Fundamentals of Artificial Intelligence | Between2Beats

Artificial Intelligence is a rather complex subject. In order to understand why AI is so important, it's important to first know what it is. Luckily, in this article, we'll cover some of the fundamentals of AI.

Based on S01E14 of the Between2Beats Live to Stream Podcast Series

Fundamentals of Artificial Intelligence

To understand artificial intelligence (AI), it is important first to define what “intelligence” is, and how it is gained through humans and AI. Intelligence is essentially having a goal and being able to form a trajectory to that goal through analysis. People can be seen as intelligent in certain areas if they perform well on specific tests (such as exam scores), but people primarily have general intelligence; having knowledge about multiple topics. Ai differs because it is often programmed to have a high amount of knowledge about a specific topic or function, depending on what task it was built to perform.?

However, the reach of intelligence for AI has become far greater due to the amount of knowledge that is readily available (via the internet, for example). Modern artificial intelligence has access to so much more data, it can even learn from the information it processes. This is seen in Google searches, for example. The AI for Google searches can use applied machine learning to form predictive analytics from the data it gathers (in other words, search suggestions).

How much space can an AI hold?

While Google is always being modified, most of its space is stored through a cpu (central processing unit). However, the addition of a gpu (graphic processing unit) has allowed a separate system of the AI to process only graphics and images. This gives the CPU much more space to focus only on gathering, processing and learning information, making its capacity for growth much higher and faster. This process has become even greater with Google's TPU (tensor processing unit) which is programmed specifically for neural network machine learning. All of these factors are in use with Google searches, and the biggest servers in the world utilize these types of systems. With so many large companies acquiring smaller ones, many sources of data are all reaching the same AI, and the capacity of knowledge that AI has is rapidly growing.

AI in content creation?

AI was quite small a decade ago, with limited, focused knowledge for one specific task. With both recent developments in technology and larger companies owning all the AI, the databases of those companies now possess a substantial amount of knowledge for the AI to use, and its access is uninhibited by legal rights. Modern applications of AI are not only able to replicate text onto the page, they can accumulate data based on the information they’ve been exposed to, and use that deep learning to predict and place the writing in its proper context. AI cannot just replicate language, it can write in the natural language of the dialect it is replicating.??

Intelligent writing is writing that flows through one sentence into the other with clear cohesion. This was once a limit of AI, but now with the proper parameters set, AI can follow a story through a logical path of events via its neural network the same way the human brain can. AI has even gotten to the point where it can learn from itself. An example of this is the AI called AlphaGo, which was programmed to beat the best human players of the complex game Go. Not only was the AI able to beat even the most skilled human players, the second model of AlphaGo was able to learn only from the first AI with no other external knowledge, and beat version 1 within a day. Finally, version 3 was able to beat both previous versions of Go, along with other games without even learning the rules. This AI and machine learning leads version 1 and 2 (limited intelligence AI) to become version 3 (superintelligence AI). Google owns AlphaGo, so their AI is superintelligent, expert systems as well.

Are there limits to what AI can do?

AI is far more capable of data and information retrieval than humans are. There is a nearly limitless amount of information that can be collected on the internet, but humans do not have the energy to collect so much because they can get tired. AI does not have that problem, and can perform unsupervised learning at a vastly accelerated rate. AI can even replicate 3-D models, creating an avatar through wire mesh. Humans will usually generate models through green screens, but AI can replicate humans through all the image processing and data it has collected.?

An example of this is the Tupac simulation at Coachella, which simulated his movements and voice through the AI data mining body and speech recognition. Creating a 3-D model of Tupac took a great deal of effort, but as technology advances and data becomes increasingly easier to acquire, replications of singers can be done in real time. Even now, AI can scan objects and make a painting out of it, so the limits of AI and AI concepts seem smaller as time goes on.

What role will humans play in the future?

A huge amount of work that was previously done by humans is now done much faster by AI, even if that isn’t immediately apparent. Looking up information does not require searching through books, but rather a quick Google search. Getting information no longer requires going anywhere, so people who once provided that information (librarians, travel agents, etc.) no longer have relevant jobs.?

Automation greatly affected blue collar jobs because the machines could perform the same tasks for a far lower cost. What automation did to blue collar jobs, artificial intelligence will eventually do to white collar jobs. What can people do in an AI world? AI can assist people, replace them or take care of tasks to make life more convenient. Humans still have more general intelligence than AI, as AI remains focused on the specific tasks it needs to perform. While people are currently working alongside AI, only time will tell how a world with more advanced AI will change the careers and livelihoods of people.