AI is confusing — here’s your cheat sheet

🌈 Abstract

The article provides an overview of the key terms and concepts in the field of artificial intelligence (AI), including machine learning, artificial general intelligence (AGI), generative AI, hallucinations, bias, AI models, training, parameters, natural language processing (NLP), inference, tokens, neural networks, transformers, and hardware used for AI systems.

🙋 Q&A

[01] What is Artificial Intelligence (AI)?

• Artificial intelligence (AI) is the discipline of computer science dedicated to making computer systems that can think like a human.
• AI is often used as a technology or entity, but its definition can be mutable and confusing as companies use it as a marketing buzzword.
• AI ultimately aims to make computers smarter.

[02] What is Machine Learning?

• Machine learning is a field within artificial intelligence where systems are trained on data to make predictions about new information, allowing them to "learn."
• Machine learning is critical to many AI technologies.

[03] What is Artificial General Intelligence (AGI)?

• Artificial general intelligence (AGI) refers to AI that is as smart or smarter than a human.
• AGI is a powerful but potentially frightening prospect, as it could lead to superintelligent machines taking over the world.

[04] What is Generative AI?

• Generative AI is a technology capable of generating new text, images, code, and more.
• Generative AI tools are powered by AI models trained on vast amounts of data.

[05] What are Hallucinations in AI?

• Hallucinations refer to the ability of generative AI tools to confidently make up responses or provide factual errors, due to the limitations of the data they are trained on.
• There is controversy around whether hallucinations can ever be "fixed" in AI systems.

[06] What is Bias in AI?

• Bias can occur in AI systems due to the biases present in the training data used to develop them.
• For example, facial recognition software has been shown to have higher error rates when identifying the gender of darker-skinned women.

[07] What are AI Models?

• AI models are trained on data to perform tasks or make decisions on their own.
• Key types of AI models include large language models (LLMs), diffusion models, and foundation models.

[08] How are AI Models Trained?

• Training is the process by which AI models learn to understand data and make predictions by analyzing large datasets.
• Training often requires significant computing power and resources.
• Parameters are the variables an AI model learns during training that determine its outputs.

[09] What other AI-related Terms are Important?

• Natural Language Processing (NLP): The ability for machines to understand human language through machine learning.
• Inference: The process of an AI application generating an output, such as a ChatGPT response.
• Tokens: Chunks of text that AI models analyze and process.
• Neural Networks: Computer architectures inspired by the human brain that help AI systems learn complex patterns.
• Transformers: A type of neural network architecture that uses an "attention" mechanism to process relationships between parts of a sequence.
• RAG (Retrieval-Augmented Generation): A technique that allows AI models to find and add context from beyond their training data to improve the accuracy of their outputs.

[10] What Hardware Powers AI Systems?

• Nvidia's H100 chip is a popular GPU used for AI training.
• Neural Processing Units (NPUs) are dedicated processors that can perform efficient AI inference on devices.
• TOPS (Trillion Operations Per Second) is a metric used to measure the capability of AI chips.

[11] What are Some Leading AI Companies and Products?

• OpenAI / ChatGPT: The AI chatbot that has driven recent excitement around AI.
• Microsoft / Copilot: Microsoft's AI assistant powered by OpenAI's GPT models.
• Google / Gemini: Google's AI assistant and models.
• Meta / Llama: Meta's open-source large language model.
• Apple / Apple Intelligence: Apple's AI-focused features and products.
• Anthropic / Claude: Anthropic's AI models, backed by major investments.
• xAI / Grok: Elon Musk's AI company and its Grok large language model.
• Perplexity: An AI-powered search engine.
• Hugging Face: A platform for AI models and datasets.