Kimera Systems: Nigel - The Path to true Artificial General Intelligence

in #agi6 years ago

IMG_8095.JPG

Imagine a world where our electronic devices truly understand our goals in life, and better yet, are constant looking for ways to help us achieve those goals. This is the enticing premise being touted by Kimera Systems, an Artificial General Intelligence (AGI) company based in Portland, Oregon, USA.

Companies such as IBM have been successfully creating Narrow AI systems since the 1990s, with Deep Blue (Chess) and later, Watson (Answering Quiz-Show Questions) both making newsworthy advances in the field. However, the difference between Narrow AI systems and true AGI systems is that key word, “General”. Narrow AI systems are unable to apply their knowledge to other tasks. For example, Deep Blue can play chess, but is unable to use that knowledge to play any other game. A true AGI would take the ‘game theory’ it learned playing chess and be able to apply it to other tasks.

So, Narrow AI systems are limited to the specific task they are built for. True AGI systems cannot, by their very purpose, be given a specific task to learn - they need to be raised, much the same way a child is raised. A true AGI system must be able to learn a piece of knowledge in one area, and subsequently apply that knowledge to other areas, in much the same way humans can extrapolate ideas and knowledge to unconnected subjects. AGI is a machine-based intelligence that has the capacity for human-level reasoning.

For those not familiar with AGI research, the general consensus for the last few decades has been toward creating an artificial ‘neural network’, which seeks to replicate the way our brains operate. Unfortunately, this method is extremely resource intensive and has yet to provide any substantive results.

However, the team at Kimera Systems, led by Founder and CEO, Mounir Shita, use a novel ‘Cause-Effect’ approach, and have developed what they claim is an AGI, which can learn from sensor data it receives via connected electronic devices, and subsequently apply that knowledge to other unconnected situations.

This Cause-Effect learning process is currently gaining popularity in the field of AGI, with prominent AI researcher and 2011 Turing Award winner, Judea Pearl, recently releasing a new book, The Book of Why - The New Science of Cause and Effect, in support of this method. Kimera’s Cause:Effect Subnet Model is also known as a Deep Causality Network, which is the term used by Pearl.

Kimera Systems have codenamed their AGI, Nigel, in honour of AGI Researcher and Kimera Systems co-founder, Nigel Deighton, who passed away in 2013.

In an early beta test, Nigel was installed onto the phones of a group of test subjects, who were then asked to go to the movies at a cinema in Portland, Oregon, and just do what they normally do. Nigel wasn’t given any specific instructions, but simply “observed” the behaviour of the moviegoers through the sensor data (phone settings and geolocation) it received from their phones.

Within 3 days, Nigel learned that people normally mute their phones when they go to the movies and began to proactively mute the phones when it detected them subsequently entering the cinema. In a later test, Nigel applied what it learned at the Portland cinema and automatically muted the phones of people going into cinemas at various locations around the globe.

Other tests have since also shown that Nigel can take a piece of information and use human-like reasoning to decide on a course of action. These simple tests are just the first steps in creating a truly useful AGI.

To achieve a human-like AGI will require trillions of pieces of sensor data. The only way do this is to launch Nigel on a grand, decentralised scale, meaning millions of devices around the world. Only in this way will Nigel be able to learn, not only fast enough, but also broadly enough, to ensure the information received isn’t tainted or skewed by governments or powerful corporate entities looking to push their own agenda and corrupt Nigel’s knowledge.

To motivate people (end-users), software developers, device manufacturers and communications networks to get involved in the Nigel project, Kimera Systems have developed an internal economy, where participants will be paid for the sensor data they provide. Additionally, contributors will be able to protect their privacy and only provide data they’re comfortable with sharing. This will be achieved through the use of blockchain technology.

Kimera Systems are presently raising funds, with a view to launching their Nigel AGI globally in 2019. Their current pre-ICO phase has an attractive 100% bonus token offer.

Dates are not yet set for the ICO main sale, as Kimera Systems are aiming to be a Gold Standard in US-based ICOs, and they are currently awaiting Reg A+ Approval from the SEC, which will allow both accredited and non-accredited US Investors to participate in the ICO.

This coming Thursday, 12 July 2018, Mounir will be giving a presentation at the BlockchainTalks BCTvol9, held in Amsterdam. More information is available at http://bit.ly/AMSJuly122018

If this sounds like something you’d like to get involved with, then you can get more information on the project at their website, https://kimera.ai.

If you have further questions, the team has a Telegram chat on http://t.me/kimeraAGI.