Solve problems. Serve joy. Save money.
How does it work?
How does it work?During the World Economic Forum, the CEO's of the biggest IT companies agreed that we have to address two major issues in order to create a truly beneficial and safe AI; a 'Black Box' problem and a 'Safe ecosystem'.
Hussar was designed with these two principles in mind. Spruce Mobile for some time a game development company, using games to conduct a research about human cognition and perception. Over one hundred games later, combined knowledge about human behaviour, biochemistry, neuroscience and algorithmic, lead to the attempt to reverse engineer human cognition.
There are three rules we've followed.
1. Safety first. The enviroment for AI have to be isolated. That's why Hussar is embeded into a software that can simulate our world but exists as a game scenario. This is his world. His rules are significanlty different from our world. So the AI cannot operate outside, but still can develop similar inteligence with which we can interact.
2. Hussar has to be a 'White Box AI'. That means, we always have to have an insight into his thoughts, decisions chains and predictions. It's very helpful during development, so we can compare its understanding process with our own. But also we can have full insight on skills and procedures it learns. We can monitor all the possible scenarios during and before execution.
3. Nature based creation. The most significant principle for an AI architect was to mimick nature. Therefore Hussar is very simple in structure and very efficient in processing. We colud also reproduce AI from blueprints outside of computer.
The most significant benefit of this philosophy is a plan to reproduce Hussar on a nanoscale and design a device with a Superhuman level AI but the size of a penny coin.
You might not agree with our way of solving this problem. Or may even consider impossible. But architecting AI is also looking beneath our current understanding and having a vision, that exceeds present mindsets by decades.
Spruce Mobile is a company developing games for quite some time. We were using games to generate income and conduct researches in various fields, like biochemistry (dopamine and addiction), cloud computing or AI.
Now, 100 games and 20 game-engines later, it's time to move forward.
The plan is simple. We are going to bootstrap while preparing product line, which is slowly emerging. I'm not looking for a financial investor but rather a bussiness savy mentor and advisor helping me to disrupt markets.
Marcin Rybicki, direcor[email protected]
Spruce Mobile LTD
590 Kingston Road
SW20 8DN London
Don't lock yourself away with a narrow mindset!(click to watch the video)
We've decided to ship to our customers a Long Term Memory core as an independend product. We called it Marie, to honor Marie Curie, Polish scientist.
Marie augments human knowledge, in a similar fashion to human memory. It can learn all kind of data and provide a simple, intuitive understanding. Augments untrained workers, providing professional knowledge, greater than the brightest minds have.
Imagine a paramedic that can successfully perform a doctors job. Or a teenager helping his community improve health in remote places of the Earth. Your private doctor, lawyer, advisor, analisyst, mentor in your pocket. It opens a new level of holistic understanding. What would you do if you had answers for everything? The solution is here already.
How it works?
We are feeding Marie memory with documents and data. She processes that just like humans do, to extract most relevant. And then creates connections, similar to spider webs or human brains. Everything is available through an easy to use interface. Helping you to understand holistic problems, synthetise and visualise complex data and relations. Marie can augment any person with any wisdom and experience he needs in order to influence the world.
What's the value for your company?
Knowledge without an education. An access to the knowledge for everybody - not just a single educated employee. Imagine hundreds of thousands spent on training that couldn't be utilised properly.
Proffessionals supported by AI.
A simple user interface allows you to offer professional services without spending time and money on updating your knowledge every day. This gives cutting edge competence each day and allows for a truly new era of data analysis.
Supports holistic research understanding.
A knowlegde structure, unachivable for a single human brain, helps you analyse and synthesize better solutions. This could give you an insight to all wisdom available on Earth, visualised in a user friendly form.
Every human language.
Now you could outsource your analisys or consulting work everywhere you want. Marie works offline (daily or weekly updates are available), so you could expand your potential beyond IT infrastructure.
We are expecting Marie to be available on the market in September 2017
To be announced in August 2017
We consider Machine Learning to be too slow, too expensive and doesn't offer proper level of security and supervision.
"White Box" AI
In basic words, you can't see what's inside a Machine Learning 'Black Box' model which is bloated with data. All you have - as a client - is the developers self-claimed accuracy. It's dangerous and can be disastrous. On the other hand, our Ai is transparent. Imagine a scenario where you get an automatic message from two different AI's:
Machine Learning (Black box AI)
"Hello Sir, this is an automatic message from your doctor. Our AI has predicted, that you have a 21% chance to develop a lung cancer in the next 7 years. We don't know why. We also don't know if AI was propertly trained, because it's too expensive to re-train. We also can't tell you why you are in the 21% risk group. Your insurance won't cover an expensive double-check. We recomend a series of costly diagnostics every 3 months."
Our solution (White box AI)
"Hello Sir, this is an automatic message from your doctor. Our AI has predicted, that you have a 21% chance to develop a lung cancer in the next 7 years. Our data shows that it's due to three major factors which are listed below along with our suggestions. If you follow our guidelines, the risk can drop to 0.5% and you can also avoid series of costly diagnostics. Please visit your profile to see your full health report."
Imagine a self driving car AI, which you could ask:
"Show all possible scenarios of an accident with humans and highlight AI decisions please."
That kind of AI is safe for humans and could easily pass even most restrictive law regulations.
Unlike Machine Learning), it's save, predictive and can be tuned to meet any safety requirements.
Our AI could appear less or more inteligent in order to give humans the best possible experience. We can also provide several AI's with individual characteristics. Or one combined Super AI.
Our clients gets a single, lightweight file which also works well on mobile devices.
Our AI does not require an internet connection. Simply upload the file before the game starts and when one of the players dissconects or the servers crash, seamlesly switch to AI and the other players won't notice. By lowering restrictions on server infrastructure, you could save up to 90% of your costs.
We want to think about our approach as a first truly General AI. The one that learns anything. On a single laptop or milion of independent machines. It also works as a milion individuals or can be merged as a super-human General AI.
But we're also aware of the challenges it brings to our world. That's why we've designed a closed game-like ecosystem through which we interact. Our algotithm is a reward driven AI, that can reach certain goals independently, but is also safe to interact. On top of our platform we've included social handbrakes. It means that AI is eager to pursue a reward - like humans - but only if it pass broader social acceptance. It can be taught to be a good silicon being.