Insights on AI: Writing, Images, Benefits for Personal Growth, and Earth-Tech
Hey everyone, after using AI for more than a year and closely studying its progress, I’ve gained some level of insight into its usage and how it can be applied to certain fields. I’m no PhD, but here are my views on writing/creating/programming with AI, what AI will be able to do for personal growth, and generating new technologies for the benefit of the planet.
Before we begin, let’s define terms at a basic level to improve our understanding.
Glossary:
AI: a system that uses information to make decisions in an environment.
Machine Learning: a model that executes comprehensive data analysis and predicts the next action in a sequence with a certain degree of accuracy.
Foundation Models: use a neural network architecture (a network of neurons like our brains), called a transformer, to generate sequences of related data elements.
LLM: an entity that uses inputs to create a sequence of related elements of language.
Generative AI: an AI that uses the output of a foundation model.
Narrow AI: AI that is very good at doing one particular thing… and has a concentrated dataset.
General AI: AI that is good at a broad set of activities.
PEAS (1.): are the key components of an AI.
Performance: how well an AI does in relation to a goal that it is given. For example, how much gold it finds, how well its art comes out, and how much an essay needs to be edited.
Environment: the space in which the AI operates. For example, a cave, the internet, or the space inside of a video game.
Actuators: the instrument(s) an AI uses to act upon the world. For example, a robotic hand, a keyboard, or a web browser.
Sensors: the instrument an AI uses to perceive. For example, a motion sensor, a camera, or an nmap scanner.
The Dangers of Large Language Models (LLMs) for Writing
Generative AI language tools are good for creating massive sets of ideas at will, but they absolutely need an editor to go in and put the puzzle pieces together.
These AI tools will often repeat themselves. They will say things that don’t make complete sense once you read them carefully. These tools can be very dangerous to those who aren’t paying careful attention to their writing. It can be very obvious when someone has not written a piece and has simply used AI to generate a smart-sounding word-soup.
Generative AI can also write software code, which is a language in itself, at a pretty good level… depending on the model. IBM can write COBOL at 90% accuracy with their models for their mainframes. Others aren’t so good. Remember, that on a fundamental level, generative AI is using a vast dataset of code/language written by humans in order to predict and create its output. It does this by predicting the next likely word in a sentence/line of code, which belongs to a paragraph/function, which belongs to a blog post/object, which belongs to a topic/package. Therefore… the code is not 100% reliable, and you should always take it with a grain of salt. You need to be knowledgeable in the field that you’re using generative AI in. It needs an a great editor at this point in time.
Multiple Modes & Mediums
AI can already create fantastic images, like these. It can also operate software… and it will continue to learn over time.
A progression of how good AI has gotten at generating images. Images from Dalle2 and Dalle3.
That being said, AI has many uses beyond generating language. Soon, AI tools will be able to analyze a company’s IT data and generate actions for Cybersecurity analysts using Cybersecurity foundation models. This is because models are being built using IT data, instead of natural language like English.
Soon, there will be generative models that combine with narrow AI models to generate things in many different contexts, and they will become quite powerful. AI is already creating extraordinary content in multiple contexts and in multiple mediums.
Generative AI can generate actions within a Cybersecurity environment, create things within software like Adobe, and become the overall operator/user on a computer (although it may have some faults in this context, do to its reliance on statistics).
By multiple mediums, I mean video, images, voice, and actions within specific contexts like Cybersecurity, Sustainability, HR, Investing, and Graphic Design. The capabilities are already infinite… and it will only get better… and soon… AI will apparently will be able to generate high fidelity virtual worlds and films. I’ve yet to see that happen.
AI for Accelerating your Path:
Imagine a companion that can understand where you are in your life and recommend actionable goals for you… All you’d have to prompt your AI is, “create a set of pros and cons for each one of these decisions”… or “teach me how to create a pros and cons list when I’m making important decisions”.
This companion might record what you tell it over time and you might even give it some data to work with… like your grades in school, the performance reviews done by your boss, and the stack of your journal entries.
Maybe you tell the AI the story of your life and it returns the best action for you to take… but using real world data produced by API’s from around the globe or actual physics information from Wolfram. At the end of the day, computers have the ability to calculate vast quantities of things that we wouldn’t be able to string together.
But… this kind of technology is going to need some kind of expert system or narrow modules to go along with it… it will likely not solely rely on generative models.
For now… I can be the one to help you Accelerate your Path!
AI for Energy and Saving the World
Now you can even start to imagine AI’s that generate actions as robots, and models that generate new hydropower turbines to optimize for greater energy outcomes. This makes me optimistic for a future with more resources. Now imagine foundation models generating new gene-edited cucumbers, or greenhouses-in-a-box that can operate in deserts, tundras, temperate forests, winter wonderlands, and deep tropics. It’s all about how we make this technology more grounded in data and harness it to build a better future.
All we really need to do is build a high-fidelity model (one that predicts reality with accuracy) that enables us to see the outcomes of putting together different modular technologies that would help us facilitate life on Earth… and from these models calculate the cost of materials and the cost of production and test things out.
Closing Remarks
An important note. It is very easy to conflate terms here. There are many kinds of AI operating in many kinds of environments, automating a host of different things like HR, others designing tools in virtual environments, and even automatic vacuum cleaners gliding through your house. If you or your company would like to learn more about how you can implement AI in your business, don’t hesitate to reach out to me... I work with IBM to help companies enhance their Cybersecurity capabilities, create chatbots for their customers, write code, embed AI in a SaaS solution and essentially design ANYTHING that a company needs.
It was great to have you on. If you enjoyed this letter, consider sending it to your family and/or friends. It would help me a lot of you circulated these ideas.
What’s a project you’d like to create with AI? Leave it in the comments.
Here’s a link to my latest podcast on Nature and Data with Darling Ngoh.
Feel free to message me at any time to find out how AI can supercharge operations within your organization.
If you know a woman who values comfortable elegance: try to comprehend the quality of this blazer at a distance.
Sources:
Russell, Stuart J. (Stuart Jonathan), 1962-. Artificial Intelligence : a Modern Approach. Upper Saddle River, N.J. :Prentice Hall, 2018.
IBM Technology Atlas: https://www.ibm.com/roadmaps/