Emerging trend
Generative AI
Group Members
Bob Gichora 664835
William Kariuki 665259
Melvin Mango 664136
Josiah Okoth 664364
Glory Kiogora 665726
Introduction
What is generative AI
Generative AI is deep learning models that can identify patterns and
structures within existing data and can generate various types of content like
text, images, audio and videos.
How does generative AI differ from other models.
Generative AI is different from other AI mainly because of its generative data.
Traditional AI mostly solves specific tasks or objectives but generative AI can
come up with new content.
Generative AI vs Traditional AI
Traditional AI Generative AI
Uses a rule based approach where the rules are pre- Uses large datasets and deep learning networks to
determined by a human being based on the human’s identify underlying patterns from the datasets. It will
understanding of the problem. They require these rules then use the identified patterns to generate new
to solve problems and give output content
Uses supervised learning where it is given labelled It can use both but works best with unsupervised
datasets and corresponding input and output is given. learning. It takes un labelled inputs and it will learn to
It learns to match certain inputs to specific outputs and identify underlying patterns with limited human
it does this by the help of a human. guidance. This enables it to generate new content.
Uses discriminative to classify objects into various Uses probability from the analysed data to generate
specific categories. Example it can separate cats from new samples that resemble the real thing. Example is it
dogs. can use Generative Adversarial Network(GAN) to
generate things from the real world.
Is less creative since it cannot work outside of the rules Is more creative and can generate new content. This is
that have stipulated for it. It cannot adapt to new because of its unsupervised learning. It can adapt to
situations without the help of a human. new data, learn from it and come up with new things.
Main components of Generative AI
Data processing layer - Collects prepares and analyses the data
Generative model layer - Generates new content using machine learning
models.
Feedback and Improvement layer - It collects user feedback, analyses
the generated data and uses it to improve the model
Deployment and integration model - Integrated the AI into a system or a
final product. For use by the end consumer.
Generative AI uses a large
How dataset to analyze patterns and
rules in images or text. It uses
generative AI probability distribution to create
new samples resembling real
works objects, making continuous
adjustments to increase its
probability and accuracy.
Generative AI Models
Generative Adversarial Network (GAN)
Large Language Models
Variational Auto-encoders (VAEs)
Transformer based Model
Flow Based models
Brief history of generative AI
Introduction of Artificial Intelligence (1950s)
Proposal of Perceptron (1958):
Creation of ELIZA Chatbot (1964)
Development of RNNs (1982)
Introduction of LSTM (1997)
Creation of Variational Autoencoders (2013)
Development of Generative Adversarial Networks (2014)
Introduction of Diffusion Models (2015)
Proposal of Transformer Architecture (2017)
Modern Generative Models (Recent Years)
Emerging Trends
Adversarial training techniques- Adversarial training has received a lot of
attention, especially in the field of generative adversarial networks (GANs).
Transformer Architecture dominance- Transformers, which were originally
developed for natural language processing, have also demonstrated remarkable
efficacy in generative modelling.
Continued scale increase- With the availability of more powerful
computational resources and the introduction of techniques such as distributed
training and model parallelism, there is a trend towards scaling up generative
models.
Cross-modal generative models- Cross-modal generative models seek to
generate content in multiple modalities, including images, text, and audio.
Fine-Tuning and Transfer Learning- Pre-trained generative models are now
routinely fine-tuned on domain-specific data and tasks.
Roles and Application of Chat GPT
Roles
Cutting edge Chatbot; Interacts with users through text conversations.
Wide Range of Uses; Suitable, for customer support virtual helpers,
teaching and entertainment.
Application
24/7 Customer Support; Delivers assistance, throughout the day, in retail
responding promptly to inquiries.
Personalized Tutoring Services; Tailors teaching methods to needs helping
students understand ideas effectively.
Creative Writing Assistance; Supports the creation of content improving
the productivity of writing scripts.
Automated Email Management; Manages email replies automatically
making communication smoother and more efficient
Features of Chat GPT
Understanding Natural Language
Artificial intelligence is quite good at understanding natural language. It is
even capable of understanding the more complex points, context, and
complexity of human speech.
Contextual Dialogue Handling
ChatGPT's capability to preserve context/information during a conversation is
honestly one of its best qualities. It considers the complete history of the
conversation rather than only reacting to the most recent input.
Importance of ChatGPT
Customer Interaction Revolution: Provides 24/7 customer support,
enhancing customer engagement. Transforms how companies connect
with their customers
Productivity Enhancement: Automates repetitive tasks, freeing up
valuable human resources. Allows employees to concentrate on more
complex, impactful endeavors.
Learning Revolution: Offers personalized tutoring opportunities,
accessible to all. Brings revolutionary changes to the educational sector.
Creative and Entertainment Enrichments: Generates rich, human-
like text, inspiring creativity. Enhances the entertainment industry with
new possibilities.
Advantages of ChatGPT
Efficiency: It reduces effort and human error by automating repetitive
processes like responding to frequently asked questions.
Scalability: Chat GPT is perfect for companies with varying workloads
since it can process several requests at once.
Cost-effective: It saves money by removing the need to hire and train
more employees.
Multilingual Support: Chat GPT can interact with users in a variety of
languages, increasing its audience base worldwide.
Disadvantages of ChatGPT
Lack of Emotional Intelligence: Chat GPT is less capable of managing
emotionally complex or sensitive situations since it does not have
emotional understanding.
Inaccuracy: It might give false information, particularly in response to
unclear or strange requests.
Over-Reliance: Reliance on Chat GPT too much might result in a
decrease of interpersonal engagement, which can harm reputation and
lower customer satisfaction.
Privacy and security issues: They are brought up when handling
sensitive data, necessitating the use of strong data protection procedures.
Initial Setup Complexity: Putting Chat GPT into practice requires a lot of
setup and training, which can be expensive and time-consuming.
Impact of ChatGPT
Increasing Interaction Capabilities: As technology advances, speech and visual
elements will be added, allowing for more versatile and intuitive user interactions.
Deeper Integration with the Web and Applications: As AI becomes more
integrated into commonplace programs like Microsoft's suite, it will eventually become a
vital component of digital systems, increasing productivity through the use of natural
language interactions.
Revolutionizing Software Development: Indicates a new era in software design and
development by having a substantial impact on code authoring, bug fixing, and other
development chores.
Economic Transformation: It is anticipated that automation and the emergence of
new job categories will lead to an economic transformation and the need for flexible
regulatory frameworks.
Enhanced Working Environment: Chat GPT features in cooperative work settings
facilitate teamwork and allow people to concentrate on more strategic, innovative tasks,
which is a step toward an improved work environment.
Conclusion
In conclusion, generative AI has shown that although it has
demonstrated its potential for future significance, such as its potential to
make user and business experiences better and easier, there are also
some limitations to it. Therefore, due to the continuous evolution of AI, it
is not only beneficial but essential to adapt to the changes since, as
clearly shown above, it is the future.
REFERENCES
OpenAI. (n.d.). [Link]
[Link]
Bernard marr, A simple guide to the history of AI .
[Link]
Dhaduk, H. ( 2023, May 23). Simform. Retrieved from [Link]: /
[Link]
Lawton, G. (n.d.). TechTarget. Retrieved from [Link]:
[Link]
Vadapalli, P. (n.d.). UpGrad. Retrieved from [Link]:
[Link]
Thank you