I know what Artifical Intelligence is...

I've given a few talks now on the role of Artifical Intelligence (AI) in the creative industries, focusing on IP and ownership law as well as some of the ethical quandaries around its deployment in visual arts. I spend a good 5-10 minutes of each of these discussions explaining what AI actually is to folks, so I figured a quicker route would be for me to write this down for future reference.

What is AI?

What is AI? Well it doesn't look this

Artificial Intelligence (AI) refers to the simulation of human intelligence in machines designed to think and learn like humans. This involves creating systems that can perform tasks typically requiring human intelligence, such as recognising speech, making decisions, solving problems, and translating languages.

All of AI's decision-making, analysis, and improvements (also known as learning) are based on algorithms and vast amounts of data to identify patterns and then make predictions. Over time, based on how correct the answer is, AI can improve through experience.

Probably one of the most famous uses of AI (coincidentally the first use of the term) was to play an online game, checkers. Three years after this invention, the term 'artificial intelligence' was used during a seminar at Dartmouth. This form of AI is a relatively simple binary algorithm of cause and effect, i.e.: if red goes here, black goes there. The main advance is that it is preemptive, meaning it knows if red goes to D2, black should go to E2 because in 5 moves, it can win, etc. This is a different, arguably more basic, form of behavioural AI we have today, which could function to know this player likes a certain move, so it will predict it and preemptively act.

Have I used AI?

Yes, yes you have. If you use Waze, Google Maps, Apple Maps (not going to lie, follow Apple Maps and you may end up in a lake...), these all use AI to predict your best route. Ever got vouchers or coupons from a supermarket for your favourite brand? That's AI. Both are using behavioural data to emulate human intelligence. Just as a friend would give you driving instructions based on the time of day, whether it is a weekend, nice weather, a directional app does the same. The pro of the app is that it also has access to live data: lots of cars on the M25? Fine, take the North Orbital instead. What the AI still will mostly not know unless it learns would be, say, the urgency of your journey. Do you need to go through town to get petrol? Unless you tell it, AI won't know.

AI actually looks a lot more lke this

In more advanced fields, AI now has use in numerous industries. In catering and hospitality, AI can be used to do your daily fresh food orders and even staff scheduling based on the number of bookings, but also if it is a sunny day or a holiday. In healthcare, AI has the potential to revolutionise major global issues such as cancer care. A survey from Wisconsin in 2023 found it to be 98.3% accurate in diagnosing cancer. How does a machine with no concept of human health or what cancer is even do this? Data. The AI can look at medical records, lab results, lifestyle data, and so on to create a picture based on probability.

Preemptive vs Predictive AI

The cognitive function of AI can be divided into two processes: Preemptive and Predictive AI. Preemptive works off pattern recognition to give a result or take an action based on the data or previous actions taken. This would be something like our checkers or chess example, where an action or prompt solicits a series of defined responses towards a goal.

Predictive AI is more theoretical and uses a series of data, suppositions, and contextual points to give a result or take action. Back to our 21st-century restaurant example, AI could predict the weather (28 and sunny), look at last year's orders, and predict some behaviours, so order more ice and so on.

Generative AI

Generative AI is the scary and sexy new buzzword that a lot of the creative sector is concerned about. Whereas playing chess online results in a simple response and sequence, Generative AI can create images, videos, and music based on a series of prompts or asks. In principle, it is no different from our restaurant ordering system or cancer-diagnosing AI; here the data are the prompts and guidance you teach the AI.

This AI is more complex as it actually creates a long-lasting output that is a tangible thing. The law around Generative AI ownership, regulation, and IP is a mess and, for the most part, non-existent (I'll cover this in another blog as it is a hefty topic), but this doesn't mean it isn't something we should be exploring and improving.

That's enough for a chill Friday morning. The main thing to take away is: AI has been around for 70+ years, we use it every day whether conscious of it or not, the priority now is the regulation of AI rather than its deployment.