網上照片
Artificial intelligence (AI) is one of the most exciting and rapidly evolving fields of technology today. AI systems can perform tasks that normally require human intelligence, such as understanding natural language, recognizing faces, playing games, and creating content. However, not all AI systems are created equal. Depending on how they are designed and trained, AI systems can have different capabilities and limitations.
Generative AI vs Interactive AI
One way to classify AI systems is based on how they generate outputs. Some AI systems are generative, meaning they can produce new and original outputs from scratch, such as text, images, music, or code. Other AI systems are interactive, meaning they can respond to inputs from users or other sources, such as questions, commands, feedback, or data.
Generative AI has been around for a long time, but it has gained a lot of popularity and attention in recent years thanks to advances in deep learning and neural networks. Generative AI can create amazing and realistic outputs that can fool humans or even surpass them in quality and creativity.
However, generative AI also has some drawbacks and challenges. One of them is that generative AI can be unpredictable and uncontrollable. Because generative AI learns from data without explicit rules or guidance, it can sometimes produce outputs that are offensive, inaccurate, or harmful. For example, generative AI can generate text that contains hate speech, misinformation, or plagiarism, generate images that violate privacy or ethics, or generate code that contains bugs or malicious code (but I think the latest generative AI has improved a lot in this area).
Another challenge of generative AI is that it can be hard to evaluate and improve. Because generative AI does not have a clear objective or metric to optimize, it can be difficult to measure its performance and quality. For example, how do you judge if a text is coherent, relevant, and original? How do you judge if an image is realistic, diverse, and appealing? How do you judge if a code is functional, efficient, and secure? Moreover, how do you provide feedback or corrections to generative AI to make it better?
This is where interactive AI comes in. Interactive AI is a new paradigm of AI that aims to overcome the limitations of generative AI by incorporating human input and interaction into the generation process. Interactive AI can adapt to user preferences and goals, learn from user feedback and guidance, and collaborate with user creativity and intelligence.
Forms of Interactive AI
Interactive AI can take various forms depending on the level and type of interaction involved. For example:
i) Query-based interactive AI: The user provides a query or a prompt to the AI system, such as a keyword, a topic, a question, or a command. The AI system generates an output that matches the query or the prompt as closely as possible. The user can then refine the query or the prompt to get different outputs or provide feedback to improve the output.
ii) Dialogue-based interactive AI: The user engages in a natural language conversation with the AI system. The AI system generates responses that are relevant, coherent, and informative. The user can ask questions, provide information, express opinions, or request actions from the AI system.
iii) Editing-based interactive AI: The user edits or modifies the output generated by the AI system using tools such as text editors, image editors, or code editors. The AI system updates the output accordingly based on the user edits or modifications.
iv) Co-creation-based interactive AI: The user and the AI system work together to create a joint output using tools such as collaborative platforms, brainstorming tools, or creative software. The user and the AI system contribute ideas, suggestions, feedback, or revisions to each other.
Advantages of Interactive AI over Generative AI
1) Interactive AI can be more creative and innovative. Because interactive AI can collaborate with user creativity and intelligence, it can generate outputs that are novel, diverse, and original. For example, interactive AI can write stories that combine user ideas and characters, generate artworks that blend user styles and themes, or generate code that implements user functions and features (some could argue this could also be the case for generative AI).
2) Interactive AI can be more personalized and customized. Because interactive AI can learn from user input and interaction, it can tailor its outputs to user preferences and goals. For example, interactive AI can generate text that reflects user style and tone, generate images that match user taste and mood, or generate code that meets user specifications and requirements.
3) Interactive AI can be more reliable and trustworthy. Because interactive AI can receive user feedback and guidance, it can avoid generating outputs that are inappropriate, offensive, inaccurate, or harmful. For example, interactive AI can correct text that contains hate speech, misinformation, or plagiarism, correct images that violate privacy or ethics, or correct code that contains bugs or malicious code.
To conclude, interactive AI is a new and exciting direction of AI that promises to revolutionize the way we work and create/consume content. By combining the power of generative AI with the input and interaction of humans, interactive AI should be able to offer more personalized, reliable, and creative outputs that can enhance our lives and work.
Thomas Kwan
Managing Director of Odysseus Capital Asia Limited
Thomas has been in the industry for more than 20 years and has extensive experience in the Greater China region. Thomas’ expertise includes leading M&A transactions, both buy and sell-side, fund-raising projects and advisory services including deal origination, target identification, business strategy advice, structural and regulatory advice and pricing and negotiation strategy.
Uploaded on 20-4-2024