What is Generative AI?
Generative AI is a subfield of artificial intelligence focused on creating new content from various data types, including text, images, video, and audio. Generative AI is not mindless copy-pasting, does not replace human creativity, and is not conscious.
Why use Generative AI?
AI can be helpful and excel at the adequacy of tasks.
Large Language Models (LLM)
LLM returns patterns of words based on its training. But these patterns of words function like predictive text on a phone. LLMS are pre- trained on large data sets, so they return comprehensible answers.
They can work with multiple inputs: text, audio, and image.
You interact with a model through a prompt. Prompts are instructions or cues provided to generative AI tools to guide them in generating the desired output. Performance benchmarks exist to evaluate their capabilities.
Each version improves in:
- Following directions better
- Omitting toxic language
- Being Less Likely to fabricate information
- Implementing improved “Guardrails”
- Words, parts of words, are broken down into “tokens”
- The next token is predictive. Usage limits are based on tokens
Responsible AI:
The person using AI is the person responsible for the AI output.
- AI can hallucinate with poor prompts.
- Fills in gaps with information that might not be true.
- AI Data sets might be out of date and make up an answer
- Results may require Subject Matter Review and Fact Checking
Prompt Basics
- Never input sensitive information into a Generative AI “chat”; not only is this unethical, but in some cases, you may be breaching regulations or other laws.
- Remember that Chat GPT is not a search engine and instead of using keywords strung together with Boolean logic, we prompt the GPT by giving it specific instructions.
- Be very specific in your prompts with concise sentences. Spend some time crafting a prompt.
- You can use punctuation
Prompt Goals:
- Increase accuracy
- Personalized response