Artificial Intelligence: Unveiling it to understand and reflect
Image by Fiorella Nario, UI/UX Designer at Onetree
Artificial Intelligence (AI) is changing many aspects of our lives and making us reflect a lot. There are more and more tools emerging every day, there are talks about Artificial Intelligence in relation to other aspects: How does it relate to each discipline? How does it impact and complement the way we work? How does it change the way children learn at school? As many other big advances in history, it challenges the status quo of how we are used to doing many actions in different areas of our lives, so it makes us question many things.
A brief tour over history: from Artificial Intelligence to Generative AI
Artificial intelligence has been present for a long time, by definition: “it is the simulation of human intelligence processes by machines”. It is the broadest family of algorithms which means that it includes from the most simple example of AI, such as a conditional statement or rule (for example: “if “) so that the system makes an automatic decision, to the most complex algorithms which are Generative IA.
Machine Learning, which is part of Artificial Intelligence, are more complex algorithms that learn by examples. In other words, Machine learning algorithms learn by themselves using Databases as examples.
Deep Learning is an even more complex algorithm that is part of the Machine Learning family. These algorithms also learn by themselves as they are also part of Machine Learning.They are also called Artificial Neural networks as they mimic the human brain activity since they consist of many layers of neurons that can even make more complex decisions..
Last but not least, Generative AI are the models that are making most of the front pages. These models are the biggest and most complex ones yet. Generative IA includes ChatGPT, Dall-e-2, Copilot, Bard and many more. These models have been trained with huge databases, for example, ChatGPT database is the whole Internet until September 2021 (this may vary in different versions). This could lead one to ask, does this model have the whole human knowledge? That includes the biases that are spread on the internet.
Dall-e-2 also from OpenIA is one of the first models to generate images, it has been trained with a huge database of images and can create new and realistic images.
Image designed by Fiorella Nario, UX/UI Designer
Last version of ChatGPT incorporates a new feature which is Multimodal, meaning that it can receive more than one type of information, as it can receive text as well as images.
Why are prompts an important tool for the development of AI applications?
A prompt is the input text fed by the user into the Artificial Intelligence to obtain a response. It is the request one makes to the tool, such as asking ChatGPT how plastic decomposes, in bullet points, the prompt would be: “How does plastic decompose? Explain it in concise bullet points”.
- Provide the necessary information for the model to understand the problem to solve and the format and style of the answer.
- Help define the limits of the AI system making it possible for the system to be more precise and effective.
Advantages and Disadvantages of Artificial Intelligence
On the one hand, Artificial Intelligence can help automate repetitive and monotonous tasks, which results in more efficiency and productivity. In addition to this, it can help us reduce human mistakes and improve the quality of our decision-making when working with data. AI models are really powerful in the analysis of data and detect patterns that cannot be visualized by human sight. Moreover, imagine when used in ecommerce, they can allow for automate personalized experiences without increasing the cost. Freeing up workers’ time to dedicate to more creative and more important tasks.
On the other hand, algorithms are as good as the database that had been train with. As every Machine Learning algorithm, it is trained with a database, which can be biased. ChatGPT, for example, collects all available information on the internet to answer, the Internet is full of discriminatory examples. Additionally, it is a threat for the dependence on technology, even though it is not any different from the moment in which cell phones started to have calculators. How many people have lost the ability to mentally solve a division?
The indiscriminate or misuse of AI models can also lead to loss of privacy or leak of data security. Each generative AI tool has its own policy regarding the use of earlier prompts for future re-training of the model. This should specially be taken into consideration when using confidential information. What is more, it may provoke displacement or modification of job positions.
So, what can we reflect about all this?
Artificial Intelligence makes many tasks easier, as it can avoid us doing long and tedious duties. For example, in one of our projects with a supreme audit institution that needed to detect fraud in purchases of public institutions, we used an Unsupervised Model, with Anomaly Detection and an algorithm called Isolation Forest. This algorithm separated all the observations that were different from a specific behavior; it separated what was distinct, not necessarily what was fraud. The anomalies are the observations that are most rapidly isolated; the amount of times that an observation takes to be isolated indicates its anomaly score: the faster an observation is isolated, the more anomalous will the observation be. Instead of having to analyze all the situations of all the public entities, the algorithm helped narrow that number down. This way, experts in fraud detection such as auditors were able to speed up the process, and study, among that smaller number, which ones were mistakes (situations that were only different from the rest) and which ones were fraud situations.
AI gives place to a lot of reflection. As different tools that help us develop tasks in an easier way appear (not only AI, other tools such as calculators or translators as well), we need our cognitive abilities less. In the past, when calculators did not exist, we had to mentally do the math, and therefore use our brain to do so. Nowadays, there are many actions for which we do not have to use our brains, as technology does it for us. Towards this, one can argue that we stopped doing certain tasks, and therefore using our brains for them, but we started doing others and are now using our brains for those. So cognitive abilities would not be lost but changed. In addition to this, some of the tasks we are no longer doing, as the one mentioned in our project’s example, are tedious and monotonous tasks that are not really challenging to our brains. Therefore, one can argue that AI tools help us with monotonous tasks so that we can dedicate time and energy to more complex tasks; even though these are not the only tasks AI helps us with, and tools keep advancing.
What is more, some argue that the advancement of AI tools leads to less human interaction and more isolation. Before AI tools, when we needed help with something, we asked a friend, a co-worker, a family member or a colleague. As AI tools keep appearing and advancing, we need to ask others, and therefore interact with them, less and less. However, others may say that we still interact with others on other occasions, such as social gatherings, or for other purposes.
AI tools have, without doubt, arrived to stay. People use them to work, study, change the way they used to do certain activities, among others. We cannot stop advancements because of the negative consequences advances -in general, not specifically AI- may bring. We cannot deny reality, the fact that they are useful and that people are using them, and their positive aspects. Nevertheless, given the place AI is gaining nowadays and the effects it is having in different aspects of our lives, informing ourselves, learning, and reflecting is enriching. We should not forget we are talking about tools, which we can use to complement and help us. All big advancements challenge the status quo, how will we adapt to this one?