What is Generative Artificial Intelligence (AI)? All you need to know
Generative AI is an artificial intelligence technology that can generate a variety of content, such as text, picture, audio, and synthetic data. The contemporary interest in generative AI has been fueled by the ease with which new user interfaces may generate high-quality text, drawings, and movies in seconds.
It is important to understand that the technology is not new. Generative AI was first used in chatbots in the 1960s. However, it wasn't until 2014, when generative adversarial networks, or GANs, a form of machine learning algorithm, were introduced, that generative AI could produce impressively authentic photos, videos, and audio of actual people.
It is important to understand that the technology is not new. Generative AI was first used in chatbots in the 1960s. However, it wasn't until 2014, when generative adversarial networks, or GANs, a form of machine learning algorithm, were introduced, that generative AI could produce impressively authentic photos, videos, and audio of actual people.
On the one hand, this newly discovered talent has created prospects for better movie dubbing and rich educational information. It also raised concerns about deepfakes (digitally created photos or movies) and damaging cybersecurity attacks on enterprises, such as fraudulent demands that realistically resemble an employee's supervisor.
Generative AI begins with a prompt, which can take the form of text, an image, a video, a design, musical notes, or any other input that the AI system can handle. Following the request, various AI algorithms return new content. Essays, problem solutions, and convincing fakes made from a person's photos or speech are all examples of content.
Early iterations of generative AI necessitated the submission of data via an API or another complex mechanism. Developers have to become familiar with specialised tools and create apps in languages like Python.
Now, pioneers in generative AI are creating better user experiences that allow you to articulate a request in plain English. After an initial answer, you can additionally customise the results by providing comments.
Early iterations of generative AI necessitated the submission of data via an API or another complex mechanism. Developers have to become familiar with specialised tools and create apps in languages like Python.
Now, pioneers in generative AI are creating better user experiences that allow you to articulate a request in plain English. After an initial answer, you can additionally customise the results by providing comments.
Approach to AI best describes the following use case
- Supervised learning
The following examples are probabilistic, machine-learning approaches to AI
- A user unlocks her iPhone with her face
- A webpage detects the image of a cat in a picture
Customers contact a company using emails or a form on a website. In order to work on them, these customer tickets need to be correctly classified, e.g. "product return", "service request", etc. the following statements are true.
- Machines can use AI to propose ticket classifications to human service agents, but not with 100% accuracy.
What is a “prompt”?
AI Prompt Writing (or Engineering) is the process of creating input (usually text) instructing the Generative AI to generate the desired response.- Instructions or cues sent to a generative AI model in order to produce the desired output
The characteristics that make foundation models a new approach to artificial intelligence
Foundation models, also known as generative AI models, are large-scale AI models trained on massive datasets to perform various tasks. They're designed to understand context, decipher nuances, and adapt to different scenarios.
- Foundation models are trained using a self-supervised learning objective.
- A single foundation model can be applied to many diverse tasks.
- Foundation models have emergent capabilities that are often discovered after the models are created.
What best describes self-supervised learning?
Self-supervised learning (SSL) has recently emerged as a powerful approach to learning representations from large-scale unlabeled data, showing promising results in time series analysis. The self-supervised representation learning can be categorized into two mainstream: contrastive and generative.
- The machine trains on labels that are created from the structure of the data itself.
The following are best practices to adapt foundation models to business context
- Put governance in place for generative AI use cases, such as an AI ethics review and designing for human-in-the-loop.
- Be aware of the limitations of generative AI, test various models, and start with task-specific instructions.
Embeddings useful for allowing large language models to access external information and ground results
- Embeddings represent information in a way that retains semantic, contextual meaning that can be retrieved to augment the output of a large language model.
Which adaption and grounding strategy fits the following use case? "A developer builds a chatbot that requires real-time data from a SAP business application. The chatbot may choose which API to call, gather information from the system, and return the results to the user."
- Orchestration tools ("agents")
The three offerings of the Generative AI Hub
- Developer tools, instant access, control and transparency
The role of the BTP Reference Architecture for Generative AI in a real-world application
- It offers a comprehensive SaaS solution to enhance customer support using large language models.
The function of Joule in the SAP Business AI portfolio
- It is a digital assistant that understands business context.
Tasks could you use generative AI via the Generative AI Hub
- Generating coding
- Summarizing text
- Writing assistance
The role of the AI Foundation in the SAP Business AI portfolio
- It provides access to a variety of foundation models.
- It provides the foundation towards accessing SAP-optimized large language models.
