Generative AI FAQs

What is Generative AI?

Generative AI, commonly called GenAI, allows users to input a variety of prompts to generate new content, such as text, images, videos, sounds, code, 3D designs, and other media. It is trained on documents and artifacts that already exist online, "learning" from these data sets so it can predict outcomes in the same ways humans might create on their own.

The rise of generative AI is largely due to the fact that people can use natural language to prompt AI now, so the use cases for it have multiplied. Across different industries, AI generators are now being used as a companion for writing, research, coding, designing, and more. And GenAI will continue to evolve as it's trained on more data.1

Imagine a world where machines can craft stories, compose music, design art, and even develop software code. Except that now you don't have to imagine this world, it's already here. And therein lies the magic of Generative Artificial Intelligence (AI)! Unlike traditional AI, which analyzes existing data, generative AI creates new content by learning patterns from vast datasets. It's like teaching a computer the art of creativity. You've heard about it in between

The rapid evolution of generative AI has ignited discussions worldwide. Startups like Together AI have achieved valuations of $3.3 billion, reflecting the immense interest and investment in this field. However, in the words of a great (albeit fictitious man) with great power comes great responsibility. The integration of AI in industries like gaming has sparked debates. Developers express concerns about AI potentially devaluing human creativity and craftsmanship, emphasizing the need for a balanced approach.

How does it work?

Generative AI models generate new content by using neural networks to identify patterns in existing data. Trained on unsupervised and semi-supervised learning approaches, organizations can create foundation models from large, unlabeled data sets, essentially forming a base for AI systems to perform tasks

There are many generative AI models, including large language models (like ChatGPT), image generation models (like DALL-E), and audio generation models.[2]

Accuracy and verification of AI outputs

Accepting AI Edits Without Review: Automated suggestions may misinterpret context or introduce inaccuracies. Overreliance on AI for Citations: While AI tools can suggest references, it's crucial to verify each source's relevance and accuracy. Blindly accepting AI-generated citations can lead to misrepresentation or inclusion of non-existent sources.​[3]

What are examples of GenAI tools?

Examples of generative artificial intelligence that you may have heard of include Google’s Bard, ChatGPT, or DALL-E from OpenAI.

  • ChatGPT or DALL-E: Generative artificial intelligence created by OpenAI, a Microsoft-backed, profit-capped company with the mission to develop artificial intelligence to serve humankind
  • Google Bard: Google’s generative AI with integrations to Google products like Google Lens and Gmail, operating with a language model called PaLM-2 that was trained on the largest data set out of all generative AI models available at the time of its release [3]

    • What can GenAI do (strengths)?

    Generative artificial intelligence has applications in diverse industries such as health care, manufacturing, software development, financial services, media and entertainment, and advertising and marketing. Let’s examine some of the different ways professionals in these industries apply generative AI to their field.

    1. Health care and pharmaceuticals
      2. Generative artificial intelligence has applications for all parts of the health care and pharmaceutical industry, from discovering and developing new life-saving medicine to personalizing treatment plans for individual patients to creating predictive images for charting disease progression.
    2. Advertising and marketing
      3. Generative artificial intelligence offers many solutions to professionals working in advertising and marketing, such as generating text and images needed for marketing or finding new ways to interact with customers.
    3. Manufacturing
      4. In manufacturing, professionals can use generative AI to look for ways to improve efficiency, anticipate maintenance needs before they cause problems, help engineers create better designs faster, and create a more resilient supply chain.
    4. Software development
      5. For a software development team, generative AI can provide tools to create and optimize code faster and with less experience using programming languages.
    5. Financial services
      6. According to McKinsey, generative AI could add $200 billion to $340 billion of value to the banking industry annually. Some of the applications of generative AI in the financial services industry include artificial intelligence investment strategies, drafting documentation and monitoring regulatory changes, and using generative AI as an interpreter to facilitate communications between clients and investors
    6. Media and entertainment
      7. Media and entertainment could embrace generative AI in several ways, considering the industry primarily engages in the same task as the tech: generating unique content. Generative AI can help create and edit visual content, create short highlight videos of sporting events, and make working with content management systems easier.[3]

    What are its limitations or risks?

    If you choose to use GenAI in your academic work, you should be aware of its limitations.

    Inaccuracy

  • Information from GenAI may appear convincing, but GenAI does not understand the content it generates, which can lead to hallucinations or factual errors.
  • GenAI generates responses based on data on which it has been trained. If you are interested in a very specialised topic, or something very new, the GenAI will have fewer sources on which to draw so the depth of its response may be limited.
  • GenAI cannot apply critical thinking, so may present misinterpreted information.

    • Bias

  • GenAI generates responses based on data on which it has been trained, which includes webpages, social media conversations and other online content which may be biased, offensive or outdated.

    • Copyright and intellectual property

  • Any information you put into freely available Generative AI tools as prompts becomes available to everybody, so you have to be careful not to put in personal or confidential information or protected intellectual property. If using a general GPT tool, you are advised to use Microsoft CoPilot through UCL, which includes some data protection. See also Guidance on the use of Third-Party AI Services at UCL.
  • If using information from GenAI, there is no way of knowing if the material you are using has infringed copyright and there is uncertainty over who owns the intellectual property of AI outputs.[5]

    • Risks

  • Generative AI's ability to quickly generate new textual, visual, and auditory content offers a wide range of benefits and risks. For example, AI-generated text, images, and videos make it possible to spread misinformation at scale, given that it’s becoming more difficult to tell human-generated content from AI-generated content. Clear ethical guidelines around AI-generated content could help reduce the spread of misinformation.

    • What are the differences between AI and GenAI

    Generative AI is a subset of artificial intelligence, which is essentially defined as the pursuit of creating machines capable of exhibiting (or exceeding) human intelligence. GenAI is a type of machine learning focused on building generative models capable of producing a wide range of AI-generated content, including human-like text, images, and audio. So while AI is typically designed to perform a narrow range of tasks repetitively, GenAI can produce original content in response to various user inputs. [1]