group of founders that included Tesla CEO Elon Musk. In February 2018, Musk left the OpenAI board, but continues to donate and advise the organization.
OpenAI was profit-making in 2019 and raised an additional $ 1 billion from Microsoft to fund its research. GPT-3 is set to be OpenAI’s first commercial product and Reddit has signed up as one of the first customers.