Generative AI

How does it work?

• Generative Adversarial Networks: Technologies capable of generating visuals or multimedia outputs from both image and language inputs. For eg. DallE, MidJourney, StableDiffusion.

• Transformer-Based Models: Technologies like Generative Pre-Trained ( GPT ) language models may leverage Internet-led data to generate textual material, such as website articles, press releases and whitepapers. For eg. ChatGPT

• Here are some generated images using StableDiffusion

• 

• 

LLMs and GANs Difference

• LLMs do not generate images and GANs do not generate text content.

• Let's understand LLMs as follows :

• LLM stands for Large Language Model

• These are statistical models that infer the next word based on likelihoods or probabilities.

• It takes what has already split out and continuously infers the next sequence.

Language Models != Knowledge Model

• Language Models exist only to create human-like speech. They do not guarantee the accuracy of statements.

• There are other models trained to provide accurate results.

• Be wary of what you read from an LLM. The results they split out are just byproducts of statistics.

How to Harness the power of LLMS?

• Understanding Prompt Engineering

• Priming the AI

• Providing details as needed

How do Transformers power Generative AI?

• Created by Google in 2017 and have a system of attention. This gives different weights for the significance of training data.

• Generative AI is just a prediction machine

• Providing the next word based on the previous word is the heart of Generative AI.

• LLMs generally pick the next word based on the existing input. When phrasing is common, it tends to hallucinate since there are more options for the next words.

Basics of Prompt Engineering

• It is a relatively new discipline for developing and optimizing prompts to effectively use language models for a wide variety of applications and research topics.

• Prompt engineering skills help to better understand the capabilities and limitations of large language models ( LLMs ).

• Developers use prompt engineering to design robust and effective prompting techniques that interface with LLMs and other tools.

Integrate OpenAI API with Application

• We can use the below-mentioned link to build a sample application usign Open AI API

  https://platform.openai.com/overview

• Here are some sample Applications available.

• 

What is best in the market?

• Bing Chat

• Google Search Experience

• Chat GPT

• Bard

• Claude

When to use and which one?

• To get up-to-date information: BingChat, Bard, Google Search Execute

• Need to do copywriting and don't need up to date information: ChatGPT, Claude or Bard

• Need to do some copywriting and being up to date is important: Bard

• To peform advanced features like chatting with PDF: Claude

Do and Don't of using Generative AI

Here are some advantages of Generative AI:

• Generate content for a blog or a website

• Create an inspiration for an art

• Create quick prototypes and mockups

Here are some Generative AI can be dangerous :

• Don't use AI to cheat on exams and quizzes

• Don't take everything an AI says at face value. Do your own research

• Don't use AI to plagiarize other artists

AI is great tool when used correctly : " With great power comes great responsibility"

************************** Thanks ********************************