Any intelligence demonstrated by machines, in contrast to the natural intelligence displayed by humans and other animals. These machines use mathematical models that identify and encode patterns in data sets, which can then perform predictions on new situations which they haven’t encountered before.
Why It Matters:
In 2023, AI matters significantly because it is rapidly integrating into our daily lives, revolutionizing industries, and altering the way we work, learn, and communicate. AI technologies facilitate data-driven decision-making, automate routine tasks, and contribute to unprecedented levels of efficiency and productivity. This swift technological evolution underscores the need to integrate AI education into curricula, not only to prepare students for academic success but also to meet workforce demands. Crucially, attention must be given to underrepresented populations, ensuring inclusivity in accessing and benefiting from the latest technological advancements.
Explore More:
Refers to a class of AI models and algorithms that can generate new content, such as text, images, music, or other data, based on patterns and examples from existing data. It includes various techniques like GANs (Generative Adversarial Networks) and language models.
Why it Matters:
Its ability to generate novel content, automate complex tasks, and optimize processes holds the potential to revolutionize industries, drive efficiency, and foster continuous advancements in technology and problem-solving. In essence, Generative AI stands as a cornerstone in pushing the boundaries of what's possible, making it a critical force in the evolution of AI applications across diverse fields.
Explore More:
ChatGPT (short for Chat Generative Pre-Trained Transformer) is a series of generative AI chatbots launched by OpenAI in November 2022. It is a Large Language Model that produces a body of unique text from a user’s specific input based on existing content from the internet. The latest version of ChatGPT is multimodal and can recognize images, generate images, engage in voice conversations, and search the internet in real-time through the same interface.
Why it Matters:
Trained on a vast 45-terabyte dataset, it excels in comprehending and generating human-like language. Its standout feature lies in consistently delivering coherent and contextually relevant responses to open-ended queries, making it a powerful and widely utilized tool.
Chat GPT (AI) Capabilities:
Explore More:
The user interface for most non-AI computer programs is a mouse, keyboard, or touch screen. Chatbots provide a different kind of user interface for AI systems, one that uses speech (either spoken or typed). These AI programs range in sophistication from relatively simple and rule-based (e.g., providing a canned response to a specific question) to more complex and AI-enabled (able to parse human language and learn from previous conversations to improve accuracy constantly).
Why It Matters:
It is easy to build a simple chatbot, but complex to build a genuine AI chatbot — which is why the arrival of ChatGPT has taken the world by storm. Because they can respond to a nearly limitless number of users at once, chatbots have the potential to provide real-time support at unprecedented scale — which, in the context of higher education, is helping institutions boost enrollment and student success while enabling advisors to focus on students who need more hands-on, personalized guidance.
A type of machine learning model that can perform a variety of natural language processing (NLP) tasks such as generating and classifying text, answering questions in a conversational manner, and translating text from one language to another.
Why it Matters:
Unlike smaller-scale language automation algorithms, LLMs, exemplified by models like GPT-3, have the potential to revolutionize communication by producing more sophisticated and human-like outputs. Their adaptability to tasks beyond explicit training sets them apart, showcasing their capability to transform how we communicate and engage with technology and information.
Explore More:
A subfield of AI that uses algorithms trained on data sets to create self-learning models that are capable of predicting outcomes and classifying information without human intervention.
Why It Matters:
Machine learning is used today for a wide range of commercial purposes, including suggesting products to consumers based on their past purchases, predicting stock market fluctuations, and translating text from one language to another.
Explore More:
OpenAI - The earliest mover with the viral web front-end and the name ChatGPT. The free version uses GPT 3.5 as the model, which is suitable for most tasks, though there are significant improvements in the latest model GPT4, particularly in the area of reasoning and avoiding hallucinations (think of AI as an intern). An iPhone app was released in mid-May 2024 with a phased launch across the world and they are going to launch an Android app soon. If you sign up for the paid plan (currently $20 a month), you get early access to features such as the ability for the bot to access the web and use plugins that supercharge the functionality of the bot. The Code Interpreter feature is in alpha currently and promises to be significantly good at analysis.
Google Bard - A free interface available to everyone with a Google account. Does reasonably well on most tasks and has access to the web by default. Currently, does not match up to the latest models from OpenAI, but it might improve fast given Google's focus on AI now. A feature similar to plugins, is being labeled "tools" and was announced in May 2023. Some AI capabilities are already integrated into Google tools.
Microsoft Bing Chat - A free interface available to all. Powered by OpenAI models. Currently requires downloading the Microsoft Edge browser to access the chat feature. Has default access to the web and is powered by the latest models from OpenAI, so for normal usage this could be considered a free version, versus the paid version from OpenAI. Microsoft has also enabled plugins for several services and will continue to add more. They have also committed to following the same standard for plugins as OpenAI, so the plugins developed once will work across both interfaces. Given that Microsoft is currently offering OpenAI models for free, and adding plugins support, it seems the best free option, and it includes references as well.
ChatPDF - One specific use case that might be very relevant for most users is to be able to 'chat' with a PDF document. This might be a research article, a printout of a web page (saved as .pdf), or a PDF of a book. It is a very simple and elegant solution. There are limitations on how much you can use if for free. This feature will be subsumed in one of the other platforms soon, but for now, it is an excellent tool.
Perplexity - If you are looking for a response from a model that includes an aggregation of search and a generative model and present it with references, this is a good option. Microsoft Bing chat also includes references in its response, so this is not specific to Perplexity, though the experience is better with this tool.
Adobe Firefly - Generative AI can be used for images, audio, and video. Adobe Firefly (beta) is currently available via university license and lets you quickly create graphics. Two powerful features include: "text to image" to generate images from a detailed text description and "generative fill" which makes it easy to remove objects or paint in new ones by supplying text descriptions. Other features include text effects, generative recolor, 3D to image, and extending images.
For a full searchable list of generative AI tools, check out the AI Scout Directory.
Every single recent breakthrough in ML and AI is a result of deep learning with neural networks. Deep learning is a subset of Machine Learning which employs networks capable of learning from data that is unstructured or unlabeled. This type of learning uses neural networks to extract increasingly subtle and complex patterns, allowing for more sophisticated tools like accurate speech and facial recognition.
AI is based on the idea of neural networks -- essentially artificial brains built from silicon. These networks provide a structural, layered approach to processing data, based on the way the human brain works: each layer of processing (made up of artificial “neurons”) provides the input for the next layer. Through clustering and classification, these layers recognize patterns that can, for example, differentiate a photo of a dog from a cat. It’s worth noting that even the most cutting-edge ML experts do not fully understand what is encoded in an individual neuron of a neural net.
A set of instructions or computations that a machine follows in order to learn how to do a particular task.
A machine learning technique where a model trained for one task is adapted for a related task. Many generative AI models use transfer learning to leverage knowledge from large-scale pretraining before fine-tuning for specific educational applications.
The initial input text or instructions given to a model to generate new content based on that starting point. It provides context and guides the model's output. The prompt can be a few words or sentences that set the tone or specify the desired content.
Explore More:
Misinformation or made-up information based on a pattern that the AI model has learned as part of its training. For example, the model could create references that do not actually exist.
Explore More:
Prompt engineering in AI is the organized creation, improvement, and fine-tuning of instructions for Generative AI systems. It helps AI produce desired outcomes and promotes smooth communication between humans and AI. This practice involves continually assessing and categorizing instructions to keep them relevant and effective.
Explore More: