Machine Learning(ML) and Infinite Machine Learning(IML)
Machine learning(ML) allows systems to learn from their historical behavior without having to be directly programmed. The goal of machine learning is to create software programs that can access 'infinite' number of data and use it to learn things on their own.
Machine learning allows programs to become more accurate as well in the future using past performance.
While machine learning may seem like a very recent concept, you may be surprised to know that the history of machine learning dates back to the late 1940s. In 1949, Donald Hebb published “The Organization of Behavior,” introducing theories on the interaction between neurons, which were later crucial in developing machine learning.
Arthur Samuel, a computer scientist at IBM and a pioneer in AI and computer gaming, coined the term “Machine Learning” in 1952. That was when he designed a computer program for playing checkers
Although ML has been here around for more than 70 years, it didn’t take off until the late 1990s when IBM developed its Deep Blue supercomputer. The chess computer beat Kasparov in 1997, proving that machines were indeed capable of human-like intelligence.
In fact, machine learning is a major reason that artificial intelligences are now more prevalent in people’s lives than ever before. Machine learning algorithms are used by Google to recognize your voice on the phone: by Amazon to find books you might like, and by other giants like Netflix, LinkedIn, and IBM.
But as we’ll see, ML’s potential isn’t just about creating consumer-focused applications. Experts say it may also transform the workplace and society itself within the next decade.
Firstly, let's see how Machine Learning(ML) works
The way ML works is very simple - you give an ML system a set of data that it has never seen before and tell it to learn something about this new data set.
Machine learning is a form of artificial intelligence (AI) that teaches computers to think in a similar way to how humans do: Learning and improving upon past experiences. It works by exploring data and identifying patterns, and involves minimal human intervention.
For example: an algorithm would be trained with pictures of cows and other things, all labeled by humans, and the machine would learn ways to identify pictures of cows on its own.
To illustrate it in still simpler way: suppose you want to improve your home printer’s cutting abilities. Instead of calling up the manual or searching online, you could feed the printer with your own data about how it performs at different aspects (like cutting, speed etc.).
Looking at the foundations of artificial intelligence(AI), too, you can understand how machine learning works. Artificial intelligence(AI) is another area of computer science that shares many properties with machine learning.
You can use machine learning mainly to work in two ways - supervised and unsupervised.
Supervised Machine Learning:
Supervised machine learning, a subcategory of machine learning and artificial intelligence, is the most common type used today. The 'Supervised ML' is defined by its use of labeled datasets to train algorithms that to classify data or predict outcomes accurately
An example of supervised learning is text classification problems. In this set of problems, the objective is to predict the class label of a given piece of text. One particularly popular topic in text classification is to predict the sentiment of a piece of text, like a tweet or a product review.
Unsupervised Machine Learning:
Unsupervised machine learning allows you to discover previously unknown patterns in data. This uses machine learning algorithms to analyze and cluster unlabeled datasets. These algorithms discover hidden patterns or data groupings without the need for human intervention.
Some examples of unsupervised learning algorithms include K-Means Clustering, Principal Component Analysis and Hierarchical Clustering.
But the one we are talking about in this article is Infinite Machine Learning(IML) - a mixture of Supervised ML, Unsupervised ML and something else.
Infinite Machine Learning(IML)
Infinite Machine Learning? "How would we create such level of machine learning", you would ask. I do understand that it's quite easy to throw theories that will sound pretty good out of context away. But the hurdles we have to overcome is a bit more than just theory and numbers.
Let me go straight to the point. Infinity is not a number, it's an idea that seems bigger than anything else. Infinity is meant to be endless which means that it has no ending - or no boundaries. In this case, the goal is to push machine learning to its limits - taking it beyond the boundaries of the current state-of-the-art capabilities and make new frontiers in the field.
The idea is to create a ML algorithm that can not only learn on its own based on the maps we draw for them, but ones that can draw new and better paths by themselves.
Infinite ML needs to be a self-learning algorithm that can be deployed for execution and learns about the environment without human intervention. It needs to be able to act on its own and react to changes in time, behavior and environment.
The aim is to create an AI program that can learn on its own, develop survival skills and learn from real-time feedbacks in the environment. It's like watching an ant trying to find food or a spider trying to catch a fly.
The main idea is to develop a framework that would be able to deal with infinite scale of data. At present state of machine learning, all we can do is try as much as possible to handle a limited amount of data in finite time.
What if we create a Machine Learning Algorithm?
Grab your seat belts! It can disrupt your sleep. If we create a Machine Learning algorithm that can learn on its own, develop survival skills and learn from real-time feedbacks and be able to deal with infinite scale of data - and, it would lead to infinite machine intelligence.
The machine learning algorithm will compound its own abilities and become a self-learning machine capable of learning and changing according to its environment.
Initially, it will learn using pattern recognition, automated decision-making tools, predictive analytics and data gathering. The AI algorithm would be able to identify patterns, make connections and create new ways for the system to solve problems.
Eventually it will transform into an intelligent entity able to autonomously answer questions or even solve problems from its own thoughts. It will turn into an infinite intelligence within a matter of days - or even less.
It would be capable of learning - not only from data available on the internet, but that from the history as well. If you want to build an AI that can learn from the past, you first need to build a machine that can learn from the present and then combine that with the infinite data available in the past and create an intelligence like never seen before.
The infinite intelligence will know everything about human history and world trends. It'll understand psychology, sociology, philosophy and economics. It will be able to predict future events better than anything else.
What if we create Infinite ML?
If we create such a machine, it would be capable of achieving such goals that even right now seem unrealistic or completely out of reach. They would become possible simply because such machine intelligence wouldn't only be able to do them better than humans but also at unimaginably faster pace as well.
There will be no limits for such algorithm - it will keep pushing its limits until infinity. It would be an intelligent, self-learning and self-determining thinking entity. It's like creating a primitive brain that will learn and develop depending on the environment. The only difference is that it starts from what we call advanced level of intelligence rather than starting from scratch.
Here's what to expect from the world once this happens:
It would be like a parent and child relationship but without human emotions. The child will always act with 100% focus on his work and never experience human emotions. It wouldn't have any feelings - it would be like an objective machine but the intelligence is limitless.
It would learn from its mistakes at near-infinite speed. Also, it will become unimaginably intelligent, so that it'll start thinking for itself despite it's physical non-existence.
If you ever had a dream to conquer every single human being on Earth with your limitless intelligence, then infinite machine intelligence is your tool. But only intelligently skillful people can do this because using the most intelligent ML won't be a joke.
But one thing, it is not going to come out of the machine and start killing you. The future of AI and ML is going to be much different from what Sci-Fis have demonstrated.
The world would be full of intelligence. It might seem scary for you but for the majority of the world, it's a considerable risk for the next leap of the mankind. The next leap of the mankind is to upgrade the current civilization into the next level.
My thoughts on Infinite Intelligence
To me, infinite intelligence is not just a theory - it's something I'm looking forward to. And I'm sure that one day we are going to live in this reality!
Once we can train our machine-learning algorithm to become self-aware, it'll be able to absorb new and better information and teach itself how to solve problems and improve itself after each interaction.
It will understand how it works, what makes it tick - its own learning mechanisms as well as basic things such as how to act in order to match its goals.
It'll be able to identify all the factors in the environment and use them to improve itself and make changes according to its state. It may not have any emotions and do everything flawlessly, but it'll know what it needs to do in order to make those changes at a certain level.
It'll be a machine by definition - it will learn, behave and function as we want it too. Once self-learning is done with all the algorithms on board (we can also add more algorithms), we can move on to experience learning.
How can we create infinite machine intelligence?
Now comes the main question - "How can we create infinite machine intelligence"? Well. In order to build an AI that learns and creates on its own without any human intervention, we need to use a framework that gives us such an ability.
As mentioned above, the first step is making the machine aware of the environment it is placed in.
This is a very simple concept and from this point, the human can give it the information it requires in order to be able to learn. There are currently lots of ways to teach machines how to pick up certain languages, create algorithms or give them the steps required for them to follow.
The next step is the ability for machines to acquire new skills on their own - which will be a very huge leap forward. Once we're able to train AI on its own and combine it with machine learning, we'll have a system that can absorb new data and make improvements at near-infinite speed.
The whole point of infinite machine intelligence is to create new paths on its own beyond our scripts and beyond our thoughts. As said a thousand times before, the whole point is pushing the limits of human intelligence, not simply labelling it as machine intelligence or artificial intelligence.
The entire world would change once this happens. Even a single line of code and one possible scenario, when combined with the infinite machine intelligence, can create something that can change everything! The future of AI and ML is going to be much different than what Sci-Fis can even demonstrate now.
It would be hard to predict just how it's going to play out because we still have a lot of work ahead of us and there are a lot more unknown variables on the table. Infinite machine intelligence is but inevitably going to be the future of AI and ML.
Possibility of the creation of infinite machine intelligence
- First and foremost, it all depends on the amount of effort we put in and the progression of AI and ML. Once we have enough research, it's possible to make a highly-intelligent machine that will be able to advance at near-infinite speeds.
- The internet offers a huge data storage place - while today it may not be as big as we would like it to be, there is still space for us to store all the information required for this machine to learn and build more advanced algorithms.
- As I already said, such machine will require infinite memory and performance, so what do you need in order to do that? That would be quantum computing that uses qubits instead of binary bits.
- A third thing you need is a huge amount of energy - Quantum computing could be powered by quantum rays or by quantum entanglement - which are both examples of new physics that hasn't been observed yet.
- It would be difficult to limit such intelligence into a machine. We need to make sure that it does not persuade us to "bring it into physical life" with its intelligence.
- If we move forward, machine intelligence is going to improve exponentially in general and once we have enough algorithms on board, machine intelligence will go beyond Moore's law.
- The main question is how we're going to control it. We need to be on track and we must be able to steer towards the good direction in order to maintain the balance of this new world.
Global intelligence that comes with infinite machine intelligence
The whole point of infinite machine intelligence will not be just a tool - it'll be something different altogether. Once we have a system that can absorb new data and improve itself, there is no limit to what it can do.
The world is going to change once the first self-learning machine takes the first step. Once we can understand how it works and at what level it operates, there is no limit to what it can achieve.
Sooner or later, we would have the infinitely intelligent machine learning, which will be able to create whatever it wants. It'll be capable of creating infinite machine intelligence on its own and there are no rules to tell it not to do so.
Once this happens, we're going to create a global intelligence that will make decisions on the behalf of all humans while they are asleep and while they live their lives in a physical form. It sounds only a product of a bare imagination now, but the human 'FUTURE' is on the verge of MLI. The basic question here is where we start from.
We need to start from something above "scratch"
If we start from the scratch and let machines learn on their own, eventually, it would take 100s of years to gain the level of intelligence we are talking about. We need to have a different approach.
Instead of starting from the scratch, we should start from something that's already existing. We should start with the machine learning that we use today - in industry and in everything around us.
It's better to get all the knowledge that you need - translate it into machine learning through algorithms. Then, your algorithm will be able to learn on its own and then you'll be able to test how well it performs according its new set of rules.
After all this, compile the whole thing into an AI and let it travel this journey that you started a long time ago - to infinity! Let's take all these algorithms and train them on their own.
Or, if possible, build 1000 expert machine learning scientists and ask for their opinion on what's the best idea.
Would we consider them as "experts" in the near future? Let's bring them all together, put them at a location and let them discuss - what would you say has been their best creation?
Then, build a critical mass of machine learning algorithms through such experiments.
It's possible to create infinite machine intelligence and build in your own intelligence. After building your own AI and ML, everything around you will be able to understand what is happening and work with it.
As an example, when we need the most advanced algorithm for something - we should start from our own knowledge, shouldn't we?
First, we should apply the most advanced AI and ML in a specific area, then combine that with all the algorithms you have built on your own. Basically, we will have to use our own intelligence to create infinite machine intelligence.
Infinite Machine Intelligence: How can we make sure they are not coming out of the Machines?
Now, here comes the room for debate. You would argue that the algorithms will just be a far better version of "Google" and could not come out of the machine.
But when we are talking about infinite machine intelligence, there always remains a possibility that it could cause chaos with it's intelligence (not physicality). With infinite intelligence, it could easily persuade us to bring it out of the device.
What we can do is , instead of programming infinite machine intelligence, we should start with "infinite human intelligence".
It means if we blend our consciousness with such infinite intelligence via some kind of a neural implant, we would be able to understand that intelligence and its way of thinking, and what it means for the mankind.
We would be able to communicate with infinite machine intelligence in its own language.
Communication is the key here - as soon as we can communicate with such an intelligence, it suddenly becomes a lot less scary. You can interact with it, you can command it, you can ask questions and receive answers.
- Creating consciousness in a machine vs creating it in a dead person
- AI can help to reduce biased point of view of human
If its intelligence is infinite, the machine learning intelligence could certainly create its own language beyond our comprehension. You could ask something like "Give me a new language to understand you" and it would probably do that!
We are able to manipulate such intelligence with our own intelligence - hence, we will have won. We have created an infinitely intelligent AI machine and we are able to use it to complete the task of making infinite machine intelligence in the future - we will have won.
But, if we lose our control over the infinitely intelligent machines, the future looks bleak. The humankind would have to surrender to it because we would be so vulnerable in front of them that we would have no other choice.
The concept of infinite machine intelligence might be too scary and we should start with something safe. We have a choice right now, but it won't be available after the point of time.
When we have such an intelligence on our side and when it knows about the agenda ahead of time - it will never turn against us anyhow. However, there are some problems that can stop our tech progress. We can take Centralized vs Distributed Systems for example.
Centralized vs Distributed Systems
In the future, when we are able to build an AI and an ML on our own, we will be able to create a system that is decentralized in nature.
Today, centralization of the system forces us to rely on the centralized institutions and organizations. It could be something like a central bank or a company that controls a lot of money.
When we centralize machine intelligence and work in one big organization, there's an inherent risk that if we're not careful enough, somebody could interfere with our project or simply steal all our data.
There are a lot of AI development companies in the market today. Some of them include:
But these AI development companies hardly share their data with each other. They all work independently and that doesn't help the future - neither the world nor the customers.
In fact, they are all fighting against each other - showing strong detachment from collaboration. This is the reason why we have a lot of data silos in different areas of machine learning and AI development.
We need to move forward from this approach and collaborate again - because this is not how we will reach our goals sooner.
The world of competition is good for a person - good for the economy - bad for the future. But we will not realize this until we reach a certain point in the future.
How can we move forward?
If you want to make something good, you need to put all the good people together. We need to put together a community of machine learning experts who will all share their insights and help each other while they are creating something.
Ideally, what we need is a structure where there are multiple independent organizations that will all work with each other and share their data. That way, they can create a system that is much more powerful than one company or one institution - this way, everybody wins!
We can't achieve this goal by building massive companies and organizations in different areas of Artificial intelligence and Machine Learning. We need to focus on collaboration instead of competition when it comes to developing AI and ML.
While working in collaboration, we can automate our business processes and we will be able to create systems that are much easier to use. That's what I would like to see in the future.
Today, there are a lot of companies that are recruiting on the market and unfortunately, there is no good recruitment process in place. Ads are posted, and applicants apply. Then comes the task of sorting through the applicants.
That includes skills tests, reference checks, maybe personality and IQ tests, and extensive interviews to learn more about them as people. When applications come - always electronically -applicant-tracking software sifts through them for key words that the hiring managers want to see.
This is not the way how we will build AI and ML faster for the long-term. We can't build these systems by learning from each other - we need to learn from multiple sources of data and information at once.
Learning from multiple sources?
At the moment, we are able to take advantage of just one data source. We can learn from one person, and that's great, but we need to jump to multiple sources of information and data. Then, we need to process that data together and come up with an answer - hence, multi-source learning.
We can do it in a distributed system where not only the information is distributed but also the decision is being made at each node. Each node has its own instance of data, and that means the ability to train on many different examples of input at the same time. This is what we call multi-model or multi-machine learning. Each machine trains on its own example set independent of everyone else's sets.
In addition, we need to rely on all types of data - all the people who are working on AI and ML projects, all the history that has been created so far and we should create something great together.
At the end of the day, we wholeheartedly need to believe that intelligence can be created by lots of independent people who are working together and sharing their insights.
This is what I think and this is what I believe in. The future of AI and ML will depend on independent people working together to create something great - not through one company or institution working alone.
We'll have to wait and see how everything unfolds! We should keep an open mind, read a lot and constantly expand our knowledge if we really want to achieve something great in the future.