Hi, in the upcoming series of blog posts, we will be discussing a couple of basic machine learning algorithms. Although they are general purpose algorithms, we will be taking natural language processing (NLP) side, especially having chatbots in mind. The target audience I assume for these articles will be anyone who's interested in this topic and may or may not have a technical background. You just need to use your common sense and logical thinking capabilities to understand those concepts. Each topic will have two parts, wherein first part, we will discuss how we would approach and break down a problem and in the second part, we will try to apply what we've learned in a demo. It's not necessary that you need to be a programmer, only intention here is to present the logical flow of implementations. Because our main focus will be on the thought process rather than implementations, I will be jumping around between two or three languages and frameworks for the sake of easy implementations.
what is artificial intelligence ? #
currently, we're leaving the peak of Gartner hype cycle for deep learning (DL). Although the impact of DL in the industry started in around 2000, the mainstream media buzz hasn't picked up the topic until late 2009. Cheap and powerful processors made by chip designers (like Nvidia, Intel), accelerated technologies for efficient Big data processing on commodity hardware (like Google's MapReduce) and outperforming deep backpropagation models, followed by a flood of papers by researchers from multiple fields of study sharpened the DL algorithms within a short period of time. Contributions to open source community in the form of DL frameworks, test and evaluation tools and datasets gave momentum to the acceptance of these algorithms deep down among people. DL changed the face of the business and industry as well. Through automation and better performing algorithms, business profit got maximized by cutting down operational costs. A lot of manual labors got replaced (and still getting replaced) while new ones being created. Some others provided free tools and training to promote their pay per use platforms. This helped serverless applications to bring ML on low-end platforms as well.
referring back to the hype cycle, I would like to admit that a lot of people (mistakenly) heard or used the terminology Artificial intelligence (AI) instead of Deep learning (DL). Media published everything as the machines taking control over human, as they were able to think themselves as we do with AI. They presented deep learning algorithm as the software design of human brain. AI is not magic. AI as a branch of science is still in its primitive stage. It has a great community of researchers contributing from different fields of study. DL is an (a set of) algorithm implementation(s), which is able to utilize the power of today's hardware, and thus outperform other algorithms. Thus, we could say that DL belongs to a set of self-learning algorithms, which are generally called, Machine Learning Algorithms (ML). We could consider ML as a subpart of AI.
ok, I got it. why should I care? #
if you are closely related to an industry where AI could influence, you should be up to date with this technology. It is proven that the current state of AI itself is powerful enough to change the way we interfaced with the technology last time. If you are an investor or a creator, you should be adopting it and see things clearly and differently from now on. Here are a few example applications that current AI can handle, just in case you are wondering:
- your devices will change its interfaces very soon, and you should be able to adopt those changes to match your business. One example is conversational interfaces.
- Analyzing your target end and the effect of efforts you place should be much more easier to keep up with the competition. Upgrading to latest tools and tech is the best strategy.
- Current AI could influence more new areas than it did previously. And of course, its growing faster than we think. Some examples are Transportation, Agriculture, Logistics, Medicine, Mining, Design, Communication etc.
- Today, everything that's connected to the internet generates a huge amount of data that we couldn't control manually. To make sense of this data we need the help of AI.
let's conclude #
We have briefly discussed how AI, especially ML could change the way we think, act and interact with the technology. From now on I will be leaving the term AI in peace ☮️. We will be discussing a small subset of algorithms in detail in upcoming articles. Don't worry, we will start from the very basics of computation itself. The topics are chosen in a way to keep them on track with a specific application of NLP, automatic sentence generation. So.., I hope we will meet again.
footnote: I have mentioned in the above article that, AI is in its primitive stage, its true for the present situation. No one neither knows when we will reach the goal of AI, which we believe to be Artificial General Intelligence (AGI) nor how machine intelligence it would look (behave) like. The growth of AI not only depends upon the algorithmic advancements, also depends upon different other achievements such as new hardware inventions. Currently, a lot of research is happening to invent the useful quantum computer (quantum computers make use of quantum mechanics and are able to solve np-hard problems that our classical computers can't solve by introducing probability concepts). People in the field of AI believes that this could bring drastic improvements in ML as well. There are new inventions in machine learning algorithms as well - many of us believe Hinton's capsule networks might cause another breakthrough in machine learning after neural networks (Hinton's one of the pioneers of Neural networks as well). Currently, capsule networks in its implementation require performance improvements to run on our current machines. It's an open engineering problem that anyone can contribute.