OpenAI
1. Category : Artificial Intelligence, Non Profit, Online Games
2. Domain : Artificial Intelligence
3. Founders : Elon Musk, Greg Brockman, Sam Altman, Ilya Sutskever
4. Established : December 11, 2015
5. Number of Employees : 120
6. Operating Status : Active
7. Funding status : Not Disclosed
8. Website : www.openai.com
9. Country : San Francisco, California, US
10. Latest in News: THE AI TIMES – THE SECRET, MESSY STORY BEHIND OPENAI
OpenAI is an independent non-profit artificial intelligence research company, which aims for highly autonomous systems that outperform human’s at most economically valuable work, which benefits all of humanity and carefully promote and develop friendly AI
The organization “freely collaborate” with other institutions and researchers by making its patents and research open to the public and they will attempt to directly build safe and beneficial AGI.
Artificial Intelligence technique explored for decades, deep learning, started achieving state-of-the-art results in a wide variety of problem domains. They freely collaborate with others across many institutions and expect to work with companies to research and deploy new technologies.
The 3 success points were first, we need to compete on applications, but cooperate on open, basic research, Second, we need to create public measurement and contests and third, we need to increase coordination between industry and government on safety, security, and ethics.
The company released PPO, proximal policy optimization performs comparably or better than state-of-the-art approaches while being much simpler to implement and tune and has become the default reinforcement learning algorithm.
The organization created a bot called Dota 2, which understands tournament rules and learned the game from scratch by self-play, and does not use imitation learning or tree search.
OpenAI Charter was released in 2018, which describes the principles we use to execute on OpenAI’s mission and reflects the strategy they have refined over the past two years and also provides feedback from many people internal and external to OpenAI and will guide us in acting in the best interests of humanity throughout its development.
The company have skilled a human like robot hand to manipulate physical objects with unprecedented dexterity. The tool called Dactyl, which learns from scratch using the same general-purpose reinforcement learning algorithm and code as Five.
OpenAI In 2019 have skilled a pair of neural networks to solve the Rubik’s Cube with a human-like robot hand and neural networks are trained entirely in simulation paired with a new technique called Automatic Domain Randomization (ADR).
