Sorry, But A.I. Doesn't Exist.
This isn't a clickbait title. It's actually true. There is no such thing as Artificial Intelligence, but there are some awesome tools!
It’s no secret that there’s a lot of hype around Artificial Intelligence (AI) right now, that there’s a lot of hyperbole, exaggerations and even some blatant lying going on. Even by Google and it’s release of Gemini. Yet, A.I. tools may be some of the most important technologies we’ve developed in a long time.
But Artificial Intelligence in and of itself, does not exist. There is no Artificial Intelligence. None. Zero. Zilch. Nada. It’s just not a thing. What does this then mean? What do I mean by A.I. doesn’t exist?
If you’re one to worry about the arrival of The Terminator and AI taking over the world, this article should help you put aside those fears. We may never actually reach what is termed as Artificial General Intelligence (AGI).
If you’re also worried about your job being lost, so far the only job being lost to AI is Sam Altman, CEO of OpenAI. So there’s that.
The words Artificial Intelligence (AI) are best understood as a branding term, another name for a toolbox of various technologies. But there is no singular AI. You might wander off into the digital forest hunting for it like pigs for truffles, but it’s not there.
Ask any computer scientist and if they’re honest, they too will admit this is true. I had occasion several years ago to work with the Canadian computer scientist Dr. Guy Lapalme at the Université de Montréal, who’s area of specialty is Natural Language Processing (NLP), a tool under the umbrella term of AI. He never quite felt comfortable calling it AI and like other such scientists I worked on projects with, often winced at the term. They just used it to help them get funding.
In the toolbox of AI are a bunch of excellent technologies that tend to do one thing very well. What is termed as Narrow AI. Every tool in the AI toolbox today is Narrow. They are most often used singularly to do a particular job.
You can’t use a screwdriver to build a bookshelf. You select a set of tools to do that job. In this way, various technologies in the AI toolbox are often combined. But there’s no way, yet, to combine all of them to do anything.
If you use Google Workspace or Microsoft365, they are a suite of tools all wrapped up into a platform. Like the AI toolbox, you may use them singularly, like Excel to do a financial forecast, but you can’t do a financial forecast in a PowerPoint deck. Yet they can be interconnected, such as pulling a table into a Word document or PowerPoint. But they are always separate.
Much of the hype today is around Large Language Models (LLMs) or what is termed Generative AI (GAI.) A tool which is very good at predicting what comes next. While LLMs are a very useful tool, they have limitations as well and are still Narrow AI.
LLMs tend to hallucinate, serve up wrong information, especially when dealing with large amounts of data. They have no understanding of the nuance of human emotions, cultures or societies. They understand patterns, but they are not creative. They’re often contradictory, have biases and don’t understand aesthetics.
Back to my work with Dr. Guy Lapalme. The project I worked on with him was around translation using Machine Learning (ML) and Natural Language Processing (NLP). My company at the time was analysing social media in fragile democracies to advise foreign policy in Western democracies.
As Tunisia had just gone through an election, we acquired a set of public data via Twitter. This complied with Twitter’s terms and conditions. No personal data was collected. We followed strict ethical rules from the start. We just wanted to try translation.
What threw us of course very quickly was that people, in one tweet, would use a mash up of hashtags, numerals and alphabet along with mixing up English, Arabic and French. One way to do this was mixing numbers and letters in a word. When you looked at it, you would see French, when you processed it in your mind, you heard Arabic. Very clever. There was simply no way to translate this. Project terminated.
All this to say, the various tools in the AI toolbox are excellent and improving rapidly. Applied in various combinations, they have already helped humanity and they will continue to do so as they improve. But it’s important to put these tools into perspective so we can cut through the hype.
If you’re wondering why there is so much hype? A number of media services have reported on this topic. Mainly it is because it helps companies like OpenAI, Anthropic and similar to raise a lot of capital, making those shareholders some nice cash. For Microsoft, Google and Amazon, it helps improve share valuations which gives executives mice bonuses. So yes, it’s all about the money and not much else.
When we understand that there is no singular Artificial Intelligence, that they can’t all be herded together to make one singular, well, thing, then we can be more practical and realistic at how we can approach regulating, governing and innovating these technologies.
One of my concerns, when I speak with people in various levels of government, is that they don’t understand that there is no AI. Once I explain this, lightbulbs go off. When they and lawmakers better understand the reality of what AI is and is not, better laws and regulations can be made.
Socioculturally, we have given a fair bit of cultural agency to AI, while not understanding it. Both the companies who create these AI tools, especially LLMs, and the ones that want to implement them in their products, prefer the market doesn’t understand the difference. One suspects however, that such learning is underway as we overcome our fears and get curious.
And should machines someday, be able to “think”, they will never think like humans because they are machines. But that’s another article.