The media is abuzz with news about artificial intelligence (AI) as of late—it’s almost overwhelming. But why the huge spike in interest? AI isn’t exactly new; many businesses and institutions have used AI in some capacity for years. The sudden attention to AI was arguably caused by something called ChatGPT, the first widely publicized AI-powered chatbot that could do what no others could.
The GPT in ChatGPT stands for [translate:Generative Pre-trained Transformer]. GPTs are a type of AI model called a [translate:neural network] that can respond to plain-language questions or requests, and those responses seem like they were written by a human. When ChatGPT and other GPTs like it were released to the public, people could experience firsthand what it was like to have a conversation with a computer. It was surprising. It was eerie. It was evocative. So of course people started paying attention!
An AI that can hold a natural, human-like conversation is clearly different from what we’ve seen in the past. As you learn in the Artificial Intelligence Fundamentals badge, there are many specific tasks that AI models are trained to perform. For example, an AI model can be trained to use market data to predict the optimal selling price for a three-bedroom home. That’s impressive, but that model produces “just” a number. In contrast, some AI models can produce an incredible variety of text, images, and sounds that we’ve never read, seen, or heard before. This kind of AI is known as generative AI. It holds massive potential for change, both in and out of the workplace.
Generative AI might seem like a hot new thing, but researchers have been training generative AI models for decades. Some have even made news in recent years. For example, in 2018, Nvidia unveiled an AI model capable of producing random photorealistic images of human faces. The pictures were surprisingly convincing, entering the public consciousness gradually.
Meanwhile, other AI researchers focused on language tasks. AI models were trained to interpret text in varied ways. For example, categorizing product reviews as positive, negative, or neutral requires understanding how words combine in everyday use — a task called natural language processing (NLP). NLP is a broad AI category encompassing many ways to process language.
Some NLP-based AIs are trained on huge amounts of text data from the internet. These are known as large language models (LLMs). LLMs capture language rules humans take years to learn, enabling advanced language tasks.
Given a sentence, if you understand how words combine to convey meaning, you can rewrite that sentence differently but keep the meaning. AI models similarly remix sentences, condensing long paragraphs into short summaries. Such AI-assisted summarization can create meeting notes or abstracts, serving as an ultimate elevator-pitch generator.
LLMs contain rules for structuring words into ideas. Each language has unique rules—English generally puts adjectives before nouns, whereas French often reverses this order. AI translators learn multiple language sets, enabling them to remix sentences between languages accurately. Programming languages also have rules, allowing AI to translate loose instructions into code, acting as a personal pocket programmer.
Even expert writers make grammatical or spelling mistakes. AI detects and sometimes auto-corrects such errors. It also fills missed words in speech-to-text tasks, greatly improving accuracy for closed captioning in noisy environments.
This task launched generative AI into the spotlight. GPT-like AIs interpret questions or requests and generate extensive text responses. For example, you could ask for a one-sentence summary of Shakespeare’s three most popular works:
- "Romeo and Juliet" - A tragic tale of two young lovers from feuding families whose love leads to their untimely deaths.
- "Hamlet" - The story of a prince haunted by his father’s ghost, grappling with revenge and existential questions.
- "Macbeth" - A drama of ambition and moral decline as a nobleman, spurred by his wife, murders to seize the throne.
Users can continue such conversations for more detailed learning, like talking with a Language Arts teacher. This real-time information retrieval exemplifies AI's utility.
LLMs can operate alongside image generation models, enabling users to describe desired images which AI attempts to create. Relatedly, some AI models can add new content to existing images, extending borders or filling context-based areas.
Similar to text-to-image, AI models convert text to speech, sometimes mimicking unique human speech patterns. To casual listeners, AI-generated speech can be indistinguishable from real voices.
These examples show LLMs’ broad capability for creating new text, images, and sounds. Any language-based task can potentially be augmented by AI, offering powerful tools for work and play.
Generative AI generates text by predicting likely word sequences, not by "thinking." It does not hold opinions, desires, or intentions; it merely predicts the most probable response based on training. For instance, when asked, “Do you prefer coffee or tea?” the AI predicts a response fitting human expectations, without preference itself.
- Trailhead: Artificial Intelligence Fundamentals
- Trailhead: Natural Language Processing Basics
- Salesforce Help: About Einstein Generative AI
Created with ❤️ for understanding the AI surge.