augusti 15, 2023
Artificial intelligence (AI) has emerged as a powerful force, transforming the way we live and work, and reshaping our future. But navigating the world of AI can feel like learning a new language. This glossary is designed to decode AI jargon and serve as a handy reference tool for general AI terminology.
Fundamentals of AI
Let's start with the core concepts that underpin AI and its applications.
Artificial Intelligence (AI)
Refers to the development of computer systems that can simulate human reasoning and perform tasks that typically require human intelligence. It encompasses various techniques, such as machine learning and natural language processing, to enable machines to learn, reason and make decisions or take actions based on patterns in existing data. AI is a broad field concerned with creating intelligent systems that can perform tasks requiring human-like intelligence, such as problem-solving, pattern recognition, decision-making and natural language understanding.
Google breaks it down further into these four commonly recognized stages of AI development (see more details here):
-
Reactive machines: Limited AI that only reacts to different kinds of stimuli based on preprogrammed rules. Does not use memory and thus cannot learn with new data.
-
Limited memory: Most modern AI is considered to be limited memory. It can use memory to improve over time by being trained with new data, typically through an artificial neural network or other training model.
-
Theory of mind: Theory of mind AI does not currently exist, but research is ongoing into its possibilities. It describes AI that can emulate the human mind and has decision-making capabilities equal to that of a human, including recognizing and remembering emotions and reacting in social situations as a human would.
-
Self-aware: A step above theory of mind AI, self-aware AI describes a mythical machine that is aware of its own existence and has the intellectual and emotional capabilities of a human. Like theory of mind AI, self-aware AI does not currently exist.
All about machine learning
Now, some general “learning” lingo...
Machine Learning (ML)
A subset of AI that focuses on algorithms and models, enabling computers to learn from data, make predictions and take actions without explicit programming. ML algorithms can be applied to various AI tasks, such as image recognition, natural language processing, recommendation systems and many others. Through ML, AI systems can autonomously improve their performance and adapt to new data or situations.
It's worth noting that ML is just one approach to achieving AI; there are other techniques and branches within AI - such as rule-based systems, expert systems and genetic algorithms - that do not rely on ML.
Automated Machine Learning (AutoML)
Automated ML can encompass many concepts including ensembles (models based on “votes” from many models, model tuning (adjusting hyperparameters based on Bayesian search or other approach), model selection/weighting and model explanation.
AutoML takes advantage of cloud computing resources to automate many tasks historically done manually by data science teams. It is particularly useful during opportunity discovery to better understand data and its ability to achieve a certain outcome. For example, with historical, labeled customer churn data, autoML can help companies learn what is driving customer retention and loss, and how these factors relate to each other.
Deep Learning
A subset of ML that leverages artificial neural networks inspired by the human brain. It involves training models with multiple layers to learn complex patterns and representations, enabling breakthroughs in areas like image and speech recognition.
Large Language Model (LLM)
Refers to AI models that can generate natural language texts from large amounts of data. LLMs use deep neural networks, such as transformers, to learn from billions or trillions of words, and to produce texts on any topic or domain.
Fine-tuning
An approach in machine learning to transfer learning in which the weights of a pre-trained model are trained on new data.
Embedding
A relatively low-dimensional space into which you can translate high-dimensional vectors. Embeddings make it easier to do machine learning on large inputs like sparse vectors representing words.
Learning and training AI systems
There are three different types of learning algorithms and processes for training AI models.
Supervised learning
Algorithms learn from labeled training data, where each data point has a corresponding label or target. The algorithm learns to make predictions or classifications by mapping inputs to outputs based on the provided labels.
Unsupervised learning
Involves training algorithms on unlabeled data, allowing them to discover hidden patterns and structures autonomously. This type of learning is useful for tasks like clustering, anomaly detection and dimensionality reduction.
Reinforcement learning
Inspired by behavioral psychology, this involves an AI agent learning through interaction with an environment. The agent receives feedback in the form of rewards or punishments, enabling it to learn optimal strategies to maximize cumulative rewards.
Practical applications of AI
There are many ways in which AI can be used to transform your business; following are several examples.
Natural Language Processing (NLP)
Enables machines to understand and interpret human language. It involves tasks like sentiment analysis, language translation and chatbot interactions. NLP helps us communicate with AI systems more naturally, bridging the gap between humans and machines.
Computer Vision
Focuses on enabling machines to understand and interpret visual data, including images and videos. It involves tasks like object detection, image recognition, facial recognition and autonomous vehicle navigation.
Recommendation Systems
Analyze user data and patterns to provide personalized suggestions or recommendations. They are prevalent in e-commerce, streaming services and social media platforms, enhancing user experiences and increasing engagement.
Conversational AI
A system that can conduct a conversation by understanding and generating human-like language. Example: Chatbots are widely used to engage with customers across multiple digital channels.
Generative AI
A type of AI system capable of generating new text, images or other media in response to prompts. Generative AI systems create output from a probabilistic algorithm based on a combination of the prompt given, the data it was trained on and the level of creativity requested.
Google AI tools and services
Vertex AI
Allows you to build, deploy and scale ML models faster, with fully managed ML tools for any use case.
Contact Center AI & Virtual Agents
Human-like AI-powered contact center experiences that enable virtual agents to converse naturally with customers and expertly assist human agents on complex cases.
Recommendations AI
Enables retailers to deliver highly personalized product recommendations at scale using state-of-the-art ML models.
See more Google AI products and services here.
Ethical and security considerations of AI
As AI becomes more pervasive, it’s crucial to address ethical considerations and potential challenges.
Bias
AI algorithms can inadvertently inherit biases present in the data they are trained on and from the people who created them. This can lead to unfair or discriminatory outcomes. Addressing bias in AI requires careful data selection, preprocessing and algorithmic fairness considerations.
Transparency and explainability
As AI systems make critical decisions, it's important to understand how and why they arrive at those decisions. Explainable AI aims to provide insights into the decision-making process, ensuring transparency and accountability.
Privacy and security
With the abundance of data used in AI systems, privacy and security are paramount. It's crucial to handle and protect user data responsibly, ensuring compliance with privacy regulations and implementing robust security measures.
AI and copyright
In the context of copyright infringement, the legal implications of using AI are still unclear. Courts across the globe are pondering how to address both Generative AI outputs and AI training processes. There are primarily two concerns:
-
Intellectual Property (IP) Leakage - When using a public service, such as Google’s Bard AI chatbot, the text or code used in the prompts could be used to train the model and could therefore show up - in some form - in a response to another user.
-
Inadvertent use of copyrighted material - The other side of the coin is accidentally using code or text created by Generative AI that is protected by some form of copyright that only allows the text or code to be used with restrictions.
Want to know more about the power of Google AI? Just ask us.
