On Monday, a group of tech moguls formed by Elon Musk, Sam Altman, Greg Brockman and Ilya Sutskever unveiled a very ambitious project called Open AI. Described as a non-profit organization that has two main ideals by which it will exist. One is to build a “safe” AI, so we can be sure this is not going to end with a giant buffed robot with an Austrian accent chasing a teenage kid around Los Angeles. Secondly, and more importantly, the members of the organization claim they want to ensure AI’s benefits are widely and evenly distributed. The latter is imperative because there is a growing problem in our modern society that no one really wants to address, and it is called wealth gap.
First things first, the AI. It is important to understand that the organization is not planning to build robots, it is a simple and common misconception. The web site clarifies that they will take “a robot from the shelf” and then, develop an AI that fits the hardware and makes it able to perform different chores. In other words, they literally want to take a lifeless robot and teach it things through algorithms like Pinocchio, but without the magic, only pure science.
Creating a digital personality cannot be a simple task
So far, the organization has two goals in its sight. One, is to measure their progress. Science was built on trial and error, and recording every result is of vital importance. It tells the researchers what works, how often it works and when it works, among many other things. To carry on with this record, the company is going to use a home-made feature called “Open AI GYM” which is very similar to the ones that get crowded in January, but then are deserted by February. The idea is to set an algorithm in a robot that allows the machine to sense the environment and interact with it. The guys behind the company will use games and other robotics- based tasks to measure the progress.
Our technical goals: https://t.co/XUpnd2QMkM
— OpenAI (@OpenAI) June 20, 2016
The announcement explains that most of the budget is going to be diverted into research. The theoretical work is very important in the area because little of what is already understood works. That is why the organization set different things that are going to study various problems not to only solve them, but to create an algorithm that can solve it by itself.
“A significant fraction of our research bandwidth is being spent on fundamental research. We’ll always be developing and testing new ideas, especially those that don’t fit neatly into our current worldview. This is important — our current ideas will not be enough to achieve our long-term goal,” reads Open AI’s official web page.
Do you remember ‘Rosie’ from the Jetsons?
The company first will attempt to teach a robot how to do house chores. There are already a couple of little droids that sweep the house or so, but the founders of the company want to create a code that makes a robot able to do basic stuff around the house. However, they didn’t say what kind of chores they have in mind, it could be anything.
One of the most exciting things Open AI brings to the table is their “natural language understanding” project. Again, they are not pioneering in these sectors as there are many chatbots and other similar things on the net. But based on the principles of the organization, it would be safe to assume that they don’t want a robot that interacts with people. They want to make an artificial intelligence capable of intelligent and autonomous conversations which is mind bending. The implications are huge because they are basically attempting to create a digital personality from scratch. If they managed to do it, they could upload all the languages in the world and create a real-life version of C-3PO.
All in all, the main idea is to develop a “AI gym” with different games that test one thing in particular (picking up things, avoiding obstacles, categorizing objects, speaking, solving math problems, answering questions and so much more), and create a group that is going to focus on solving one specific problem at the time. Since all the groups are somehow interconnected, if one of them discovers something, all of them can benefit from it. Lastly, when they think it’s the time, patch all the projects into one unified effort which will hopefully result into a new AI, something never seen by human eyes.
The organization expressed its desire to involve the community, both updating the improvements, new projects, and information, and also extending job offers to anyone capable of contributing. Yes, the company made available a link where people that have the knowledge, talent and motivation can join them in their quest for a new kind of artificial intelligence.
The company also opened a chat for people to discuss topics that could inspire the researchers into working new ideas, and it is all available starting yesterday, June 20, 2016.
The members commitment to ensure that the fruit of their work is going to be distributed in an even and fair way is a critical point to address. Specialists report that while technology has given humanity the means to dominate the world, its cost and viability are designed so a few people with a lot of money can benefit from it. This fact has further widened the gap between the people that have with the people that don’t have creating a barrier and potentially killing the American dream, but that is a topic for another article.
— VentureBeat (@VentureBeat) June 20, 2016
Source: Open AI