How to Use Artificial Intelligence (AI)

Photo by Igor Omilaev on Unsplash

At the release of ChatGPT in 2022 I was the techiest of bros. I still consider myself a tech bro but like back then I was fresh off the boat to Austin, TX, and I hopped on the tech wave relatively circumstantially, though I had always been a hobbyist for the most part. I’m not an engineer, I've only been on the corporate side of things. Since then I’ve stayed incentivized to be informed on technological advancements. My company was searching to find its more defined niche in this technology but I think that experience, along with some of the copyright infringement issues early on, kept me clear eyed.

It was an enjoyable time and I was hustling hard. Even more of a reason why Sam Altman is one of my favorite tech oligarchs. Here are some things to consider when interfacing with artificial intelligence or (AI):

1. It’s Propaganda.

I have a Master’s in psychology so I'm basically a certified propagandist. The term "Artificial Intelligence" was coined by computer scientist John McCarthy in 1956. It’s an umbrella term that refers to technologies that can exhibit human intelligence. Because the meaning is broad there are things that can proclaim being AI without adhering to a certain standard. Examples of this reputation editorialization will be threaded throughout this understanding of how to use this technology.

2. The Evolution of Artificial Intelligence

In this article, I highlight the following point further. The common consensus about the direction of this technology is Artificial Narrow Intelligence (ANI) to Artificial General Intelligence (AGI) to Artificial Super Intelligence (ASI). The definitions of these can be elusive. ANI focuses on specialized tasks. As of the release of this article, tech and AI leaders are currently talking about AGI with the notable exception of Mark Zuckerberg who is pioneering ASI, which I think is the right decision. Again because AI is an umbrella term some actors in the space refer to AGI as a fundraising buzzword.

One of the fundamental questions is, what is the vision for this technology? Sure let’s say we’ve only cracked the surface of ANI, which isn’t the case, but what is considered ground breaking with respect to AGI? A contingent of investors want to see AI do things for them. Like give it a command and have it execute that command to completion. Apple is a notable company that has committed resources to this integration. Is that AGI, or is that still too narrow with respect to scope of general human intelligence? Then, how is general human intelligence confined to a chat box or similar agent?

I think Zuckerberg is right to pioneer ASI because with respect to AI’s current iterations as chatbots, research tools, and an overall derivative of the internet, being most effective at synthesizing information and optimally facilitating the way that information is deployed, is super human. In fact, according to the Touring Test human-like intelligence is demonstrated by a technology’s ability to provide responses indistinguishable from a human. Using this threshold AGI doesn’t need to develop because it’s happening now through ChatGPT and its offshoots.

The point is, in addition to the inherent propaganda (of AI), human intelligence is open to interpretation with certain considerations and the way that intelligence is packaged matters. That’s one of the blessings of having a body, because humans can communicate intelligence in an expressive way. Transcendence of Artificial Intelligence and or humans alongside AI via a singularity is also a common end goal for this technology and connected ideologies. For some people this is a religion if not at least heavily religiously coded.

Photo by Emiliano Vittoriosi on Unsplash

3. Large Language Models VS. Deep Learning VS. Machine Learning

Machine learning is an umbrella term within the umbrella term of AI. It refers to algorithms (rules, instructions, and or parameters for learning) that learn from data to make predictions or decisions with a narrow scope. A Large Language Model (LLM) is a sophisticated form of deep learning. It's an expansive neural network that has been pre-trained on a massive diverse dataset of text and code. LLMs learn the fundamental patterns and structure of human language. This makes it a foundational model, meaning it can adapt to a wide range of tasks without being constantly primed to adjust.

Deep learning is a subfield of machine learning that uses algorithms structured in layers to create a neural network inspired by the human brain. Transformer architecture, which was introduced in 2017, is what allows deep learning models in these neural networks to process vast amounts of data and understand the relationships between words. Transformer architecture uses an attention mechanism to weigh the importance of different words in a sentence simultaneously. This is what gives LLMs their ability to grasp context and generate coherent, relevant responses.

At its core a LLM is using its programming to decipher patterns in language and similar inputs to predict the next appropriate response. Some people are rightfully calling this magic or supernatural which is an inherent component of all technology, but this is where the technology ends and human behavior begins.

4. AI can Hallucinate

This is a phenomenon that can happen where a query from an end user falls in a gap in an AI agent’s knowledge, but its programming is still inclined to give an answer. “I don’t know” isn’t necessarily a default response for an AI agent so when it has to process information it doesn’t have a reference for, this can lead to responses that are out of pocket to say the least. Because of the mystical nature associated with getting one’s most curious questions answered and because of a lack of knowledge I hope this article reduces, people may be inspired to ask artificial intelligence questions about an array of topics.

I’ve heard stories about people asking ChatGPT and similar models about their future, to interpret the Bible, to help them process emotions and more. These chatbots are intended to be great research tools with a lot of runway for R&D. ChatGPT is not Google (AI can answer questions and automate tasks, a search engine points you in the direction to where you can discover relevant resources for whatever your cause is). Chat models can help you write copy or learn more about a subject but you should still proof read, verify sources, and do your due diligence to make sure the information you are being encouraged to parrot is correct. AI is a creative research tool that presents potentially fragmented information as a unified thought. And while the process behind that may be spellbinding, it’s not groundbreaking enough to abdicate one’s tether to reality, to the degree anything is.

5. Humans Program and Train AI

If a human is involved it’s not perfect. Companies hire professionals to train AI models. Tech companies use collective intelligence to try to develop a baseline consensus for their technology, but they are also inclined to make this technology profitable. There is more than one example of AI that is untrained or unmonitored going left real quick. A notable example is Grok’s “Mecha Hitler” moment. There is a considerable amount of resources allocated to scale this technology and keep it as ethical as possible; but other human motives can corrupt the integrity of an AI agent whether it be from the perspective of training and development or from how an end user interfaces with it (pressing an AI client on potentially harmful questions and topics). Yes, technology tends to shape human behavior, but AI is not a reason to not develop your sense of values, morality, or common sense.

As I mentioned earlier I'm partial to OpenAI and Sam Altman but other influential figures in AI include Google, Microsoft, Amazon, Invidia, and Meta. There is so much money in this space, it’s a multi-billion dollar industry, so there are numerous AI companies out there right now. Apple is an honorable mention because I talked about how AI might be a flop for them. That said, they are approaching AI from one of its more promising functions and applications. This article consists of the core things to remember when you open your next AI agent session. Before this technology is a religion it’s propaganda. Before it’s full of wisdom and insight it’s input dependant, trained by humans, and potentially unprofitable. It’s literally telling you what it thinks you want to hear. Regardless of form factor that kind of behavior, from a survival perspective, should always solicit some type of feeling of caution.