Artificial Intelligence - Human Required

Our modern Artificial Intelligence (AI) is an amazing technology, made possible by our ability to process and analyze more data than ever before. But the chat bases using Large Language Models (LLMs) are simply that - highly analyzed data. Our current generation of AI is certainly artificial, but it is not actually intelligent, the way we typically define the term.

An LLM-based chat bot works by mapping our words to "tokens" - which are basically just a numerical identifier. By processing all the words written and digitized, the model learns some words are more likely to be used together in a particular order. By processing multiple languages, it even begins to find common patterns between words meaning the same thing - giving it the ability to process and produce the same ideas in each language on which it’s been trained.

But an LLM doesn't understand any of the words you type, or those in its response. When I say "fish" it conjures up all your experiences with the water animal. Whether from an aquarium trip, a fishing trip, or a restaurant menu, you have a strong mental concept of what the word fish means to you. For an LLM, fish is a number, and that number is closely mapped to many other numbers. Those other numbers are things like "aquatic," "animal," "food," and so forth - all the concepts you think of too - but just as statistical numbers.

If you ask an LLM if it knows what a fish is, it types its response, making you think it does.

But the answer is just the statistically best arrangement of words (tokens) based on your prompt. You gave it a math problem when you asked, and it gave you the mathematical result as a response. Admittedly, the actual process under the hood is far more complicated, such that we don't truly know how some of it works, but the basic concept is these statistical predictions. So why does the AI feel intelligent, even if it has no comprehension of what we're asking it or what it's saying back?

Because we are intelligent. We know what we asked and we can understand the response. The mathematical arrangement of words on the screen creates thoughts and ideas within our minds, thus allowing our intelligence to create an understanding of what we're reading. The AI's response has no meaning without us there to observe it. If people aren't there to interpret the results, random gibberish has the same amount of meaning to the system as a well-structured response.

The old question, "if a tree falls in a forest and no one is around, does it make a sound?" has its AI modern version. Is an AI's response intelligent if no natural intelligence is there to read it?

Does this distinction even matter? Is token based processing all that different than how we understand things? Maybe, maybe not. But the analogy of actors is helpful to explore this idea. "I'm not a doctor, but I play one on TV" is the phrase which comes to mind. An actor who plays a character so long they know all the vocabulary still doesn't have the skill to do the job. A TV surgeon might be able to say all the right things, but they don't have the training and understanding to do surgery in the real world. Or imagine a captain of a science fiction spaceship. They speak convincingly about a topic which doesn't even exist. Actors understand the prompts and speak the right words - but without needing to understand what's really being said.

Our AI bots are basically actors. They pretend to understand and act out the part accordingly. In the end, this makes for an impressive show, but they are not taking the real role away from people who do understand the topic. This is also why AI bots are more likely to make mistakes and hallucinate. They know how to say things that sound good, but because they don't understand what they're saying, they don't know when they are wrong. It still takes a human to read it, understand it, and either fix it (or take the risk if leaving it unchecked). The human in the loop is a key part of making an AI useful and helpful. It's our role to put the intelligence into artificial intelligence. Without us playing our part, the results truly are just artificial.

When we're not there to understand, AI doesn't make a sound.

Upcoming Events

The week after this is Atlanta Tech Week - a fully packed week of events all around Atlanta (more than 50 in a single week). If you're looking to go, ATL Tech Events is a community partner, and you can use ATW20 for 20% of the cost of the general ticket.

For events this whole month (and beyond), be sure to follow ATL Tech Events and visit the calendar!

Latest Episode of Perimeter Business RadioX

We had Kelly Kierans and Scott Williford from CelticQA Solutions on the latest episode, talking about the importance of Quality Assurance (QA) and their new solution QAConnector.

Week In Review

It was a short week, with Memorial Day on Monday, but packed in a lot of conversations and two great events! Check out the topics from this past week:

• Job Event Calendar?

• SSO is more secure. (A poll)

• Two nights of great events.

• CelticQA on Business RadioX

There's still time to join in on the conversation, vote in the poll, and share your thoughts!

In Conclusion

It's June! This first half of 2025 is almost over, but there's still more to get through and grow through!

I know I mention it every week, but ComplianceXO is coming out of soft-launch and into a full launch! If your company has IT and has risk (every company does) - please reach out and let's talk about how ComplianceXO can help you get a handle on managing that risk!

If getting back to work is a goal of yours, let me know if you would like to be in the You Just Found ME™️ job seeker spotlight. It is free to all jobseekers! Send me a note!

I want to remind everyone I offer referral bonuses to any work you bring me through Mirability, LLC - if you're interested. If there's anything I can help you with, I'd love to hear about it.

I hope this coming week is what you need it to be and that everyone stays safe!

Be sure to check out my online merchandise. Remember, 100% of the profits for any You Just Found ME merchandise goes to support that program for job seekers!

https://www.cafepress.com/shop/Mirability

If you want to keep up with everything I’m posting, click here and also the bell (🔔) to be notified when I post!

Follow ATL Tech Events to be kept in the loop about exciting Atlanta area technology events!

Subscribe to the ATL Tech Events newsletter here: https://eventhorizonmedia.substack.com/s/atl-tech-events

Follow You Just Found ME™️ to help support job seekers!

Follow Mirability, LLC to learn more about ComplianceXO and our goal to take your company from risk to reward! Be sure to check out the YouTube channel too!

Subscribe to my Substack here: https://ebspoke.substack.com/

I'm on Medium as well: https://ebspoke.medium.com/

About Erik

Erik Boemanns is a technology executive and lawyer. His background covers many aspects of technology, from infrastructure to software development. He combines this with a "second career" as a lawyer into a world of cybersecurity, governance, risk, compliance, and privacy (GRC-P). His time in a variety of companies, industries, and careers brings a unique perspective on leadership, helping, technology problem solving and implementing compliance.

He's available to help you with any of this now too!