Page 349 - Ai_417_V3.0_C9_Flipbook
P. 349
• The outputs are scrutinised for bias and misinformation. Implement tools and processes to detect biases in
AI-generated content.
• Prioritising user privacy and informed consent. Apply strong encryption and data anonymisation techniques to
protect user data. Ensure users are aware of and consent to how their data is being used by AI systems.
• Educating stakeholders on ethical use and risks. Provide ongoing education for developers on ethical AI practices
and the potential risks of AI misuse. Educate users about the capabilities, limitations, and ethical considerations
of AI technology.
• Establishing clear guidelines on ownership and attribution. Define and enforce guidelines regarding the
ownership of AI-generated content. Clearly attribute AI-generated content to its sources, distinguishing between
human and AI contributions.
• Engaging in public discussions around the social and ethical implications. Foster open discussions with the public
about the benefits and risks of Generative AI. Collaborate with policymakers, ethicists, and other stakeholders to
develop guidelines and regulations that ensure AI is used in socially beneficial ways.
All these points ensure the responsible use of Generative AI. By emphasising ethics, creating trust, limiting
negative repercussions, defining legislation, and encouraging innovation, we maximise generative AI’s potential
and use it in ways that benefit society.
At a Glance
• AI-generated images are created using AI algorithms.
• Distinguishing between a real image and one generated by AI can be challenging as AI-generated images continue
to become more sophisticated.
• Artificial intelligence shows inconsistencies if observed closely, although it tries to piece together its creations from
the original work.
• AI-generated images may include elements that seem unrealistic or improbable, such as impossible perspectives,
mismatched colors, or objects that defy physics.
• In a supervised learning model, a labelled dataset is given to the machine.
• A labelled dataset is the information which is tagged with identifiers of data.
• Discriminative Modelling is an approach in Machine Learning where the focus is on learning the boundary or
decision boundary that separates different classes or categories directly from the data.
• Unsupervised Learning is a type of Machine Learning where the model is trained on input data without any
corresponding output labels.
• In Generative Modelling there is no labelled dataset, and the model can generate structured data from the Random
Noise dataset.
• A "random noise dataset" typically refers to a collection of data points or samples where each data point is generated
randomly.
• Generative Artificial Intelligence also called Gen AI, refers to the algorithms that generate new data that resembles
human-generated content, such as audio, code, images, text, simulations, and videos.
• Generative AI is trained with existing data and content, creating the potential for applications such as natural
language processing, computer vision, the meta-verse, and speech synthesis.
• GANs are neural networks that work to produce fresh data.
• Variational Autoencoders (VAEs) produces fresh data, learn the distribution of the data and then sample from it.
• RNNs are a special class of neural networks that excel at handling sequential data, like music or text.
• Autoencoders are neural networks that have been trained to learn a compressed representation of data.
Introduction to Generative AI 347

