infoShare '18 flashback: Breakthroughs in AI. Pawel Noga's speech at Inspire Stage
Every year we’re getting better and better at developing artificial intelligence and applying it in practice to make our lives easier, safer and more comfortable. Already, there’s plenty of projects under way with some stunning and jaw-dropping results. In this 30 minute speech, Paweł Noga talks about the biggest breakthroughs in the field of AI within the last years.
Although many of us fear that AI might eventually pose a threat to humanity, things don’t look like this at the moment. On the contrary, it’s the other way around. In his speech Noga focuses on deep learning, the fastest growing part of AI. When applied on neural networks and big data, deep learning allows the systems to learn extremely quickly, achieving a much higher efficiency than humans.
So what have been the major breakthroughs in the last years? Let’s have a look.
- Magenta. This Google project is exploring the role of machine learning in the process of creating art and music. What can it do? Give it the music of a certain composer, say Chopin for example, in return Magenta will create for you new music that sounds just like if it was composed by Chopin.
- A Neural Parametric Singing Synthesizer. This product can sing. Yes, in male or female voices and a variety of languages. Give it notes, adjust parameters and the computer will sing a new song for you.
- Baidu. An app made in China that is a real-time translator. Talk in Chinese and hear the app translate you instantly to English. According to the creators it has a better accuracy than professional translators. Offering free translations, at the moment it translates Chinese dialects as well as English.
- Alpha Go Zero. A Google project that shows how deep learning systems can learn to play any game and achieve basically champion level within a few hours or, depending on difficulty of the game, even minutes..
- Data Center Power Optimizations. Another Google project that uses deep learning and big data to create data center designs that highly outperform projects done by humans.
- High Value Patent Searcher. This is the work of IBM. The goal of the project is to teach the neural networks to scan patents and find the best ones as well as to find those that when joined together the combined patents will generate much more potential.
- Solved Scalability Problems. This is an area where Facebook is active. Last year they presented a new way of scaling networks with the use of AI.
- Nvidia. Noise Filtering. This project is about the use of neural networks in creating high-quality graphics.
- Seeing AI. A Microsoft project that can change lives of the visually impaired. Seeing AI is an intelligent camera app that provides the user with information about who and what is around you. Just hold up your phone and hear the description. The app can recognize products, currency, people and even... people’s emotions.
- Self-Driving Cars. Developed by automobile companies, self-driving cars are the cars of the future. As Noga says, Uber is especially interested and highly engaged in this project - drivers are for them a high cost that could potentially be fully substituted by AI.
- Medical Diagnosis Assistants. In these projects, machines are learning to present the right diagnosis. The accuracy of machine-generated diagnosis for early stages of cancer is at the moment 80%. Human doctors score 60%.
- Urban Planning. Although we don’t have much success with artificial city planning, most probably AI can outperform us. Such experiments are being carried out in China, where machines decide about the best locations for streets, schools, parks, hospitals or police stations.
We might be a little scared but Artificial Intelligence is here to stay and there’s no way back. As Noga says, statistics will eventually convince everyone to AI-solutions. How?
For example, if self-driven cars are much safer, they will become favored by the insurers and slowly become the cheapest driving option. Another big advantage is AI’s adaptability. Deep learning means systems can adapt and learn quickly on their own so there’s no need for humans to foresee all possible scenarios and program everything.
What else will happen? Our lives will change forever. Just as the previous ‘revolutions’ of the past AI will for sure render many jobs obsolete. Translators, doctors, even programmers are at risk, not to mention those doing various manual or repetitive jobs that require little skill.
But then maybe... on the other hand... when machines take over so many tasks we can finally have more time to just relax and enjoy our lives? Isn’t that a vision worth looking forward to?
Pawel Noga, an Intel Employee, is a manager of benchmarks and tests team in Artificial Intelligence Products Group.
HAVE ANY IDEA FOR CONTENT?
Contact the editorial team at:firstname.lastname@example.org