Text-based generative AI

- Using generative AI to help draft and write new content is revolutionizing entire industries, as well as enhancing personal and professional workflows everywhere. Let's drop back and talk about how this works at a high level. Generative AI language-processing technologies, large language models like GPT-3 and GPT-4, learn on massive training data sets. In this case, it's text and code from the internet, from articles, books, webpages, Wikipedia, and more, many, many billions of words. Then the large language models learn by using probability to guess the next words of text and answer to a prompt. During initial testing phases, people supervise the process. As the AI model provides answers to prompts, these human supervisors teach it correct answers, helping to build its knowledge. Now, subsequent training phases ask the AI model to offer multiple answers to a prompt, and then human supervisors rank the answers from best to worst. Each of these training phases is important in teaching prediction, comparisons, relationships and more. Over time, the AI model improves its understanding of prompts and increases the likelihood that it will produce good, correct, well-rounded responses using human-like text. ChatGPT is an AI chatbot built on GPT technology. Drawing from its vast knowledge base, it communicates and generates human-like text in response to a wide range of prompts and questions. Here's a summary of just some of the high-level things ChatGPT can do. That is a lot of power. Let's get more specific for a moment and assume the role of a small business owner creating various social media materials. I can use ChatGPT to ideate a collection of potential marketing campaign themes, draft a cross section of social media copy for different social media platforms, translate my marketing copy to other languages, write the first draft of a tone-specific article that I can then modify and make my own, and so on and so forth. If we assume the role of an academic, Chat GPT can help me summarize all sorts of dense research content, ideate and generate new research ideas, generate interview questions on any given topic, take a first pass on grading the most basic aspects of student work, allowing me to dive deeper with my own personal evaluation, and so on and so on. The possibilities are truly endless across a wide variety of roles and professions. Now realize that these prompts are often carefully crafted and honed over several trial and errors. Arriving at the best nuanced wording for AI prompts is an art in itself. It's called prompt engineering and it's a growing and important part of successful generative AI strategy. There are many resources available on the web to help you write better prompts. So far, we've just been using ChatGPT, which was created by OpenAI, a research laboratory whose mission is to ensure that artificial general intelligence benefits all of humanity. Other companies have also integrated GPT technology including Microsoft when they created Bing Chat. One main difference is that while ChatGPT is currently limited by its training data, which at the time of this recording has an upper limit of data until September 2021, Bing Chat has access to the entire internet. This means it can provide answers related to current events and it also lets you confirm your sources, a huge plus to ensure accurate results. It also provides various fine-tuning controls. For example, it offers different response modes, creative, precise, and balanced. Creative mode offers responses that are more original and imaginative. Precise mode offers responses that are more factually accurate, and balance attempts to provide a nice mix between the two. Google Bard is a similar chatbot, but it uses a different large language model called Lambda. Like Bing Chat, it leverages the power of the internet. More specifically, Google's search engine. As of this recording, Google Bard is still in beta and will undoubtedly evolve rapidly, but one valuable feature it offers right out of the gate is to automatically provide multiple drafts of responses. These big three, ChatGPT, Bing Chat, and Google Bard, are currently the most powerful text-based generative AI tools and serve broad applications across many industries, but there are also AI solutions that have been fine-tuned to cater to specific sectors. For example, AI-powered writing assistance like Jasper AI also use GPT technology, but have been fine-tuned for a specific business and marketing applications and customer bases, resulting in optimized and SEO-friendly content. Here's a list of other powerful AI writing assistance that are likewise optimized for different types of business use cases. Most have template libraries, grammar checkers, plagiarism checkers and more. Some are free to use, others require a subscription. Depending on your industry, some may be more helpful and optimized than others. As with all generative AI, the goals of these tools aren't necessarily to totally replace human content creation altogether, rather, they aim to enhance the creative process so people don't start from a blank slate. By giving creators an informed, optimized, fleshed out starting point, they offer greater and deeper opportunities to research, ideate, strategize, and create.