Artificial Intelligence is our most important discovery…it might actually be our last ….  

What exactly is AI has been a very hot topic nowadays. Some people say simple “machines that think” whereas others incorrectly interchange AI with “big data and stats”. AI is, in fact, a broad discipline constituted of numerous disciplines, ranging from robotics to computer vision to natural language understanding etc..

The greatest objective of AI, most of us affirm, is to develop machines capable of executing tasks and cognitive functions that are otherwise only inside of the scope of human intelligence. In order to get there, machines have to be able to discover these abilities on their own as a baby gradually builds it’s intellect.

It truly is incredible how significantly progress the field of AI has accomplished more than the last 10 years, ranging from self-driving cars to speech recognition and synthesis.

Because of this progress there has been also lots of hype and exaggeration about the real potential.

Without a doubt, the well-known press reviews on AI practically everyday and technologies giants, one by 1, articulate their substantial long-term AI approaches. While numerous traders and incumbents are keen to comprehend how to capture worth in this new world, the majority are nevertheless scratching their heads to figure out what this all signifies. Meanwhile, governments are grappling with the implications of automation in society along with the dangers and the potential for new types of warfare

While it is ok to speculate, it is even greater to know what to watch for the future and where most of the value will be captured.

Right here are some examples of AI that are particularly noteworthy in their potential to disrupt the future of digital merchandise and solutions. I describe what they are, why they are crucial, how they are becoming utilized right now and contain a listing (by no signifies exhaustive) of businesses and researchers functioning on these technologies.

Reinforcement learning (RL)
RL is a paradigm for studying by trial-and-error inspired by the way people learn new skills. In a common RL setup, an agent is tasked with observing its present state in a digital environment and taking actions that maximise accrual of a prolonged-term reward it has been set. The agent receives feedback from the surroundings as an end result of every single action this kind of that it understands regardless of whether the action promoted or hindered its progress. This technique was produced common by Google DeepMind in their operate on Atari video games and Go. An illustration of RL functioning in the real world is the activity of optimising power efficiency for cooling Google information centers. Right here, an RL method attained a 40% reduction in cooling costs. An important native benefit of using RL agents in environments that can be simulated (e.g. video games) is that education data can be created in troves and at very lower price. This is in stark contrast to supervised deep studying duties that usually demand education data that is high-priced and tough to procure from the genuine world.

Applications: Numerous especially in industrial environments and robotics. Imagine a world where cars are driverless and robots are as good as a human or even better in most of the tasks currently done by people (robotic batlers for all at last..)

Principal Researchers: Pieter Abbeel (OpenAI), David Silver, Nando de Freitas, Raia Hadsell, Marc Bellemare (Google DeepMind), Carl Rasmussen (Cambridge), Wealthy Sutton (Alberta), John Shawe-Taylor (UCL) and others/

Generative models
These are AI systems that have 2 components: A creator and discriminator. The job of the creator is to cheat the discriminator. Yes, ladies and gentlemen, we are going to experience an era of fake videos beyond imagination even whole fake worlds that are not easy to discriminate. This opens the potential for entertainment industry and art into realms that we cannot even fathom currently. We can also create systems that can simulate realistic future scenarios (plan and imagine strategies) or systems that can cheat other systems by posing realistic images that are actually cheating product AI systems.

Principal Researchers: Ian Goodfellow (OpenAI), Yann LeCun and Soumith Chintala (Facebook AI Analysis), Shakir Mohamed and Aäron van den Oord (Google DeepMind), Alyosha Efros (Berkeley) and numerous others.

Memory Networks
Intelligence is as good as its capacity to imaging and to imagine we need intelligence that has memory. Current systems do not have this capacity not at least in a way that can generalize. It is a very difficult task to endow memory that can be trained in artificial neural networks, nevertheless, lots of progress has been made. Expect lots more to come.

Principal Researchers: Alex Graves, Raia Hadsell, Koray Kavukcuoglu (Google DeepMind), Jürgen Schmidhuber (IDSIA), Geoffrey Hinton (Google Brain/Toronto), James Weston, Sumit Chopra, Antoine Bordes (Honest).

Miniature AIs
In order for AI to be more pervasive and useful in our life we need chips and neural network designs that can fit into smaller and smaller devices. Currently we need big GPUs to run them and some models run in mobile phones although with compromises in accuracy.
A dilemma of developing smaller sized networks is finding out architectures with state-of-the-artwork efficiency employing a comparable amount or drastically less parameters. Positive aspects would incorporate far more effective distributed training simply because data needs to be communicated in between servers, much less bandwidth to export a new model from the cloud to an edge gadget, and enhanced feasibility in deploying to hardware with restricted memory.
Principal Researchers: Peter Warden (Google), Zoubin Ghahramani (Cambridge), Yoshua Bengio (Montreal), Josh Tenenbaum (MIT), Brendan Lake (NYU), Oriol Vinyals (Google DeepMind), Sebastian Riedel (UCL).

In Manxmachina we have experience in developing products on all the above examples especially when it comes to deployment on the edge and mobile devices or gadgets. Reach out if you wish to engage with us for a project that will bring success to your organization.

Alexandros Louizos, MD
Alexandros Louizos, MD

Alexandros Louizos, MD is a vascular surgeon that left his career in surgery in 2013 to join the artificial intelligence revolution. After working for 2 years as a data scientist he decided to leave the corporate career to start his own company in 2015. He is a 2x entrepreneur of AI-related companies, Galaxy.AI (VC funded with $2.9M), and his latest venture is bootstrapped. He has designed and executed artificial intelligence systems currently in production is Fortune 500 companies. What gives him happiness is helping other dreamers to learn data science.

Leave a Reply

Your email address will not be published.