1/15
Looks like no tags are added yet.
predictive machine learning
Uses past data to predict future outcomes.
generative AI
Creates new content based on information it receives.
Large lanugage model LLM:
An AI program trained on huge amounts of text to understand and generate human language.
[redicts
What LLMs Can't Do
LLMs cannot reliably perform complex math on their own.
They predict numbers instead of actually calculating them.
Enterprise AI Systems (AI Agents)
AI Agents: AI that can use tools, such as Python or calculators, to perform calculations and complete tasks automatically.
does LLM have memory
no memoery of past interactions
what is lack of world model in LLM
LLM don’t model reality only language about reality
LLMs understand patterns in words, but they don't truly understand how the real world works.
LLMs learn patterns in language, not the real world itself.
hallucination liability
when a model can;t calculate, retain, or perceive is forced to give an answer anyway
snowball effect
When AI makes one small mistake, that mistake can lead to more mistakes as it continues working.
gen ai: base model
An AI trained on general information, without being specially trained for a specific task.
RAG (Retrieval-Augmented Generation)
AI that looks up information from documents or databases before answering.
Claude Cowork:
An AI agent that can work directly with files and applications on your computer without needing coding skills.
what is augemented
means improved, enhanced or supported by adding something extra.
Model monoculture:
When many banks use the same AI models or technology providers, meaning they can make the same mistakes
(Herding)
1. Same brain, same blind spots
Fine-tuned model:
: An AI trained further to specialize in a specific area.