Imagine if electricity was patented.
If you stuck your smartphone charger into a wall plug, you’d have to pay a royalty fee to the private company that owned the patent on the passage of electrons through wires (in addition to the cost of the energy).
If you wanted to build a product that ran on electricity, you’d have to pay a massive licensing fee.
At Mind AI, we believe artificial general intelligence (AGI) will be as transformative as electricity. It is a technology so powerful, it cannot be owned or controlled by any one company or government. That’s why we’re building Mind AI, the world’s first reasoning engine that will be built collaboratively and available for everyone to use.
Artificial general intelligence refers to the ability of a machine to successfully perform any intellectual task that a human being can. Despite many impressive advances in machine learning and neural networks, AGI still only exists in science fiction. We are setting out to change that by bringing a completely new approach to AI development.
Mind AI will think and learn like humans think and learn.
To understand what sets our approach apart, you need to understand what the vast majority of modern AI projects are truly doing.
Machine learning — the technology behind mainstream AI projects — does not mimic human intelligence. Machine learning algorithms do not perform reasoning, they simply take in data and match it against a massive spreadsheet of data to calculate probabilities. Humans don’t look at a cow and calculate what percent of the image matches ten thousand other examples of ‘cow’ and compare that to what percentage of the image matches ten thousand examples of ‘horse.’ We make observations, and then generalize based on the facts we’ve accumulated throughout our lives. We see four legs, a large head, maybe an udder, and note that the animal is grazing in a field, and we conclude that it is a cow.
Correctly identifying a cow grazing in a field is an example of deduction. We refer to general principles we’ve learned over time to decipher what we’re seeing in front of us.
But humans also engage in two other types of reasoning — induction and abduction.
We use inductive reasoning when we identify patterns in our observations and expect those patterns to continue. For example, if we have only ever seen cows eating grass, we might determine that all cows only eat grass. This would be a reasonable, although inaccurate, conclusion.
Through abductive reasoning, we find the best explanation for our observations. For example, we might see a field full of round piles of poop, and abduce that cows graze in that field, even if cows are not present.
Today’s AI performs one of these types of reasoning — pattern recognition through induction — but not very well. Mind AI’s canonical data structure will combine all three types of reasoning to truly mimic human intelligence.
Like a human, Mind AI needs to be educated by building up its database of knowledge. We have developed a working product that can read and interpret texts. In the next few months, Mind AI will be reading 7th grade level text and be able to hold unscripted conversations about the material it has read.
Mind AI will be affordable to use.
Mainstream AI is prohibitively expensive to use because it relies on supercomputers that crunch massive amounts of data.
In contrast, Mind AI will be lean enough to run on a smartphone. This is because Mind will be able to connect to a database of ontologies, which are facts and rules about how the universe works and how objects relate to each other. For example, a simple ontology would be the fact that “objects that have mass are subject to the earth’s gravitational pull.” Knowing this, Mind will be able to deduce that an object with mass will fall to the ground, without having to study a pattern of thousands of objects falling to the ground.
These canonicals will act like shortcuts, taking Mind AI from point A to point B by the fastest route possible. Using a canonical is the computational equivalent of using a formula to calculate the area of a circle. Instead of filling in a circle with thousands of smaller and smaller squares and adding them up, we use the formula π r² to cut straight to the answer.
Mind AI is based on a new business model. It doesn’t rely on selling personal data, it’s not massively subsidized by tax dollars, and it’s not a supercomputer or data center for rent. The Mind reasoning engine will be available as an API for developers to access. When they create a project that relies on a reasoning engine, they can utilize Mind AI and pay for the ontologies they use incrementally.
Mind AI will promote collaboration among the world’s brightest minds.
To build a reasoning engine that accurately mimics human thinking, we need the input of all kinds of humans. That’s why Mind AI is developing an open-source development ecosystem using blockchain technology.
We’ve already mentioned the ontologies (or rules) that will form Mind’s education. To give Mind an understanding of the world that’s as unbiased and global as possible, we will crowd-source these ontologies from bright minds around the world. In exchange for submitting ontologies based on their own areas of expertise, participants can earn royalties in the form of tokens whenever a developer uses the ontologies they submitted.
This method of crowdsourcing will help fight the bias that today’s AI experiences. Mind’s understanding of any one domain will be based on multiple ontologies, giving it a nuanced understanding of every topic. For example, a foundational ontology would be the answer to the question, “What is a man?” One ontologist might submit the definition they use in their daily life: “A man is a human being with male genitals.” But another ontologist might submit an explanation that accounts for transgender identities. These ontologies will not cancel each other out, they will build on each other, creating a more complete understanding of the issue.
Mind AI will be interoperable with current and previous AI developments.
AI research has generated incredible advancements so far, especially in the areas of voice-to-text and machine vision. Our reasoning engine will pair with these and other deep learning technologies to create powerful use cases.
We like to compare it to the desktop computer. Without a CPU, a monitor and a speaker are useless, no matter how crystal clear the screen or how many decibels of sound the speaker can pump out. Mind AI will be the CPU that connects and drives tools like machine vision and voice-to-text.
Mind AI will also provide valuable understanding of context to current applications of machine learning. One example would be computer vision in a self-driving car. If an image recognition software knows that it is looking at a road, it will be more accurate in identifying objects that are on the road. Take this example: An image recognition system identifies an object as 68% boulder and 79% whale. Since whales live in the ocean, it’s improbable that a whale would be on the road, so a reasoning engine would improve the accuracy of image recognition results by understanding context.
Mind AI will make humans smarter and more efficient.
We understand the fear that AGI will make humans obsolete. Automation has made many jobs obsolete in the past, causing pain for those unable to transition their skills to stay relevant. But history has shown that on the whole, technology and automation creates more jobs than it eliminates. We believe that will continue, even with the introduction of AGI.
However, our biggest concern is that more people have a voice in determining how AGI gets used. Under capitalism, most technology gets developed to solve the problems of the people who can pay for it first. Eventually, life-altering technology may trickle down to the rest of the world. But the use cases are tailored primarily to the rich and elite.
Through our crowdsourced development ecosystem, we will empower people at all income levels to contribute to the development of Mind AI. They will be able to influence what problems get solved and have access to those solutions.
The initial applications of Mind AI will be based on the ability we outlined earlier. When we release our MVP, Mind will be able to read and digest documents and interact conversationally with human beings, without relying on decision-trees or scripts. This means Mind will be able to serve as a personal research assistant for anyone, provided the worldwide community of ontologists provides enough foundational ontologies related to that subject matter.
Mind AI is a community movement.
At Mind, we are working to ensure that the next evolutionary step for humanity will be shared by all, instead of falling into the hands of corporations and governments. We’ve developed a revolutionary ecosystem that allows contributors of intellectual property to be paid for their work, and for the public to co-own that work through a public blockchain. Our job as the founding team of the ecosystem is to spark the movement.
To stay up-to-date on our progress, sign up for our email list or follow us here on Medium.
