Since the growth of digital data, the progress of AI has been peaking every year and it’s tending to grow every year. Even this year, there has been numerous breakthroughs in Artificial Intelligence. AI is reaching organizations in many different ways compared to last year, and it’s growing every year. There have been many researches and papers for the development of Artificial Intelligence. And Tech companies and researchers
from across the world have been experimenting in various fields with Artificial Intelligence, and like previous year these are some of land marking results in the year 2019.
Open AI solving Rubik’s Cube with a Robot Hand
Open AI’s system called Dactyl in 2018, was trained entirely in simulation and transferred its knowledge to reality adapting to real world physics. Dactyl learned from scratch using the same general purpose reinforcement learning algorithm and code as OpenAI Five (Dota2).
However in 2019, same Dactyl was used to solve Rubik’s cube. Open AI was able to rain a pair of neural networks to solve the Rubik’s Cube with a human like robot hand. The system can handle situations it never saw during training, such as being prodded by a stuffed giraffe. It still hasn’t perfected its technique though, as it solves the Rubik’s Cube 60% of the time (and only 20% of the time for a maximally difficult scramble).
Rise of DeepFakes
DeepFakes are already used to put pictures of other people’s face on the bodies of porn stars and put words in the mouths of politicans. DeepFakes started in 2017 putting celebrity faces into pornographic movies, however this year DeepFakes have disrupted politics, mocks the most popular TV show in the world, and inspired action by US senators and the Pentagon. In 2019, fake videos of Speaker of the House Nancy Pelosi, Facebook CEO Mark Zuckerberg, and “Game of Thrones” character Jon Snow went viral and inspired responses from some of the most powerful people in the world.
DeepMind’s AI masters the StarCraft II
Back in January, DeepMind announced that its AlphaStar system was able to beat top pro players 10 matches in a row during a prerecorded session, but it lost to pro player Grzegorz “MaNa” Komincz in a final match streamed live online. The company kept improving the system between January and June, when it said it would start accepting invited to play the best human players from around the world. Its actions per minute were also controlled, with the AI only being able to complete 22 actions every five seconds, something that pro player Wünsch said was comparable to the human professional level.
And, after 44 days training the neural network, DeepMind was not only successful in trouncing its human opponents under similar human-level constraints, but it also managed to climb its way into an elite group consisting of the top 200 players of the game, giving it coveted Grand Master status and securing its place in the top 0.2 per cent of players.
Open AI hide and seek
Hide and seek is a simple game, which requires a a lot less methods than other games. Researches at Open AI taught a gaggle of intelligence artificial bots to play hide and seek. The goal was to observe how competition between hiders and seekers would drive the bots to find and use tools (objects in an environment). The idea is familiar to real life game, the hiders would do their best to hide and seekers would do their best to find. The hider agents not only found a best to spot to hide but they used a tools or objects to hide themselves from seekers in the environment. Whereas on the other hand seekers also learned to use the object to find them.
Samsung’s AI brings photos to life
Samsung impressive AI is able to bring Mona Lisa to life. They have demonstrated that it can take a single image of a person and turn it into a talking head. It does this by front loading the facial landmark identification process with a huge amount of data, making the model highly efficient at finding the parts of the target face that correspond to the source. The more data it has, the better, but it can do it with one image — called single-shot learning — and get away with it. That’s what makes it possible to take a picture of Einstein or Marilyn Monroe, or even the Mona Lisa, and make it move and speak like a real person.
Open AI Text Generating AI
This February Open AI released a full version of a text generating AI system aka GPT-2 which is a part of a new breed of text-generation systems that have impressed experts with their ability to generate coherent text from minimal prompts. The system was trained on eight million text documents scraped from the web and responds to text snippets supplied by users. Feed it a fake headline, for example, and it will write a news story: give it the first line of a poem and it’ll supply a whole verse. However, the full version of the program was withheld out of fear it would be used to spread fake news, spam, and disinformation.
DeepMind gamifies memory for reinforcement learning
DeepMind changes the game how the reinforcement learning is trained. We humans learn about various events and make decisions based on past experiences. When humans do something they realize if it was a right or wrong decision based on our past experiences. For one, if we place a object too close to the edge of the table, we realize that it might accidentally hit the floor a moment later. Predicting such instance even before the disasters occur is to make human superiors to machines.
So DeepMind created something in software that is like what people do when they discover the long-term consequences of their choices, the DeepMind’s software is a deep learning program called “Transport of temporal value” (TVT). It is a way to send lessons from the future, if you wish, to the past, to inform actions. In a way, these are “gamifying” actions and consequences, which show that there may be a way to make actions at a time obey the probability of further developments to score points. They are not creating memory, per se, and they are not recreating what happens in the mind. Rather, as they put it, “they offer a mechanistic description of behaviors that can inspire models in neuroscience, psychology and behavioral economics.”
The “Reconstructive Memory Agent” uses multiple objectives to “learn” to store and retrieve a record of past states as a kind of memory.
Solving Three-Body Problem with AI
Solving Three-Body problem is a mind blending calculations that requires to predict how three heavenly bodies orbit each other which have baffled physicists since the time of Sir Issac Newton. Now Artificial Intelligence has shown that it can solve the problem in a fraction of the time required by previous approaches.
The baffle calculations involves calculating the movement of three gravitationally interacting bodies- such as the Earth, the Moon, and the Sun, for example — given their initial positions and velocities. the neural network produced by researchers from the University of Edinburgh and the University of Cambridge in the UK, the University of Aveiro in Portugal, and Leiden University in the Netherlands.
The team developed a deep artificial neural network (ANN), trained on a database of existing three-body problems, plus a selection of solutions that have already been painstakingly worked out. The ANN was shown
to have a lot of promise for reaching accurate answers much more quickly than we can today.
The algorithm they built provided accurate solutions up to 100 million times faster than the most advanced software program, known as Brutus. That could prove invaluable to astronomers trying to understand things
like the behavior of star clusters and the broader evolution of the universe.
Upside Down reinforcement learning
A team from Swiss AI Lab introduced a new methodology called it an upside-down reinforcement learning. They succesfully carried out reinforcement learning in the form of supervised learning. This allowed the team to
provide rewards as input, which is contrary to how the traditional reinforcement learning work. Such a technique also enabled them to imitate and train the robot for carrying out the strenous tasks just by imitating in front of a machine. The command based methodology assists ML models in expediting the training process, thereby decreasing the time required in AI workflows.
AI is making great strides, but understanding the methodologies inside the black box is crucial for bringing thrust. Over a couple of years there been huge concern that even an AI creator doesn’t understand what’s running inside a neural network, or how it is truly working. Therefore companies released services to allow businesses to underline the prime factors that lead to outcomes from their machine learning models. For the first time, firms are able to clear the cloud and gaining insights into the way the black box works. Although one cannot yet obtain all the aspects of a conclusion form models, it has a more prominent role to play in democratizing AI.
Thanks for Reading