I Don't Know If I Can Do It, Let Me Calculate First

I find AI technology as fascinating as I first knew about the mobile phone, my Nokia 3310. I was able to send SMS and play the snake game for hours, and I wasn't complaining about its limitations. I was a happy kid.

Then, I witnessed the evolution of mobile phones, from no camera to 3 camera heads and front camera heads, smile-sensor photo capturing, face recognition, built-in RFID, to the latest Samsung AI 'Circle to Search' without switching apps.

Technology has never failed to amaze me.

When OpenAI launched ChatGPT, I was a happy kid again, marveling at how quickly I could get answers, how it wrote jokes, poems, song lyrics, and was the world's most wonderful personalized Wikipedia. And then, Midjourney dropped the bomb, and I was even happier. I have tested more than 50 AI platforms (actually, I lost count), from paraphrasing AI tools to video generation AI platforms. I try them whenever I have spare time

And then, I took a pause. I realized I needed to stop being just a user. I needed to know more, and I did! As I discovered new and interesting AI platforms, I eagerly shared my findings on LinkedIn.

One year, 90 articles - not something I had done in the past, but I did it happily. Today, I want to go back to basics, to explore the fundamentals of AI.

What needs to be calculated before AI can prompt me with an answer?

When an AI is prompted with a question, it needs to perform various calculations to generate a response. Here's a breakdown of some of the calculations that take place:

• Retrieval: Before calculating an answer, the AI must retrieve relevant information from its knowledge base. This involves searching through the vast amount of data it has been trained on to find the best match for the question at hand.
• Data analysis: Once the relevant information is retrieved, the AI must analyze it to extract relevant facts and ideas that can be used to generate a response.
• Inference: Based on the analyzed data, the AI must make inferences and deductions about the question to determine the best answer. This involves using various algorithms and heuristics to identify the most likely response.
• Language generation: Once the answer is calculated, the AI must convert it into natural language that humans can understand. This involves selecting the appropriate words, sentences, and phrases to generate a coherent and informative response.

These calculations are performed in milliseconds, allowing the AI to quickly provide you with an answer.

When you ask a question, the AI needs to retrieve relevant information from its knowledge base. This knowledge base can include anything from general facts about the world, to specific information about a particular topic. The AI uses sophisticated search algorithms to find the most relevant information for your question, and this can involve using a variety of techniques such as keyword matching, topic modeling, and semantic search.

Once the relevant information has been retrieved, the AI must analyze it to extract the key facts and ideas that will be used to generate its response. This analysis involves identifying patterns, relationships, and dependencies in the data, and extracting the most important information from the data.

Based on this analysis, the AI must make inferences and deductions about the question to determine the most appropriate response.

Here are some examples:

Scenario 1

Question: Who invented the light bulb?

Thought process: The AI searches for information about the invention of the light bulb and identifies Thomas Edison as the most well-known inventor associated with the light bulb.

Answer: The light bulb was invented by Thomas Edison in 1879.

Scenario 2

Question: What is the largest country in the world by area?

Thought process: The AI searches for information about countries and sorts them by their land area. It identifies Russia as the largest country in the world, spanning 11 million square miles.

Answer: The largest country in the world by area is Russia, which covers approximately 11 million square miles.

Scenario 3

Question: What is the largest mammal in the world?

Thought process: The AI searches for information about mammals and sorts them by their size. It identifies the blue whale as the largest mammal in the world, with some individuals reaching up to 100 feet in length.

Answer: The largest mammal in the world is the blue whale, which can grow up to 100 feet long and weigh up to 200 tons.

I would like to turn the question back to you, dear readers.

What have been your frustrations with the current state of AI platforms such as Google Bard, Claude AI or ChatGPT? Have you found it to be too slow in responding, or have you struggled to get research-based findings that are reliable?

About Jean

Don’t forget to subscribe to my newsletter and follow me on X (Twitter), LinkedIn for more updates and development of AI and Web 3.0