top of page
Rechercher
  • Photo du rédacteurHerve Blanc

Small AI dictionary

Dernière mise à jour : 18 mars

The AI community is very inventive and is always coming up with lots of new terms and acronyms. This AI dictionary page is just intended to provide a non-exhaustive list of those and their definitions to make the AI field easier to understand.



books in library

AGI (Artificial General Intelligence)

The Artificial General Intelligence (AGI) is also known as “Strong” AI. The best example of Artificial General Intelligence or “AGI” is what we see in Intelligent Robots who can “interact” with us and learn. “Sophia” developed by Hansen Robotics is the best example of AGI and you can see how far we are from human like intelligent robots! AGI (when fully developed) can successfully perform any intellectual task that a human can. The best example of a developed AGI robot is the iRobot Movie.


ANI (Artificial Narrow Intelligence)

Artificial Narrow Intelligence (ANI) is also known as “Weak” AI. What we see today as the result of development by different companies around the world is ANI. Every sort of Machine Intelligence that surrounds us today is Narrow AI. Google Assistant, Google Translator, Siri, and Factory Robots are all examples of limited AI. The scope of ANI is to perform only "Single Tasks" on a "specific data set". This can be done offline or on a real-time or near real-time basis.


API (Application Programming Interface)

Software interface allowing communication with a remote computer program.


ASI (Artificial Super Intelligence)

Artificial Super Intelligence (ASI) is the ultimate level of Artificial Intelligence beyond the capabilities of the human brain! Oxford philosopher Nick Bostrom defines Super Intelligence as “Any intellect that greatly exceeds the cognitive performance of humans in virtually all domains of interest”.


Bias  

AI systems that produce biased results that reflect and perpetuate human biases within a society, including historical and current social inequality.


Chatbot

computer program designed to simulate conversation with human users; ChatGPT is an example of chatbot where users can ask their questions in natural language and the LLM respond in natural language.


Clustering

Statistics of data points which have similar values. Clustering identifies groups within real-world embeddings and enables applications such as identifying which books are about the same topic.


Decision tree

A decision tree algorithm is a machine learning algorithm that uses a decision tree to make predictions. It follows a tree-like model of decisions and their possible consequences. The algorithm works by recursively splitting the data into subsets based on the most significant feature at each node of the tree.

 

Deep learning

Deep learning is the subset of machine learning methods based on artificial neural networks with representation learning. The adjective "deep" refers to the use of multiple layers in the network. 


Embedding

The process of representing the real world as data in a computer is called embedding and is necessary before the real world can be analyzed and used in computer applications. Data scientists use embeddings to represent high-dimensional data in a low-dimensional space.


Epoch

Each time a dataset passes through a machine learning algorithm, it is said to have completed an epoch. Epoch can therefore be defined as the one entire passing of a training dataset through an algorithm.

 

F1-score

F1 score is a machine learning evaluation metric that measures a model's accuracy. It combines the precision and recall scores of a model. The F1 score combines precision and recall using their harmonic mean, and maximizing the F1 score implies simultaneously maximizing both precision and recall. Thus, the F1 score has become the choice of researchers for evaluating their models in conjunction with accuracy.


Generative AI

Generative artificial intelligence is artificial intelligence capable of generating text, images or other data using generative models, often in response to prompts. Generative AI models learn the patterns and structure of their input training data and then generate new data that has similar characteristics.


GPT

Generative pre-trained transformer


Ground truth

known as the target for training or validating the model with a labeled dataset.

 

Hallucination (LLM)

When you prompt an LLM with a question/subject that was not part of its training dataset, the LLM is likely going to output gibberish. This phenomenon is referred to as a hallucination. The LLM could not infer a meaningful answer, since the LLM had not learned nor stored this new information somehow in its billions model weights.

 

Inference

applying a machine learning model to a dataset and generating an output or “prediction”

 

LLM  Large Language Model

Huge deep learning models typically based on the Transformer architecture. LLMs are trained on huge datasets (trillions of token) resulting in several billions model parameters/weights.

 

LMM  Large Multimodal Model

advanced type of artificial intelligence model that can process and understand multiple types of data modalities. These multimodal data can include text, images, audio, video, and potentially others.


Machine Learning

the use and development of computer systems that are able to learn and adapt without following explicit instructions, by using algorithms and statistical models to analyse and draw inferences from patterns in data.

 

NLG Natural Language Generation

 

NLP Natural Language Processing

 

NLU Natural Language Understanding

 

Precision

Precision measures the ratio between true and false positives


RAG Retrieval Augmented Generation

RAG is a key Generative AI framework to take advantage of LLMs with your own company’s data

 

Recall

Recall measures the ratio between true-positives and false-negatives

 

Shadow AI

AI tools usage with no company’s IT oversight, thus risking confidential data leakage (e.g. using free chatbots to help generate proprietary presentations). Professional AI tools comes with SLAs ensuring data fed to the tools are not used for creating new models (that may be potentially be used by your competitors).

 

Similarity

Similarity finds how similar real-world embeddings are to each other and enables applications such as product recommendation. Similarity is the distance between the two points. Points are more similar if they are closer together in space.

 

Training

a process in which a machine learning (ML) algorithm is fed with sufficient training data to learn from

 

XAI ore Explainable AI


That's it, I hope this AI dictionary got you motivated to want to learn more about AI.


Just don't hesitate if something is mission in the above list or if you are looking for advices.


And don't forget to spread the information if you enjoyed this blog post, just click on the social network buttons below. Sharing is caring :-)

8 vues0 commentaire

Comments


bottom of page