AI Tools

My attempt to list all the AI tools, which is changing faster than Trump’s tariff changes in 2025 🙂

This is a Not a list of all the AI Tools out there, but only those I have tried or use.

Machine Learning
– Supervised Learning – Regression, Classification, Neutal networks
– Unsupervised Learning
– Reinforced Learning . Reward

Regression – Analyse relations between features x and results y. Estimates Continuous values.

Neutal Networks – imitate the structures of the human brain, process input data to recognise patterns, make decisions or predictions

Classification – Identifies discrete class labels, assign class label y base on input features x. Process of predicting class of given data points.

Features – use features to predicts the class of given data points , use training data for input-output relationships.

Classificarion types – Decision Trees, Logistics Regression, Random Forst, Support Vestor machines,

Training – using a learning alorithm, determine the model parameters.

Data set – training (traing the algorithm), Validations (validate results, adjust algorithm parameters), Test (unseen data, access model’s performance)

Model effectiveness – accuracy, precision, recall.

Machine Learning categores:
– Supervised Learning: Use labelled data for model building
– Unsupervised Learning: Discovers patterns in unlabelled data
– Reinforced learning: Employs a reward system to guide actions

Supervised Learning Categories:
– Regression: Estimate continuous values
– Neural network: Imitates the strucure of the human brain
– Classificatio: Focus on discrete values

Training a model:
– Training set: Train the algorithm
– Validation set: Fine-tune and validates the model
– Test test: Evaluate the model’s performance

AI > ML Machine Learning > Neural Networks > DL Deep Learning (> = subset)

Deep Learning – layers algorithms for neural network, Enables AI to learn, improve quality and accuracy, learn from unstructured data.

DL – Input layer, Hidden layer, Output layer . Multiple layers.

DL configure layers, select function to connect each layer output to next layer’s input. train model by providing examples.

Deep Learning
– Layers algorithms to create a neural network
– Allows continuous improvements and learning
– Enchanges AI’s natural language understanding by grasping contect and intent
– excels in these tasks
– Imaging captioning
– Voice recognition
– Facial recognition
– Medical imaging
– Language translation
– Driverless cars

Neural Networks

NL – Input layer, one or more Hidden layer, Output layer . > 3 layers mean DL

ML vs DL

Feed forward. Back propergation.

Gen AI : Text, Art, Music, Video

Type of Gen AI:
– VAE (Variational auto encoders) . Eg Image generation, anomaly detection
– GAN (Generative Adversarial Networks). Eg Image synthesis, style transfer, data augmentation, styleGAN
– Autregressive models: eg text, music.
– Transformers: eg NLP tasks,

VAE – Encoder: converts inout to latent space representation
– Latent space representation: holds key features of the data
– Decoder: Generates new outputs based on latent space representation

GAN – Generators: Generates new data samples, creates data that looks real
– Discriminator: verifies genrated data, continuous process

Autoregressive models
– create data sequentially
– consider the context

Transformers: NLP tasks, Encode/decode layers, Text sequences. languages translations.

LLM Transformers: Eg ChatGPT, Gemini

Gen AI models:
Unimodel: generate outs in same modality (types of data processed). eg GPT-3 (text only)
Multimodel: takes input in one modality, creates outputs in a different modality. DALL-E (text to image)

Gen AI > Foundation Models: large data, unsupervised, unstructed data > LLM

Tuning – train small specific data to do specific task,

LLM:
Advantage: 1. Performance 2. Productivity
Disadvantage: 1. Compute Cost – to train and run inference. 2. Trust. (error, biase, hate,)

Prompt Engineering

Foundation Models:
– LLM
– Vision:
– Code:
– Scientific
– Chemistry:
– Climate Changes.

Example of AI
01. Smart Assistants . eg Siri. STT, NLP, TTS.
02. Smart thermostats. Preference, phone app control, smart sensors, run on schedule
03. Recommendation system. collect data, create profile, understand preference, ML to analyse data, identify patterns, predict content. Display recomendation
04. Facial recognition system. Take pic, video of face. DL, CNN convolutoin neural netowrks extract key facial features. ML algorithms compare features against a db. If match, verification is successful.
05. ChatGPT. prompt. use NLP to analyse text. using Transformer architecture, generate & display results.
06. Fraud detection system: gather transaction data. ML to analyse and learn pattern. monitors transaction in real time and compare against pattern. Flag unusual transaction. High risk transactions lead to automatice actions, like freezing of account.

AI > ML > DL > Foundation models > LLM

NLP – Natural Language Processing
STT – Speech to text
TTS – Text to speech

AI – Computer Vision