Artificial Intelligence, or AI, is now no longer a thing of fantasy. It has arrived, and that too, in style. If you take a look at social media feeds, AI is trending, and for good reasons too. With tools such as DALL-E or ChatGPT, AI is now able to undertake creative tasks such as drawing a picture or writing code.
But have you ever wondered how the AI knows how to do these jobs? Who taught AI these things? Has the AI already become self-aware? If you, too, are curious and want to learn more about the workings of AI, then scroll down for more information.
Table of contents
Is Artificial Intelligence really intelligent?
Before delving deep into the topic, first, we have to ask a serious question that must have been on the minds of some of our readers. And this question is: Is AI really intelligent? Despite popular belief and portrayal, AI is not all-encompassing. Rather, it is the result of human ingenuity. Remember that AI, at its core, is only a set of algorithms that rely on external data to produce results. So before marveling at how AI knows stuff, just keep in mind that it was we humans that programmed AI that way. While it is highly efficient, it is not sentient.
Programmable
We have established that AI is not capable of independent thought. By the same extension, it is also proven that the reasoning applied by AI during problem-solving is not an inherent quality, but rather the result of rules and algorithms programmed into the central unit. That is why an AI can be efficient at what it does but not actually knowledgeable or smart, yet. It can correctly point out a cat or identify a cat, but it can’t understand what a cat is or what it means.
Lacking Common Sense
Then there is the lack of common sense that doesn’t allow AI to be compared to real intelligence. A human’s comprehensive ability or logical reasoning can be trained into an AI model; however, it is not the same as good old common sense.
It is common only to humans and is gained while experiencing life and their surroundings. And we agree that an argument can be made that the same can also be taught to an AI, but with so many different cultures, ethnic groups, and traditions, it will be hard for the AI to understand the nuances of human behavior. It is the same reason why we can train AI to recognize a smiling face, but the AI can’t infer why the person is really smiling, but a human can with minute clues.
Limited “Thinking” capabilities
And lastly, we have the final nail in the coffin, and that is the limited learning capability of AI. Now, we do not mean limited in the term of data consumed or its ability to process data, but rather, its functioning, which is entirely dependent on the data it is trained on. AI can’t think outside the parameters of the data it knows. Hence, it has limited thinking. While a human can learn or experience new things by itself, an AI has to rely on humans to feed it new data parameters to expand its thinking capabilities.
AI Memory
| Memory Type | Description | Example |
|---|---|---|
| Short-Term Memory | Temporary storage for recent information. | Recent conversation context |
| Long-Term Memory | Permanent storage for learned knowledge. | Pre-trained knowledge from training data |
| Episodic Memory | Records of specific events or experiences. | Past interactions with users |
| Semantic Memory | General knowledge about the world. | Facts, concepts, and common knowledge |
| Procedural Memory | Knowledge of how to perform tasks. | Skills, algorithms, and procedures |
| Working Memory | Current focus of attention. | Information being processed in real-time |
Then how come AI knows things?
If it is true that the AI isn’t intelligent, then how come it knows things? And especially the things that weren’t taught? You must have read a few articles where even the scientists behind ChatGPT were unable to explain how the AI tool arrives at a certain result or its thought processing. Does this mean AI is able to learn things on its own?
Well, no. AI is more of a highly functional parrot, which, while undeniably clever, is still only drawing results from a set of pre-defined data. Like how a parrot imitates human speech, an AI imitates human logical reasoning, or at least it tries to. But just like the parrot doesn’t understand what it says, neither does the AI.
Does AI really understand the things it says?
And this brings us to yet another intriguing question, which is: Does AI really understand the things it says? It is a question we can answer with quite a bit of certainty. And which is absolutely no. While there may come a time when the AI will truly be sentient or self-sufficient, today, the AI is highly dependent on human programmers and thus, incapable of really understanding what it actually says. We can train AI to recognize a particular fruit and assign a word to it, but the model will be unable to comprehend what that fruit is or what the word means.
How does AI work?
But the question still remains: how does AI actually work? While AI is a tool made by humans, it is very good at what it does, which is solving complex problems. Today, almost every industry incorporates AI in one way or another. The AI uses a pre-defined data set as a training model and works on algorithms and rules placed beforehand. It uses these algorithms to manipulate and simulate results from the available data.
The entire process can be divided into smaller segments, such as:
Input Process
Like the first step in any computing system or program, the AI also needs certain input to work. Without data, AI won’t be able to function. So the programmers collect data and compile it accordingly for the AI to learn from and train. However, this data can be versatile; it can be text, images, videos, or even audio. But the programmers also have to ensure that the data is clearly defined and presented with enough context for the AI to understand the basic subtleties.
Data Processing
Then comes data processing, meaning, using the data from the input step and processing it as per the rules and algorithms in place. When the AI is asked to perform a task, it first checks the data and processes it for possible outcomes. With the advancement of technology, AI can even process data in real time and derive accurate results accordingly.
Outcomes or Results
After processing said data, it is obvious that AI will present results. These outcomes, whether positive, negative, accurate, or inaccurate, are not of concern in this step. Rather, it is the ability of AI to present outcomes and their delivery.
Adjustments
Then comes the adjustment process, where the AI can learn from its mistakes. If the derived outcome is incorrect or inaccurate, the AI can try to process the input again. It adjusts some of the algorithm parameters to derive a better solution. It is this step that differentiates AI from other computing tools, as the AI, when programmed accordingly, can adjust the rules and algorithms to better itself.
Final Assessments
And finally, the AI also does an assessment of the task it performed after making adjustments to not only retain the experience but also learn from it so that it can perform better in the future. This step also allows the AI model to make inferences and predictions based on the data used. Also, this step can be used to provide much-needed feedback that can be helpful in making future algorithms.