With artificial intelligence (AI) becoming increasingly popular, OpenAI is the toast of tech enthusiasts. So what is OpenAI exactly, and what has this company achieved to become so famous?
The Origins of OpenAI
Founded in 2015, OpenAI focuses on researching and developing AI-based tools. On its official website, OpenAI states that its goal is to “ensure that artificial intelligence benefits all of humanity.”
Sam Altman, Elon Musk, Wojciech Zaremba, Greg Brockman, John Schulman, and Ilya Sutskever founded OpenAI, which is based in San Francisco. However, Elon Musk stepped down in 2018 due to potential conflicts with his AI developments at Tesla and SpaceX.
OpenAI aims to avoid a future where tech conglomerates like Google and Microsoft monopolize AI tech. They aim to transparently improve AI tech and make them open-source—available to all.
In December 2015, Sam Altman, Elon Musk, Wojciech Zaremba, Greg Brockman, John Schulman, and Ilya Sutskever founded OpenAI, based in San Francisco. However, Elon Musk stepped down in 2018 due to potential conflicts with his Tesla and SpaceX involvements.
OpenAI is committed to researching AI advancements and sharing its knowledge with the rest of the scientific community. The AI firm has released numerous papers on AI, including breakthroughs in neural network architectures and AI training methods.
OpenAI was initially a non-profit organization but has evolved into a hybrid project with non-profit and for-profit components. The cost of creating AI products, which needed extraordinarily large amounts of data, made it difficult for the group to continue running as a non-profit.
The for-profit subsidiary began in 2019 with OpenAI Limited Partnership, as the company wanted to attract funding for its research.
OpenAI’s Main Projects and Achievements
OpenAI announced its goal of $1 billion in donations but has only received a little over $130 million so far.
Then came OpenAI LP.
Crunchbase reports that OpenAI LP has amassed over $11.3B in funding over five rounds, from its founders and contributors to a $1 billion investment from Microsoft in 2019. So it’s no surprise that this research company has covered huge milestones.
ChatGPT (Chat Generative Pre-trained Transformer), was launched on November 30, 2022. It can be used for text generation and translation, content creation, and even casual communication.
In 2020, this language model created by OpenAI passed the Turing test. A machine passes this test when it behaves in a manner indiscernible from that of a human.
ChatGPT-3 is the first generation of ChatGPT released in 2020. It was trained on a huge text and code dataset with about 175 billion parameters.
The latest generation, ChatGPT-4, has 45 billion parameters, which means it has been trained in a more extensive dataset of text and code than its predecessor, making it more accurate, creative, and fluent.
DALL-E is a generative AI model capable of generating images from text descriptions. It was created in January 2021 and has since undergone several improvements.
The initial models could only generate images with a resolution of 256×256 megapixels and were heavily criticized and shunned for their inaccurate and blurry images. However, DALL-E 2, released in April 2022 and made public in July 2022, could create images of 1024×1024 megapixels and is more accurate and realistic than its predecessor.
This AI model can generate visual content like paintings and memes for product designs, fliers, commercial and campaign posters, research, and more.
Released in 2021, Codex is based on the language model GPT-3 family—an AI model that can generate code. It can create code that is as good, or even better, than those written by human programmers.
Codex can generate code in various programming languages, including Python and Java. It can also translate over 200 languages, text, and code, and even translate between different programming languages.
Challenges and Controversies Faced by OpenAI
OpenAI was founded to advance AI research and prevent its monopolization. But the research company faces challenges on different fronts.
1. Regulatory Issues
AI has advanced faster than we expected. There are now models that can generate text and images within seconds. And there are concerns that malicious entities could use this to create misinformation, spread fake news and chaos.
For instance, there have been instances where DALL-E-generated images were used to spread fake news due to how original they looked.
OpenAI has begun creating regulatory measures trying to prevent these scenarios. But some of these measures have also prevented the use of the models, even within normal parameters.
2. Secrecy vs. Openness
OpenAI’s mission is to make AI accessible to all. Initially, they stuck to their mantra, publishing all of their research and advancements.
However, as AI advances, so do its applications and chances of being used for malicious purposes. This has led OpenAI to try to find a balance between hiding and sharing research, despite the former going against their mission.
3. Ethical Issues
There have been cases where OpenAI’s models have shown biased content due to the type of training data they were built with. Addressing this problem is complicated as it will need a lot of fine-tuning and comprehensive data selection. And even at that, it would be almost impossible to eradicate biased results.
OpenAI has numerous for-profit competitors, such as DeepMind (a subsidiary of Alphabet Inc.) and NVIDIA Corporation, which is also developing AI solutions. For OpenAI to remain at the forefront, there must be continued innovations and collaborations. This has led them to move from their non-profit declaration to a hybrid that welcomes funding.
What Does the Future Look Like for OpenAI?
OpenAI remains a huge player in the AI field. Despite challenges, many possibilities are ahead, especially with partners like Microsoft and Shutterstock.
Moreover, the AI firm has reiterated its commitment to ensuring its technology is only used for good.