Ones and Zeros

AI Practitioner Exam Prep -  Aspects of AI Cost Considerations   Cost factors are responsiveness and availability, redundancy and regional coverage, performance, token-based pricing, provisioned throughput, and refining your custom models. Responsible AI Aspects: controllability, explainability, fairness, governance, interpretability\transparency, privacy\security (theft and exposure risk), safety, and veracity\robustness.   Business Benefits: trust, regulatory complain, mitigate risks, competitive advantage, improved decision making, and improved products.    Model Selection: a)  Narrow the use case to tune your model to your use case.  Ex: favor recall or precision , b) pick by performance with some test data sets,  c) responsible agency, d) environmental reasons, e) economical reasons.   Dataset Prep: want  balanced dataset  so  inclusive and diverse  in data collection, curating by 1) preprocessing, 2) augmentat...

AI Practitioner Exam Prep -  Products Abbreviations KMS  =  Amazon   Key Mgmt Service  product Developer Products (in ML frameworks layer) SageMaker   AI  =  IDE  plugin that is  fully managed service that automates ML lifecycle ( from data prep to production) with " no-code" environment and handles infrastructure to streamline building, tuning , and deploying models.  Capabilities: Predictive analytics, computer vision, NLP, and fraud detection. Operations: Features auto-training and integrated logging. Security & Data: Secured via IAM ; integrates with S3 (storage), Lambda (triggers),  CloudWatch ( monitoring)  and API Gateway (endpoints) .     Auto Model Tuning feature exists on Sagemaker.     Auto Pilot = uses Clarify to show how ML models could make  predictions. uses  SHAP values. Auto finds the  best hyperparameters.     Canvas =  No-code ML tool to crea...

  AI Practitioner Exam Prep -  Training Training (Learning)  Types Terms Continued Pre-training  =  FM gets latest raw U and updates its weights to teach domain knowledge. extends initial pre-training phase. Deductive  =  using general rules to specific outcomes Emergent  =  at large scales, these models develop skills that are not explicitly programmed into them . Federated Learning  =  instead of bring data to central server (traditional), this brings model to the data. Good for data privacy and local compliance. "Fine tuning"  =  improves already deployed pre-trained LM  using small L  ( ex: industry-specific data )  or input/output  pairs often for behavior or task specialization .  Most important task for fine tuning is labeling with accurate and relevant labels.  Types are instruction tuning, RHLF, adapting models for specific domains, transfer learning, and continuous pretraining. ...

  AI Practitioner Exam Prep - 3C Prompts Abbreviations ART prompting = Automatic Reasoning and Tool-use, CoT prompting = Chain-Of-Thought,  RAG = Retrieval-Augmented Generation,  ToT prompting = Tree of thoughts Prompt Terms Adversarial Prompting  =  protects against prompt injection attacks ART  prompting  =   breaks complex into steps using external tools (search/calculators) for accuracy.  Complexity Based prompting  =  u ses complex chain-of-thought examples to improve multi-step outputs. Context of prior messages = append prior message(s) to current message. CoT prompts  =  user prompts for answers to steps from AI. Directional Stimulus prompting  =  guides desired outputs by providing hints, keywords, or structure. Examples in Prompt = helps tailor the output better  Exposure  =  PII or privacy issues in the output Few-shot prompts  =  user prompts for something that AI is giv...

AI Practitioner Exam Prep - Data Abbreviations IDP = Intelligent Data Processing, Data Terms Classification = SL. groups data into known labeled groups. ex1: data = car pictures labeled by maker and model, ex2: customer sentiment grouping. Types are binary and multi-class classification. Clustering = UL. groups data with no labels into previously unknown groups. Curating = structures the data for processing prior to learning. See Labeling. De-identification = removing PII such as social security Encoding = converts from non-numeric to numeric. Governance = managing, securing, and monitoring data throughout its lifecycle IDP = extracts and classifies unstructured data in docs. gives summaries and actionable insights. Labeling = id and tags with content labels of each piece thus classifying. See Curating. Multi-Modal = uses multiple data types (such as text, images, audio, video, and computer code). Multi-Modal Embedding = uses multiple data types embedding them into a shared ...

AI Practitioner Exam Prep - Model + Inference Abbreviations ATLAS = Adversarial Threat Landscape for AI Systems, BERT metric = Bidirectional Encoder Representations from Transformers, BLEU metric = Bilingual Evaluation Understudy,  CNN = Convolutional Neural Networks ,  GLUE benchmark = General Language Understanding Evaluation, GPT = Generative Pre-trained Transformers,  FM = Foundational Models ,  L = Labeled Data,  LLM = Large Language Models ,  METEOR metric = Metric for Evaluation of Translation with Explicit ORdering , NL or NLP = Natural Language Processing ,  NTM = Neural Topic Modeling ,  PDP = Partial dependence plots ,  PRA = Privacy Reference Architecture , RAG = Retrieval-Augmented Generation , RL = Reinforcement Learning , RLHF = Reinforcement Learning from Human Feedback , ROUGE metric = Recall-Oriented Understudy for Gisting Evaluation , RWK metric = Real World Knowledge score, SL = Supervised Learning ,...