Over the past few years, artificial intelligence (AI) has become the center of world attention, and this rapidly developing technology is often a source of anxiety and even fear. Afraid in some cases. However, the evolution of AI doesn’t have to be a scary thing — and there are many ways that this new technology can be used for the benefit of humanity.
Writing in “AI for Good” (Wiley, 2024), Juan M. Lavista Ferres And William B. Sundayboth senior directors at Microsoft’s AI for Good Research Lab, reveal how AI is being used in dozens of projects around the world today. They explain how AI can improve society by, for example, being used in sustainability projects like using satellites to monitor whales from space, or by mapping glacial lakes. AI can also be used in the aftermath of natural disasters, like the devastating 2023 earthquake in Turkey, or for social good, like curbing the spread of misinformation online. There are also significant health benefits to be gained from AI, including studying the long-term effects of COVID-19, using AI to manage pancreatic cysts, or detecting leprosy in vulnerable populations.
In this excerpt, the authors detail the recent emergence of large language models (LLMs) such as ChatGPT or Claude 3 and how they have risen to prominence in today’s AI landscape. They also discuss how these systems have had a significant beneficial impact on the world.
The emergence of language models
At the heart of current linguistic technologies, such as GPT, lies the concept of a language model. Imagine you start a sentence with “This morning I woke up and saw a beautiful blue _____.” What should happen? The language model predicts the continuation based on probabilities derived from large amounts of text data. For example, words like “sky” might be a very likely continuation. However, the sophistication of the model allows it to consider multiple possibilities, such as “bird” or “car,” each with a certain probability, demonstrating its nuanced understanding of different contexts.
Despite its limitations, the impact of LLMs has been remarkable, especially in late 2023. For example, GPT-4 has achieved significant milestones, such as passing multiple-choice and written tests. The real power of these models lies in their ability to learn from a vast source of information: the World Wide Web. This vast resource contains the vast majority of our collective human knowledge and is by far the most important dataset in the world. By training on this vast dataset, LLMs can build representations of the world that replicate the complex relationships found in human understanding.
Related: 32 times artificial intelligence made huge mistakes
In short, while these advanced LLMs lack real understanding or consciousness, their capacity to process and mimic human language and thought is remarkable. As this technology continues to develop and refine, it is important to understand its capabilities, limitations, and ethical implications.
LLM as a language tool
When we discuss LLM, most of the discussion revolves around the power that AI has in areas such as healthcare diagnostics, but an area that is not often discussed is the power that LLM has as a language aid in helping non-native speakers write fluently.
This opportunity really touched my heart. I am a non-native English speaker leading a research lab with over 70% of its members from the global south and also non-native English speakers. About 95% of research is published in English, yet only 4.7% of the global population is native English speakers. When I give talks in Uruguay, my home country, I always emphasize the importance of English and coding. I was fortunate, my parents made sure I learned English from a young age. However, many smart people don’t have that opportunity.
With GPT, the ability to write confidently in English is now within reach for everyone. LLM, like GPT, is not a panacea, but it has the potential to bridge the language gap in a remarkable way. A good translation tool does more than provide a literal translation of words between two languages. It must also convey meaning, tone, cultural connotation, and context.
LLM works by refining text that may not be well-structured, transforming it into native-like expressions. This is especially important for non-native speakers who often struggle with the nuances of English grammar and syntax. The model helps not only by ensuring grammatical accuracy, but also by improving vocabulary to match the quality of native English publications.
LLM for democratizing coding
I consider myself fortunate in many aspects of my life, especially since my parents gave my brother and I a computer when I was eight years old. This early experience gave me an invaluable opportunity to learn to code. The profound impact coding has had on my life is undeniable. However, among my friends and classmates, I was the only one who had this privilege. More than three decades later, less than 0.5% of the world’s population knows how to code.
Learning to code is like mastering a new language, which serves as our interface to programming computers. While there are significant positive impacts from more people learning to code, it is difficult to foresee a drastic increase in the number of coders in the coming decades.
However, LLM advertising can bring about substantial change. Advanced systems like GPT-4 have the ability to translate natural language into real programming languages. These models empower people to write programs and automate processes in their native language, be it English, Spanish, Mandarin, or others. This technology has the potential to democratize programming, expanding its reach to hundreds of millions of people around the world and bridging the gap between those who can code and those who cannot.
LLMS in fields such as medicine
In April 2023, John W. Ayers and colleagues published a study in JAMA International Medicine comparing doctors’ responses to GPT-4’s responses to patient questions. The study found that GPT-4 not only provided more accurate answers than doctors, but also showed greater empathy.
It is worth noting that GPT-4 has not been extensively trained on most medical knowledge, much of which is still paid. Furthermore, the model was not specifically trained for medical scenarios. Despite these limitations, its impressive performance highlights the potential impact of the model.
Today, about 4 billion people — nearly half the world’s population — do not have access to a doctor. While medical access has improved in recent decades, especially in the global south, the challenges remain significant.
These AI models aren’t ready to replace doctors. However, if they can provide accurate responses to human questions, they could free up doctors to focus on their areas of expertise. While we haven’t seen these models deployed in a production environment for medical consultations, the promising results point to a path forward in addressing global healthcare disparities.
This excerpt has been edited for style and length. Reprinted with permission from “AI for Good: Applications in Sustainability, Humanitarian Action, and Health” by Juan M. Lavista Ferres and William B. Weeks, published by Wiley. © 2024 by Juan M. Lavista Ferres and William B. Weeks. All rights reserved.