What is an AI model? A Plain-English Guide

A few years back, artificial intelligence was like something extracted from a sci-fi film. Nowadays, it’s an integral part of us. When you ask ChatGPT a question, unlock your phone with facial recognition, or watch a movie suggested by Netflix that you don’t know is available anywhere else, an AI model is doing its work.

Which is why there are a lot of people asking, “What is an AI model?” The word is used in news articles, in job descriptions, in online courses, and in business discussions, but many still don’t know what it is completely. Is it software? Is it an algorithm? Or is something different going on?

In this guide, we’ll break down what an AI model is in simple terms, of course, but also delve into the types of AI models, understand how AI models are trained, get a look at an AI model example in day-to-day life, and discuss how to build an AI model. And, at the end, you will understand how contemporary AI systems think, learn and make decisions.

Quick Summary

Everything is AI, from chatbots and voice assistants to recommendation engines and self-driving technology. However, there is an AI model behind each AI-powered tool. It provides terminology for an AI model, how it functions, types of AI models, and the applications of AI models in daily life.

6-Point Comprehensive Summary

• What is an AI model? An AI model is a program that is trained using data to make predictions or recognise patterns, or it may be used to generate text.
• Types of AI models: AI models are of different types for different tasks, like language generation, image recognition, and decision-making.
• AI Model Training: At the time of training AI models, a vast amount of data is utilised to make these models more precise.
• AI Models list: In today’s context, AI models are reshaping challenges in the real world, for example, with ChatGPT, Gemini, etc., and also particularly with Stable Diffusion, which is easy to use.
• AI learning models: AI learning models use supervised learning, unsupervised learning and reinforcement learning to help them learn.
• How to Build an AI Model: There are 4 steps that follow to create an AI model: Formulating & Defining a Problem, Collecting Data, Training the Model, and Testing the Model.

Key Takeaways

• What is an AI model? is a topic that comes up often today, especially given that AI is used in various sectors, such as searching, chats, healthcare, finance, and much more.
• Knowing how to explain the working of AI learning helps make sense of the models that are better.
• It’s important to select the right AI model for your specific use case, given your available data and objectives.

What Is an AI Model, Exactly?

Whether you’ve thought about how ChatGPT forms answers, how Google Photos can accurately identify faces, or how YouTube’s AI makes great suggestions for viewing, you’ve already been considering AI models. The AI model is essentially a piece of computer code that has been trained with information and can make use of it to accomplish a task.

The Simple Definition

The simplest way to comprehend an AI model is to equate it with a pupil getting ready for an exam. Learning with data, similar to how students learn from examples, practice questions and see patterns, is a fundamental part of decision-making or prediction-making for the AI models.

• An AI (artificial intelligence) model is a mathematical structure trained on data to recognise patterns and make predictions, categorise things, suggest, and generate new content based on the patterns seen.
• AI model quality is directly linked to data quality and quantity used for AI model training.
• The latest AI learning models continue to train, test and fine-tune, gleaning better and better results.
• All AI model examples, like chatbots and fraud detection systems, work on the same principle of learning.

AI Model vs. AI System: What’s the Difference?

An AI model and an AI system are not synonymous but are pretty equal. This distinction is helpful to grasp to better visualise how AI applications are actually constructed.

• The “brain” is an AI model that is trained with data and is capable of making predictions or decisions based on the training it has received.
• An AI system comprises the AI model and software, databases, interfaces, security, and hardware necessary to make it a fully functional solution.
• As an example, the AI model used by ChatGPT is the language model, which is an AI system.
• This isn’t an irrelevant distinction to make before examining other AI models employed in the sectors nowadays.

How an AI Model Actually Works

AI models may appear incredibly complicated at first glance, but the main concept is incredibly straightforward. All AI models have to go through a process of input, processing based on the information, and output.

Input, Processing, and Output

An AI model always has an input, a teaching step to process the information and, lastly, an output. The general workflow from query to result is always multiple steps, which often includes query, recognition, and recommendation, like when notifying if a photograph was taken, sequencing a movie, recommending a film, etc.

• Input—It refers to the information fed into the AI model, whether text, images, video, audio or numerical.
• When processing, the model will compare the new input with the pattern that is learned from the AI model training to have an understanding of what the new input is.
• The output is the model’s understanding, which can be a prediction or a recommendation, a generated response or a classification.
• Different AI models have been fuelled by this workflow, such as virtual assistants, translation tools, medical software diagnostic systems, and fraud detection systems.

The Role of Algorithms and Parameters

An AI model does not have human thinking capabilities. Rather, it processes algorithms and thousands of mathematical values which are fed into the process gradually during training to enhance the predictions the system makes over time.

• An algorithm dictates the learning method, and parameters contain the means by which the model learns from the training data.
• The latest AI models, such as ChatGPT, have billions of parameters, which enable them to grasp context and produce natural-sounding replies.
• The higher these parameters are, the more they work well with the unseen and new data, the better they are tuned and trained.
• One of the key reasons for using extensive computing power and careful analysis of data to train the most current and best AI model is the need for its support.

AI Model Training: How Models Learn

A lot of people think that AI models have a set of preprogrammed solutions. However, that’s not the case! AI models are believed to be able to access a matrix of all the answers and be prepared to deliver them. This isn’t true. Instead, they will imbibe as much information as possible and offer comments on the recurring patterns they detect and will integrate the errors and gradually improve over repeated trials.

What Training Data Actually Does

The training data constitutes the learning content of an AI model. Solving more questions helps students learn the underlying concepts, and as more data is analysed, AI becomes more intelligent.

• The model is taught how various inputs are connected with expected outputs in a variety of situations by the training data that are fed into it.
• In the real world, good-quality data can reduce any bias and improve the prediction accuracy of the model.
• Good performance and accurate forecasts rely upon the quality and completeness of the data.
• At the very start of AI model training, reliable, balanced and representative data collection is always key.

Training, Validation, and Testing Explained

If you can use more than just data to build an AI model. Developers also should try to verify that it has learnt correctly before putting it to use, running tests to verify that it has learnt correctly.

• During training, the process model executes on many facts and adjusts its internal parameters accordingly to the training facts.
• The developer can validate the model during the development process to detect potential overfitting or underfitting in the model and rectify those.
• Also, during the test phase, the trained model will be able to identify the pictures in the test data that it has never been exposed to during the training phase.
• The three steps create a smooth and consistent pipeline for the reliable and consistent performance of AI learning models.

When can training be completed?

The time it takes to train a model for AI varies from a few days to several weeks, depending on the complexity of the AI model, the amount of training data and the computing resources utilised. Some models can take a few hours to learn; others can take weeks or months.

• Small models that can be trained in a few hours with common computer hardware can do so.
• A major drawback is that the large language models can require thousands of GPUs, all working simultaneously for weeks of time before the AI can provide any meaningful results.
• Models are often “retrained” as new information arrives to make them more accurate and change with the changing demands of users.
• With advancements in AI technology, the development process and operating costs are being streamlined with more efficient training methods.

Different Types of AI Models

You don’t have to use every type of AI model for every job. Some specialise in pattern recognition, others generate totally new content, and a few are designed to make decisions by experience. Knowing different types of AI models makes them more suitable for use in various real-life applications.

Discriminative Models

Discriminative models are about understanding, if not generating, new content. Their applications are broad, wherein the crucial task is to make proper classification.

• These models help to classify a spam email, have the ability to recognise faces in a photo or detect fraudulent financial transactions.
• They develop an understanding of the connection between inputs and outputs but do not attempt to produce entirely new data.
• Discriminative models have been widely applied in many image classification and medical diagnosis systems due to their good prediction performance.

Generative Models

This is where generative models have become some of the most successful AI technologies today, as they can produce totally new content instead of just analysing data.

• They can produce text, pictures, sound, video or software code from simple instructions – and even a real-sounding voice.
• Well-known AI model example applications based on generative AI include ChatGPT, Claude, Gemini, and Stable Diffusion.
• There are numerous businesses today that utilise an AI model generator to accelerate content production, style development, software development, and customer support.

Foundation Models

Foundation models are larger, more capable AI models trained using vast amounts of data that are able to handle a host of different tasks without having to be built specifically to each one.

• The models share the same structure and are used to answer questions, summarise documents, write code, translate languages and analyse images.
• Today, the most popular and widely known foundation models are GPT, Gemini, Claude, and LLaMA.
• Companies have many special uses built atop foundation models, ranging from enterprise technology to medicine and education to finance.

Reinforcement Learning Models

A reinforcement learning model incrementally learns from trying things out instead of labelled examples. Over time, they practise the right course of action, sometimes rewarding them for decisions and sometimes correcting them for poor ones.

• They are frequently used in robotics, self-driving cars, AI-powered game playing and the industrial sector.
• The system continuously learns by interacting with the surroundings and making changes to its behaviour based on outcomes.
• Reinforcement learning is one of the different AI models leveraged when making decisions when all examples are known, but the more they try, the better they get.

AI Learning Models: Supervised vs. Unsupervised

There are different types of learning used by different AI models. Some will be labelled, with the correct answer provided, and others will find patterns of their own. These learning strategies are critical to the current methods that make up AI and play a key role in a model’s precision when it comes to various tasks.

Supervised Learning, Simplified

One of the most popular supervised learning methods in artificial intelligence is the model-based method, where the model is trained with labelled examples. Consider this to be studying as if there were an answer key to follow before testing.

• The model was trained on pre-acquired data sets of classifications and true classifications.
• It slowly “learns” the correlation between the input and output, becoming better and better at predicting output on new data.
• Some applications of supervised learning include email spam filtering, insurance companies’ credit risk analysis, disease prediction and image categorisation.
• Many AI learning models are adopted in business applications for their high accuracy, especially with supervised learning models.

Unsupervised Learning, Simplified

In unsupervised learning, the data is not labelled. Instead, the model is allowed to explore the data on its own and seek out parts of the data set that, for lack of a better term, humans might not instinctively be able to detect.

• The model clusters similar data items into groups without hand-rigorous knowledge of what the “correct answer” is.
• Unsupervised learning is applied in businesses to segment customers, recommend products or services, detect fraud, and perform market basket analysis.
• This is especially useful with large amounts of untagged data.
• There are several different AI models that, by using unsupervised learning along with others, improve the performance.

When to use each approach

The type of learning you select is dependent upon the problem you’re attempting to solve. There is no one “best” way; each works best with certain types of data and business goals.

• The best use of supervised learning is when a sufficient amount of historical data is available, and accurate predictions are needed.
• Unsupervised learning is best suited for finding hidden truths, clustering or looking at new data sets.
• In practice, some sophisticated AI systems may utilise more than one learning model to enhance accuracy and versatility.
• Once you have grasped these methods, it becomes easier to evaluate what each industry uses for different types of AI models.

AI Model Examples You Already Use

If you haven’t programmed any AI models, you’ve definitely come across AI models today. Even though humans don’t realise it, AI is behind many of our daily activities, such as unlocking your smartphone or watching the videos that are considered to be recommended.

ChatGPT and Large Language Models

Large Language Models (LLMs) have changed the manner in which individuals engage with technology. They are not bound by set instructions to perform repetitive tasks but comprehend natural language and respond in detail within seconds, like a human.

• ChatDPT is powered by a massive language model that can respond to questions, summarise information and provide content, as well as help to solve problems with large amounts of text.
• In addition, modern LLM models can translate between languages, develop software code, clarify concepts and carry out research.
• One of the most well-known AI model example applications out there today.
• LLMs are also one of the most cutting-edge types of AI models in the education, business, and software development sectors.

Image Recognition in Your Phone

Ever see the ‘people’, ‘pets’ or ‘places’ sorting feature – automatically done by the phone? And that is image recognition made possible by AI.

• AI algorithms use colours, patterns, facial features, and shapes to make sense of what is in a picture, which can include objects and people made up of images. For images featuring people and objects, AI can analyse the colours, patterns, facial features and shapes to identify what’s contained within the picture.
• AI has also been used in smartphone cameras that detect scenes, take photos in low-light conditions, use facial recognition to unlock the phone and take pictures of faces.
• Also, image recognition is used in hospitals, manufacturing firms, and even for security purposes in diagnostics and quality checks.
• The scenarios illustrate the use of various AI models to enhance everyday technology.

Recommendation Engines on Netflix and Spotify

You often won’t find random suggestions on streaming websites. Rather, they use an artificial intelligence model that analyses user behaviour to suggest films, music, podcasts and TV programmes that you’re more likely to enjoy.

• Recommendations get more and more tuned with continuous learning as more interactions occur.
• The same kind of AI models are used on e-commerce websites to suggest products based on the patterns of browsing and purchasing.
• Today, there are many different AI models, among the most commercially successful being recommendation systems.

A List of AI Models Worth Knowing

Over the last couple of years, artificial intelligence has progressed by leaps and bounds, and powerful language, image, coding, reasoning, and creativity models have been born. To learn about AI for education or business, knowledge of these models is beneficial.

OpenAI Family and GPT-4

One of the most popular examples of modern generative AI is known as the GPT family. They are trained to understand context, generate text, deconstruct questions, write code and help with a variety of professional tasks.

• GPT is employed in software development, customer support, and summarising documents for businesses.
• The wide usage and versatility make it a top choice for any popular AI models list.

Google Gemini

Google Gemini is Google’s multimodal AI solution that leverages text, images, audio and code to better grasp and link them in one system. It is meant to fuel search and productivity applications, as well as enterprise AI options.

• The Gemini system enables reasoning, coding support, content generation, and multimodal understanding.
• It integrates with many Google products, bringing AI abilities to millions of people.
• It’s no surprise that Gemini is considered one of the best AI models for companies that have already adopted Google’s ecosystem.

Meta LLaMA

LLaMA is a family of open-source language models created by Meta to enable the research, innovation and commercialisation of AI. It’s open for everyone, making it particularly appealing to developers and startups.

• For developers, they can not only use LLaMA models to handle target scenarios but also fine-tune them according to specific requirements and tasks.
• With open-source developments, AI has now proliferated in the education, healthcare, research and software development sectors.
• The expanding developer community behind LLaMA is another reason it’s often featured in the latest updates to the list of AI models.

Anthropic Claude

Claude is a family of AI assistants, designed with a strong emphasis on safety, reasoning, and responsible AI behaviour. It is general purpose across its myriad applications for writing, research, coding, and productivity in the enterprise.

• Claude aims to generate useful, reliable, and contextually relevant answers while minimising negative outputs.
• Claude is used for document analysis, customer support and knowledge management in businesses.
• Now, it’s one of the most popular models among the different AI models. available today.

Stable Diffusion and Image Models

Stable Diffusion revolutionised how anyone creates digital artwork by generating high-quality images out of pretty simple text prompts. It brought more awareness in the creative sector to one of the most remunerative aspects of having AI generate images.

• Stable Diffusion can be applied to create illustrations, product imagery, or concept art for designers, marketers, educators, and content makers.
• Today, there are numerous AI model generator platforms providing similar image generation capabilities.

AI Model Generator Tools: What They Do

In the last several years, artificial intelligence model generators have made artificial intelligence more accessible than ever. Today, you cannot always create a more complicated program or develop a model from the ground up. Many platforms enable the creation or modification of an AI and utilise simple interfaces, providing access to AI for businesses, students, and professionals.

Text Generators vs. Image Generators

Not all AI generators are used for the same purpose. Some specialise in developing textual materials, and others develop visuals, videos, music or even software code when users, or user prompts, request it.

• An AI model generator generates articles, emails, reports, code snippets and answers based on natural language prompts.
• Image generators convert text into a realistic image, illustration, logo, artwork or marketing creatives in seconds.
• These platforms are going multimodal – many are becoming able to manipulate text, images, audio and video together.
• The best choice is determined by users’ usage, creative objectives and workflow needs.

No-Code AI Generator Platforms

The creation of AI models is currently accessible to anyone with some basic expertise. No-code platforms have opened up the door for the development of AI-powered solutions without needing to write large blocks of code.

• These platforms provide drag-and-drop user interfaces which make it simpler to prepare your data, train your models, check them, and deploy them.
• No-code AI for Businesses: Customer service, predictive analytics, workflow automation, and document processing.
• They streamline the development process and increase the accessibility of AI for non-technical users.
• If you are new to AI, no-code AI platforms are a great starting point since you don’t need to undergo the process of learning how to build an AI model using AI programming frameworks.

AI Models vs. Machine Learning Models

Raised as one of the most confusing from the realm of AI vs machine learning models. These words are frequently used synonymously, but they may not have the same meaning. Once you grasp this, you may have a better idea about what an AI model is and how contemporary AI technologies are created.

Where They Overlap

Machine learning is a subset of artificial intelligence, and therefore, all machine learning models are part of artificial intelligence. AI is a large domain, however, containing reasoning systems, planning, robotics and other forms of intelligent systems.

• The machine learning models are integral to most of the AI applications people use today.
• Under the umbrella of AI are the technologies of deep learning, natural language processing, computer vision and recommendation systems.
• With every advancement in AI technology, machine learning will play one of its pivotal roles.
• This explains why there is often a mix between the terms ‘parallel’ to each other when assessing different AI models.

Key Differences to Know

While closely related, AI models are not necessarily machine learning models and vice versa. When deciding what type of technology to use for specific projects, it’s important to be familiar with the differences.

• AI is a catch-all term for making machines capable of intelligent actions.
• Machine learning specifically aims at learning patterns from the data without manual programming.
• All machine learning models are AI models, but not all AI models are based only on machine learning.
• It helps solve one of the most common issues that might arise after learning what an AI model is.

How to Build an AI Model from Scratch

Once grasping the concept of what an AI model is, many newbies are tempted to form their own. Fortunately, the technology of creating a simple AI model has become more accessible with the development of various frameworks, cloud platforms, and tools. The process is logical, although advanced models require a lot of computing power.

Step 1: Define the Problem

Any successful project dealing with AI starts with a well-defined goal. The most sophisticated model is not likely to be useful without a problem it is solving.

• Determine exactly what the AI model should do, such as forecasting sales, image recognition, language translation, fraud detection, etc.
• Making it clear what you are focusing on will help figure out what data, algorithms and evaluation will be necessary.
• Well-defined problems also save unnecessary development time and help to make the project more successful.

Step 2: Gather and Prepare Data

All AI models are based on data. Without accurate, complete, and unbiased training data, no complex algorithm will work very well.

• Gather appropriate data from reliable sources that show an accurate account of the problem being solved.
• Prepare and clean the data, meaning drop missing data, fix errors, and get rid of duplicates, prior to training.
• Data is properly prepared, and using it speeds up the AI model training, which in turn improves the accuracy of the AI.
• The amount of time spent coding and preparing data for most projects that involve AI is drastically less than the amount of time spent on all other activities in that project.

Step 3: Choose a Framework

Once all the data is completed, the next step is to select the right media! The good news is that there are a number of viable frameworks that ease the journey of developing AI.

• These are commonly used frameworks such as TensorFlow, PyTorch, Keras, Scikit-learn, and Hugging Face Transformers.
• There will be distinct benefits in each case for language as well as images, forecasting, or deep learning.
• Many people prefer to use pre-trained models as a first step and then come to creating their own.

Step 4: Train, Test, and Iterate

Creating an AI model is an ongoing journey. A developer continually tweaks the model based on the performance, errors, etc. noted by the developer while learning the model and continually modifies the model with new, better data.

• Use an existing data set you’ve prepared and observe the model metrics as it is trained.
• Test the model using test data that were not used in training the model to check its predictions.
• If necessary, make additional tunings on parameters and the data set, and retrain to boost accuracy.
• Continuous improvement is one of the key factors of this successful AI model training projects.

What Makes the Best AI Model for Your Needs?

No single model is perfect for any circumstance. The ideal AI model will truly rely on the choice targets, available resources, budget and type of problem that you want to resolve. It can be good for generating content, but it may not be the best for making medical decisions or making financial projections.

Accuracy vs. Speed Trade-offs

There are AI models that are developed specifically to be accurate, and there are those that are developed specifically to respond quickly. The type of balance required will be influenced by the use to which the model is put in real-world situations.

• It might be that an enterprise application always craves accuracy, even if it takes some time to generate the response.
• AI-driven customer front-facing tools like chatbots typically prioritise speed to enhance customer experience.
• Establishing performance and efficiency are crucial factors to consider when choosing an AI solution that works best for your business.

Open Source vs. Proprietary Models

A second crucial question is whether an open-source-based model or a commercial proprietary one will be used. Each of these techniques offers benefits based on the technical expertise as well as business needs.

• The open-source models have more flexibility, transparency, and development options.
• Proprietary models generally have more support and managed, enterprise-grade reliability.
• The appropriate option will depend on the budget, the scalability needs, security concerns, and the long-term maintenance needs.

Matching the Model to the Task

However, when selecting an AI model, it is important to first define the problem you are trying to solve, not just the latest and hottest technology.

• Language models are very good when writing, summarising, translating or conversing.
• Image recognition, quality tracking, and medical imaging are the areas of application where computer vision models work best.
• Predictive models help with forecasting, catching fraud, and doing business analysis. Selecting a correct model for better performance also reduces avoidable development costs, and then you don’t waste time.

Risks and limits of AI models

Even though AI power has, sort of, reshaped industries and improved a lot of daily life, it’s still not flawless, not really. The influence of AI is real, and it does enhance everyday tasks, but it remains imperfect. Like other technologies, there are drawbacks involved, and users, developers, and businesses should know about them before they make important decisions around using AI. If you understand these challenges, you can use AI in a more responsible way and keep your expectations in a reasonable lane.

Bias in Training Data

If a model is wrong, it will be wrong in the same way too. The training data can be biased or simply not diverse enough, and that can quietly create unintentional skew in what the model outputs or how it “sees” things.

• AI models can end up predicting more favourably on one slice of data than on another if that data is already biased.
• The way training data is modelled is regularly reviewed and improved, so there’s more balance and inclusivity built in.
• Ethical AI development is largely about cutting down bias and increasing fairness, plus transparency and accountability.
• More sturdy AI systems across different industries are being developed because practices for AI model training keep getting improved.

Hallucination and Factual Errors

Generative AI models are sometimes able to give answers that seem believable but aren’t completely accurate. This is called an AI hallucination and is quite a common occurrence.

• Machine learning algorithms predict what the reader is more likely to look for an answer to, rather than ensuring factually accurate responses.
• Important data, such as medical, financial, educational, or legal information, should be double-checked by the user at all times.
• There are no AI models that are 100% accurate today, but the continuous improvements of the model diminish hallucinations.
• With accuracy crucial, human review is still necessary.

Privacy and Data Security Concerns

In a time when AI is being used more and more in companies and in normal daily life, data privacy seems to matter a lot. It’s worth noting that organisations should manage the information inside an AI system carefully, kind of responsibly, not just casually.

• Firms really need to stick to solid privacy guidelines and also to the laws that apply. Also, anything sensitive should not be sent over to an AI platform that doesn’t offer a solid guarantee on the security safeguards.  
• More and more, organisations are leaning toward private AI models, so the sensitive material stays in their own internal systems. At the same time, businesses are using private AI models more frequently to protect sensitive data while it remains inside their internal environments.
• In the task of creating AI, there is a need to weigh innovation against user privacy, security, and trust.

Conclusion

So if you now ask, What is an AI model? Hopefully, now it should be much clearer. An AI model is a system that learns from data, identifies patterns and applies that learning to make predictions, solve problems or generate new content.

We covered the types of AI models, learnt about AI model training, looked at an AI model list, discussed several real-world examples, and even walked through how to build an AI model from scratch. More importantly, you’ve seen that there isn’t a single “perfect” model. The best AI model always depends on the task you’re trying to accomplish, the quality of your data, and the outcomes you expect.

Whether you are a student, a developer, a business owner or merely an AI enthusiast, the knowledge of these underlying concepts will prove useful as AI technology advances. And as you learn about AI models from Amquest Education, you understand the innovations helping to design the future.

FAQs

What is an AI model?

An AI model is software that learns from data, recognises patterns and trends, makes predictions, and categorises information or generates content. AI model training increases its accuracy, and continuous improvement completes it.

How does an AI model work?  

An AI model gets trained on a large set of data, and then it uses that information to pick up patterns. After that, when it sees new inputs, it relies on what it learnt to guess outcomes and/or generate responses, depending on the task.  

What are the different types of AI models?

The major model categories exist: discriminative, generative, foundation, reinforcement learning and other task-specific AI learning models.

What is the difference between an AI model and an algorithm?

The algorithm is the mathematical approach to learning, and the AI model is what is created upon applying the algorithm to data.

What are AI models used for?

In practice , AI algorithms end up in chatbots, recommendation systems, fraud detection, and diagnosing medical conditions , plus image recognition, language translation, autonomous systems, and more. Basically, they show up everywhere.  

So what about a large language model (LLM)?  

Large Language Models, or LLMs, are a type of generative AI. They get trained on huge amounts of text, so they can learn to grasp and also produce human language. You’ll often hear about ChatGPT, Gemini, Claude, and LLaMA, as well as a few other well-known options.