The state of AI in 2019: 3 discoveries you should know about
A collection of some of the great AI breakthroughs in 2019
Over the past years, Artificial Intelligence has managed to impress everyone with numerous breakthroughs, and 2019 was no different. During the whole year we witnessed inspiring innovations in deep learning, natural language processing and computer vision. We highlight some achievements in the world of AI :
OpenAI publishes their latest AI for automatic text-generation, and launches a debate
OpenAI - a well-known research laboratory based in San Francisco, has released the full version of a text-generating AI system, named GPT-2. This discovery caused a harsh debate early 2019 as its developers originally refused to disclose their model.
GPT-2 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. If feeded with a headline, the system is capable of outputting news stories, reviews and explanations. Apart from the praise, GPT-2 faced some internal and public critics concerning the possible use of the system to help generate "synthetic propaganda" for ideological groups. OpenAI had released several models of text-generating systems in the past, and the latest model received a "credibility score" of 6.9 out of 10, making it marginally greater than the scores of the previous models. Perhaps the best way to get a feeling of the system, is by testing the GPT-2 abilities on the following web version of the system, and enter your own prompts.
GPT-2 is a living proof that the progress of Natural Language Processing and Generation is moving forward at a very fast pace.
Researchers manage to reconstruct neural networks following human brain waves in real time
Researchers from the Russian corporation Neurobotics and the Moscow Institute of Physics and Technology have found a way to reconstruct human thoughts from brain waves in real time, by applying AI neural networks. The brain-computer interface relies on artificial neural networks and electroencephalography, or EEG, a technique used to record brain waves via electrodes which are being placed on the scalp. By analyzing the brain activity, the system reconstructs the images seen by a person undergoing EEG in real time.
The current solutions for extracting observed images from brain signals either use functional MRI or analyze the signals picked up via implants directly from neurons. Both methods have fairly limited applications in clinical practise and in everyday life. The system developed by MIPT and Neurobotics can empower development of new post-stroke rehabilitation devices controlled by brain signals. You can read the details about all stages of the research here.
We can think about many good (or less good) applications for such an application. Overall, it is an impressive progress that opens the door to many future applications.
Google detects breast cancer using mammographic screening
Google's AI team DeepMind, has been working on developing artificial intelligence to help doctors in identifying breast cancer, according to their research paper published in Nature. The model, which scans X-ray images known as mammograms can reduce the number of false negatives by 9.4 percent - a hopeful leap forward for a test that currently misses 20 percent of breast cancer assessments. Radiologists omit about 20% of breast cancer from mammograms, according to the American Cancer Society, and half of the women who get cancer tested over a 10-year period have false positive results. While mammograms are the most common detection tool, they can miss a large number of cases, and there's still significant problem with false negatives and false positives - says Shravya Shetty, a researcher at Google.
The results of the study conducted by Google represent a major advancement of the potential for early detection of breast cancer.
Today, breast cancer is the second leading cause of death in women, beat only by lung cancer and in its deadliness and overall prevalence. Early detection is the best defence to treat the disease.
Kantify helps companies be successful in their AI journey. We help our clients develop their AI strategy, and we develop best in class AI solutions. If you are curious to find out how your organization can leverage data to boost growth, don't hesitate to contact us, and one of our team members will be back at you shortly !