Understanding AI: Opportunities and Challenges Ahead

In the fourth episode of this series, we want to take a look at the hottest topic around: Artificial Intelligence (AI). In this article we are going to look briefly what AI is doing for us today and focus particularly on where we see us, as humans, fitting in.

What is AI

AI, or Artificial intelligence, is a broad term that means many things to many people; In the modern world today, it can be seen in large language models, like ChatGPT, driverless cars and artificially generated images. But the term AI covers much more than these and has been around for a lot longer than most people think. The groundwork for AI was made in the 1900 mathematically, but really took shape in the 1950 with the likes of Alan Turin who coined the Imitation Game or Turing Test.

There are also those in the tech world who try to convince tech investors that data processing and complex algorithms are AI, but this is an exaggeration. AI really refers to deep learning that use nodes based trees to perform answers to new inputs given previous outputs. AI models are given huge data sets of correct information to help them ‘learn’ and this helps them predict future answers.

A simple of example of this is an AI model for selecting if an image input is of a cat or a dog. First, the model would be given a massive library of correct answers (labelled pictures of dogs and cats), and from that it can then start to build the model of what the differences between dogs and cats really are from a pixel sense. Then when new images are fed in, it can try to guess with a percentage of predictability whether the new images are of dogs or cats.

Is that it’s very hard to predict why the model chose something to be a cat or a dog and this can make it very difficult to reverse engineer. The other problem is there are no ways to validate new answers and retrain the model from them as it is only really guessing whether something is a cat or a dog. If the model was to feed its new answers to rebuild its AI model nodes then errors will appear with increasing frequency.

Chat GPT

ChatGPT is what is referred to as a Large Language Model (LLM) from OpenAI. It has digested the history of the internet (quite literally) and using that breadth of knowledge to guess what the next letter or word will be for a given input and the history of the conversation. When ChatGPT is visually typing, this isn’t just some UI gimmick; it really has to process the letters in real time and doesn’t know what it’s going to say before it says it.

The Imitation Game was a game devised to test whether we can reliable check whether an individual we are talking to is human or an LLM. In 2024, ChatGPT was able to pass the Imitation Game and come across as human 54% of the time. Wiki

Driverless Cars

Driverless cars, like those from Tesla, are increasingly being able to operate complex roads and driving situations. While not yet perfect, these cars are showing the clear trajectory that driverless cars will be commonplace on our roads very soon. The main thing these vehicles struggle with is complex roundabouts and areas where road markings are not very clear. They are good at following the rules, but not so good at choosing to break them when the situation arises (driving past a broken down car, going on the embankment to avoid potholes etc).

Generated images

Another big use of AI is generating images and video. Cool services like dall-e from OpenAI are capable of creating extremely impressive images given the right prompt.

Take this prompt for example: “draw me an image of a mythical creature with giant claws and a large beak, fighting a robot with 1 arm “

AI Trajectory

It is undoubtedly the case that these services are only going to improve, however there is some argument over exactly how much and how quickly. The theoretical goal is an Artificial General Intelligence (AGI) system that would be capable of learning new tricks on its own but it’s still unclear if LLM’s and the current format of AI will ever actually get us there.

Regardless, we can expect that AI will be playing an increasingly large role in our lives and will start to cause massive disruption to people in almost all jobs.

As the world increasingly relies on AI for a boost in productivity, we will most likely see a drop in the number of people required to provide solutions such as development. Where a studio would need 4 engineers to create something, they can likely do this now with 3. This is something we’ve already experienced in other aspects of our lives, such as the slowly increasing number of self service tills in supermarkets or McDonald’s. These tills can operate without salaries, coffee breaks and sick days which usually more than offsets the cost of purchase, maintenance and the increase in thefts that take place with these tills.

Working Together

Therefore, working together, AI and humans can create truly amazing things. AI can take the pressure off repetitive tasks, while humans can focus on different and unique challenges while still having immense productivity benefits.

As a fun little experiment, all the images in this article were generated by AI – but, as will always be the case, the wording of the article is authentically human.

Contact

We hope you enjoyed reading this blog post. If there are any other topics you would like us to cover or you’d like to chat about any of these topics, please fill in the details below:

At Cog Mobile, we specialise in building high-quality apps and websites for our partners. With over 12 years of industry experience, we cater to organisations of all sizes.