AI: Artificial Intelligence Explained the non-AI background way

The most popular term of 2023 has to be Artificial Intelligence or AI. This term has brought both fear and excitement to the world as this technology - which has existed for decades - is now at the forefront of the Fourth Industrial Revolution or as I like to call it, the digital era. In this article, I explain the components of AI with a focus on Generative AI and Large Language Models (LLMs) which have the highest number of users globally through platforms such as OpenAI ‘s ChatGPT and 谷歌 ’s Bard.

What is Artificial Intelligence (AI)?

Artificial Intelligence is an academic discipline and a field of study (think of the depth of subject matter in Biology or Psychology). There are areas of specialisation and sub-division within the field. Machine Learning is a sub-field of AI and within it includes Deep Learning. Deep learning is therefore a subset of AI that mimics the human brain's neural networks. It processes vast amounts of data to learn patterns and make decisions. Imagine teaching a computer to recognise cats: deep learning enables it to independently identify feline features in diverse images, transforming how machines understand and interpret information, much like how we learn to recognise objects through repeated exposure.

Artificial Intelligence -> Machine Learning -> Deep Learning        

Two fields within the deep learning category include Generative AI and Large Language Models.

Generative AI

Generative AI produces new outputs that are similar to the data that was used to train the model. To determine if something is Generative AI then the output is natural language text or speech, an audio or an image. It has various model types including the following:

• Text-to-text where text input is given and text output is provided (eg. ChatGPT and Bard)
• Text-to-task where a model is trained to perform a specific action based on the input data (eg. asking Google’s Bard to summarise the last document you saved in google drive)
• Text-to-image where a model is trained on a large set of images and I can then produce new images (eg. Midjourney, DALL-E by OpenAI)
• Text-to-video where a model can produce and edit videos (eg. Meta AI Make A Video, CogVideo)
• Text-to-3D where text input can be used to generate gaming assets (e.g. SHAP-E by OpenAI)

Large Language Models (LLMs)

LLMs are models that are pre-trained with a large set of data and then fine tuned for specific purposes and use cases. A pre-trained model would be regarded as a generalist and a fine tuned model would be considered a specialist in a particular field. These models are usually pre-trained with one or more of the following capabilities:

1. Text classification
2. Text generation
3. Document Summarisation
4. Question Answering

Once the appropriate LLM has been selected, it can then be fine tuned to any particular subject matter using industry specific datasets such as those from banking, healthcare, mining and media to name a few use cases. Y Combinator’s last cohort (S23) had a significant AI presence of startups with fine tuned solutions.

examples

• Elythea (YC S23) - Preventing maternal mortality with machine learning
• Synaptiq (YC S23) - AI-powered medical training that boosts training and learning speed for Doctors
• Nowadays (YC S23) - AI co-pilot for event planning

Why should you learn more about AI?

Reading and learning more about AI helps in the following ways:

• Improves your understanding of the developments in the field
• Optimises your use of AI tools and platforms
• Fosters creativity to build AI aligned projects
• Makes you future-ready for changes in the workforce and daily life

The AI industry is growing fast and it is here to stay. It presents a tremendous number of opportunities for first-time and experienced entrepreneurs as well as corporates, development organisations, institutions and individuals looking to enhance outcomes in any field of interest. As various advancements are built globally, one of the best methods to adopt through the emergence and growth of AI is to keep learning and using existing tools to familiarise with its power and potential. The digital era is getting exciting and we all have an opportunity to contribute to this important aspect of human development.