What is AI?

So, what is AI? Is it just a newer, higher level of programming? Hiding from us from what the computer actually does ? Is it evil incarnate? For this article I’m going to concentrate on Generative AI, what people see when they talk to chatbots

AI

AI accomplishes, amongst other things, the ability to allow the user to ask questions, and to give answers back, in a human language. Most people that deal with AI know nothing about the computer it is running on. It is ingests huge amounts of data and is able to respond to inquiries about that data in a human understandable way. You could ask for instance:

“What color is the sky?”

It would probably respond “blue”

Of course, it only knows what information you feed it, so, for instance, it has no concept of current time.

What time is it?

ChatGPT said:

The current time depends on your time zone. If you let me know your location or time zone, I can give you an accurate answer! Alternatively, you can check your device's clock for the exact local time. 😊

So what does AI do for us?

If AI only knows what we tell it, what good is it? It organizes our data in ways we can’t easily do. It can collect so much data, organize it, and spit it out in ways that we may not expect, but it’s not really inventing things. It needs to ingest huge amounts of data so that it can create what it can. Which is why it is so expensive to set-up. It takes up so much energy that Microsoft will be using a whole nuclear reactor just to process AI.

It can give you opinions:

I asked it about red wine…

The best red wine depends on your personal preferences, budget, and the occasion. However, here are some top-rated red wines across different styles and price ranges:

1. Classic & Iconic Red Wines

• Cabernet Sauvignon – Bold, full-bodied, with dark fruit flavors and strong tannins.

  • Best Pick: Château Margaux (Bordeaux, France)

  • Budget-Friendly: Louis M. Martini Napa Valley Cabernet Sauvignon

• Pinot Noir – Light-bodied, smooth, and elegant with red fruit and earthy notes.

  • Best Pick: Domaine de la Romanée-Conti (Burgundy, France)

  • Budget-Friendly: Meiomi Pinot Noir (California)

• Merlot – Medium-bodied, soft, and smooth with plum and chocolate notes.

  • Best Pick: Château Pétrus (Bordeaux, France)

  • Budget-Friendly: Duckhorn Vineyards Merlot

2. Bold & Intense Reds

• Syrah/Shiraz – Spicy, smoky, and full-bodied with black fruit and pepper notes.

  • Best Pick: Penfolds Grange (Australia)

  • Budget-Friendly: Côte-Rôtie by E. Guigal (France)

• Zinfandel – Rich, jammy, and fruit-forward with spice.

  • Best Pick: Turley Old Vines Zinfandel

  • Budget-Friendly: Ravenswood Zinfandel

3. Elegant & Complex Reds

• Nebbiolo – Known for its high acidity and tannins, found in Barolo and Barbaresco wines.

  • Best Pick: Gaja Barbaresco (Piedmont, Italy)

  • Budget-Friendly: Vietti Barolo

• Tempranillo – The signature grape of Spain, used in Rioja wines.

  • Best Pick: Vega Sicilia Único (Ribera del Duero, Spain)

  • Budget-Friendly: Marqués de Riscal Rioja Reserva

4. Everyday Drinking Reds ($15-$30)

• Bogle Petite Sirah (California)

• 19 Crimes Red Blend (Australia)

• Alamos Malbec (Argentina)

• Chianti Classico by Ruffino (Italy)

If you have a specific preference (bold, light, fruity, or earthy), I can refine the recommendations further. 🍷 Cheers!

I was curious and asked ChatGPT for its sources. Under bold intense reds, two of the sources were Australian. So, did it make some of those choices on the data it new about? What would happen if the AI only had French sources. This to me, it the interesting point of AI. Was the LLM skewed for the data we’re interested in? In the past, ChatGPT was behind by a couple of years. For instance, it may not know who won the 2024 World Series. They now have provisions to keep up.

So, now I get to talk to computers in English?

Most of my career, I served as kind of a translator. Someone said they wanted the computer to calculate sales tax on a sale, I would be given the formula, in English and I would write code so that the computer would ask for amount , and say 6 percent to it, and display the tax and the total

Of instance , For a $1.00 item , calculate 6%, and the amount , then add the two together.

In BASIC that might be.

REM this is mostly correct, depending on the BASIC flavor
REM do not use this at home.
PRINT "Enter amount of item "
INPUT item
LET tax = item * 0.06
LET total = item + tax
PRINT "Item " ; item
PRINT "Tax: " ; tax
PRINT "Total: " ; total

In ChatGPT I typed
I want you to ask me for the price of something, then take that value, calculate 6% of it, then show me the item price, the tax amount, and the total of the two

It’s response was:
Sure! Let's do this.

What’s the price of the item?

You said:

1.00

ChatGPT said:

Got it! Here's the breakdown:

• Item Price: $1.00

• Tax (6%): $0.06

• Total: $1.06

Want to try another one?

So I explained everything, in English. And it understood.

Of course this is a very simple examples, and as I said above, out of the box, there are somethings AI (GenAI LLMs) don’t know, but you can give the process hooks into, say finding out the current time.

My current applications in GenAI

I am working on an application that purposely limits a generic LLM by giving it my own knowledge base, and telling it to ignore stuff out on the web. The idea is to be very specific, and to offer different advice to what is out there.