More on AI

"AI" is often a misnomer

A broad field of study, artificial intelligence (AI) includes many subsets such as machine learning (ML), neural networks (NL), robotics, computer vision, etc. Within this umbrella is "artificial general intelligence" (AGI), however it's only theoretical and does not exist yet.

AGI and AI are often conflated by non-technical people, and it's important to distinguish the two in conversations and decision making.

Subsets of AI transform industries

It's subsets of AI that make an impact on industries, where finely tuned "narrow" AI is focused on particular problems and contexts.

The gamechanger in our current era is ML, and to be specific a subset of ML, Large-Language Models (LLMs). This is the "AI" that currently touches everyone, whether it's on a mobile device, website, bank ATM or subway advertisement. In this regard, people often use the terms "AI", "ML" and "LLM" interchangeably.

AI is a resource

Just like other resources such as an external API service, a font, an icon library or even a human developer on a team, AI is just another resource to be used when & where appropriate.

AI is only as good as its inputs

Large-language Models (LLMs) are the form of AI that's currently transforming the (tech) world, however it's effectiveness only goes as far as its training data (e.g. millions-of-x... millions of images, millions of articles, etc). As such, problems it can solve are essentially problems that have been solved before. Inversely, AI in this form has great difficulty with new problems, unique ideas, creativity and generally anything subjective.

As well, another form of input occurs when interacting with AI, where questions-asked, and context-given can greatly effect the quality of output.

The old adage, "garbage-in, garbage-out" remains true.

AI is highly subsidized, has diminishing returns and requires huge resources

AI is no silverbullet for business, as companies are in a war of attrition... bleeding money, running on limited VC funds. AI takes huge amounts of energy, even for menial tasks. And what happens when all available pratical data has been consumed, and the AI stops "growing"?

Lots of flipside considerations.

Do I use AI?

The quick answer is: Yes.

However, AI is a broad topic, and when asked the question "Do you use AI?", one has to quickly get into specifics for the conversation to go anywhere.

Do you mean "use AI"...

• as an assistant, as part of a workflow?
• to generate a complete product or service?
• within a product or service?
• to-do-everything-and-make-you-an-overnight-billionaire?

How I use AI

My relationship with AI is primarily as an "a-la-carte" resource to accelerate my workflow. I employ different AI mechanisms, turned on-and-off like a tap, as I need them.

Sometimes I don't use AI because...

• it's quicker just doing something myself
• it's causing too many bugs or going wild
• it costs too much
• it's taking too many (system) resources

AI as a product

Another wild frontier is the concept of having AI as an actual feature in a product or service. This is wholly different than using AI to improve one's productivity.

At a recent workplace, AI-derived content was a component in auto-generated reports which I built the UI template for. Just like any other resource, if it fails or is malformed in some way, consideration must be given to fallbacks and edge cases.

In another instance, for my own marketplace project, I have on mind integrating chatbot(s) as an alternative to the regular fuzzy search. Whether the chatbot becomes the primary user interface would be a design decision, informed by A/B testing.

In any case, if well-matched, AI can be a powerful component of a product and be a core feature of the business.