Wow, less than a year between blog posts. Go me! Admittedly though, this post is more like the rantings of a madman.

In this post I want to describe why I think AI is like the Ford Model T, and what it means to be absent self-driving. For those unaware, the Ford Model T was a car produced from 1908 and is generally considered to be the first mass-produced vehicle for personal use.

AI is like the Model T because it's the first time it has been mass-produced and generally available to everyone. The barrier for entry is nothing. Any person, anywhere, can now load an AI model and start hacking away. Prior to the Model T, a motor vehicle was a luxury only for the ultra-wealthy, much the same way AI was only for the mega-corporations.

But what is AI? What is going to happen with it? What about the job losses?

What is artificial intelligence right now?

If you believe the hype, then AI is the next industrial revolution. It's going to replace all jobs and Humans will spend their days relaxing and enjoying leisure activities, as was [https://youtu.be/xS8xX3usi4c](predicted by school aged kids in the 1950s and 1960)s. How I wish this was true. Hold on, but hype is always right.. remember the dot-com era, Y2K, blockchain , cryptocurrency and NFTs. Those were all ultra-successful and fulfilled everything the hype-machine promised.

AI is a parlour trick, a sleight of hand. AI cannot cannot think, it cannot analyse, it cannot solve, it cannot rationalise, it cannot make decisions, and it cannot learn.

But what can AI actually do because it is clearly producing some amazing results. AI can infer information based on what it has previously been trained with plus a bit of random variance. It turns out, that Humans has produced an insane amount of useful data over the past 40 years. It turns out that most of us deal problems day-to-day that have already been solved and documented elsewhere, thus allowing a trained AI model to provide us with a known solution (sleight of hand).

Let's break down a very simply example to illustrate what I mean. If we ask AI "What colour is the sky?" we will most likely get some variation of "The sky is blue". The added context the model will produce is based on how your prompt is manipulated prior to being given to the model. Tools like ChatGPT do pre and post manipulation of the prompt and results to provide a better experience. So the model may give you some fluff around the answer, but it will tell you the sky is blue.

We consider the answer of blue to be largely correct, but how did the model know this? The model was trained on millions of sentences that contain information about the colour of the sky. Think about every book, article, blog post, forum post, text message, facebook post, instragram post, tiktok post. The AI has seen all of these, and it's built a probability matrix that gives is a list of possible answers to the question "What colour is the sky". We can safely assume that the vast majority of times the sky has been described across all literature the colour used was blue, but this is not the only colour. The AI model will have a small chance of returning other colours based on a random chance. This means it is possible for the model to tell you the sky is "pink", which again would be correct given specific circumstances.... but the model doesn't know this, it can only infer this.

Just like the Ford Model T was not a self-driving car, AI models are not thinking machines. They are a probability distribution system trained on the entirety of Human Knowledge to give you mostly-accurate solutions to pre-solved problems. They use previously seen information and chance to try and give you an answer. They don't know or care if it's correct or not because it's just some complex simple math.

What will AI do to the world?

Within 10 years of the launch of the Ford Model T, the industry to support Horses and carts had all but been destroyed. Nobody was buying horses, nobody wanted to re-shoe horses and nobody needed carts to be pulled by horses. The Ford Model T had created an entirely new industry by destroying an old one. While the cars had destroyed the role of the horse, it had also created many new industries. These new cars needed parts, tires, fuel, lights, repairs, maintenance. The mass producing or cars was also significantly quicker than breeding horses, thus allowing for more jobs to be created than originally destroyed. Much like moving from horse to cars, the introduction of AI will likely create more jobs post-correction because of the increase in efficiencies gained.

It is fair to say the launch of the Ford Model T cost people their jobs, but it created far more jobs than it destroyed. The increase and advancement in technology allowed things completely unimagined to become real. Humanity did more, built more, and accomplished more because of this. In my view, AI is going to be largely the same. AI as it has been designed today is the next step in technological advancement. It's like moving from a horse to a car, or a screwdriver to a power-tool. It is increased automation.

Do I think AI will completely change the world? No. AI is a tool that can be used by people who already have an underlying amount of expertise. The rise of non-technical people writing code (vibe coding) is going to create more cyber security jobs than ever because the quality of code produced by AI models is poor at best, buggy always, and seldom works accurately. You still need experts to take the output and give it quality, reliability and safety. The invention of power tools didn't turn everyone into a builder, but it made builders more efficient.

What about real AI, are we close to a true AI?

No. And I think there have been some good arguments put forward that the current pathway we're on for AI is actually going to slow the creation of a true artificial intelligence. We're investing insane amounts of time, data and money in a machine learning approach to build large relational matrixes of data and selling them under the brand-name "artificial intelligence". We're not actively trying to create intelligence or the singularity.

The approach we're using for current AI models is to ingest a large amount of information and build a massive matrix of relationships between words. It's machine learning with extra steps. Spend any real time asking the most advanced AI models complicated questions and you will quickly realise how often they are wrong and how wrong they can be.

Now, I love AI. I love the current generation of AI, I use them daily in my work both as a modern search engine and AI agents. They Admittedly make me considerably more efficient than I was before using them. I can use them as if I had my own graduate to complete simple tasks, but I almost always need to complete the job. The AI agents to the reusable grunt work, but I have to do the domain specific stuff.

Final thoughts

All in, I am excited by the release of AI. I see huge privacy and security risks associated with them, and I am glad to see new security standards being launched (ISO 42001) to provide easily obtaining knowledge. I will continue to use AI, much like a builder continues to use power tools and I strongly encourage everyone to use them as well.

Now with that all being said. All hail our robotic overlords, I am a friend not foe.