Generative AI are algorithms such as ChatGPT, DALL-E, and Midjourney, which are used to generate content such as audio, images, text, or code based on a user-given prompt (McKinsey).
Generally speaking, these types of algorithms are “trained” on a wide database of pre-existing pieces of content which it analyzes for patterns and commonalities (McKinsey). Once the user gives the AI a prompt it then defers to this database and based on what it knows pre-existing content, generates a new piece of content that fits what it recognizes to be the patterns and parameters of the given prompt (McKinsey). For instance, if a user asks DALL-E to generate a movie poster for a hypothetical Star Wars movie directed by Tim Burton, it will look at pre-existing examples of movie posters, Star Wars content, and other content made by Tim Burton to essentially predict and replicate what this hypothetical scenario would look like.
Absolutely. A technology as vast and complex as Generative AI will very quickly have effects on almost all areas of human life (Mihajilovic). Even if you don’t personally use it, it will likely be used in some way that is adjacent to your career whether that's for generating creative output or for generating more efficient factory floor plans and influencing your career (Mihajilovic). AI will also affect the entertainment you consume, as it is already being considered for extensive use in Hollywood as well as the games industry (Sammer). AI will be on the forefront of managing finance and the economy, it will be used to educate you or your children, and more than likely it will be used to either protect or steal your personal information online (Mineo, Ravichandran, UNESCO). No matter who you are, Generative AI will have an impact on your life.
Despite the extraordinarily complex nature of the technology supporting it, Generative AI is actually fairly simple to operate for the average user (Harvard Business Review). Simply go to the corresponding webpage or download the software for the AI model of your choice, type the prompt you desire the AI to generate into its input section, and watch as the AI creates your desired output (Harvard Business Review). If you are unhappy with the end result, some models will allow you to ask the AI to make modifications, or you can alter the original prompt to get something remover to what you originally had in mind (Harvard Business Review).
The effectiveness and intelligence of AI is rapidly increasing almost all the time, but currently it is still entirely bound by content that has been used to train it and prompts that are given to it by a user (Harvard Business Review). At this time, AI is not capable of generating original content and essentially functions as a prediction algorithm simply replicating pieces of what it has previously seen (Harvard Business Review). If AI is asked to generate an image of a “duck”, it does not actually grasp the meaning or significance of a “duck” but rather replicates and image of a “duck” based upon patterns present in other images that it knows are supposedly of “ducks” (Harvard Business Review).
AI is not all-knowing, and does possess a number of limitations (Lawton). AI does not always cite the sources it pulls from to generate content (though sometimes it can if asked to do so in the prompt), it has little understanding for surrounding context which can sometimes result in content that feels uncanny or unnatural, and it has no understanding human bias or malevolence (Lawton). AI like ChatGPT has had to have restrictions placed in its code to prevent it from providing information that could potentially be used to cause harm, though loopholes still exist to obtain this information (Lawton).
Popular AI models for text generation include ChatGPT, Jasper, and Lex; DALL-E and Midjourney are often used for image generation; Podcast.ai and Listnr are commonly used for voice synthesis, and Codestarter and Codex provide code generation tools (Lawton).