AI is changing the world around us. It eliminates work and floods the internet with slops. AI is taking over the internet completely thanks to ChatGpt's huge popularity to Google's crammed AI summary at the top of the search results. With AI, you can get instant answers to almost any question. You can feel like talking to someone who has a PhD in everything.
But that aspect of AI chatbots is just part of the AI landscape. Certainly, it's cool to help ChatGpt do your homework or create attractive images of mecha based on the country of origin in Midjourney, but the possibilities of generative AI can change the economy completely. According to the McKinsey Global Institute, it could be worth $4.4 trillion in global economy each year. So you should expect to hear more and more about artificial intelligence.
It has appeared in dizzy products. A short list includes Google's Gemini, Microsoft's Copilot, Claude of Mankind, and Prperxity Search Engine. Read reviews and hands-on ratings of these and other products along with news, explanators and how-to posts on AI Atlas Hub.
New terms are appearing everywhere as people become accustomed to the world of AI intertwining. So whether you're trying to sound smarter than a drink or trying to impress in a job interview, here are some important AI terms you need to know.
This glossary is updated periodically.
Artificial general information, or AGI: The concepts suggest a more advanced version of AI can teach and advance unique abilities while performing tasks that are far better than humans than we know today.
agent: A system or model that illustrates an agency with the ability to autonomously pursue actions to achieve its goals. In the AI context, agent models can act without constant supervision, such as high-level self-driving cars. Unlike the “agent” frameworks in the background, the agent framework is at the forefront and focuses on the user experience.
AI Ethics: Principles aimed at preventing AI from harming humans are achieved through means such as determining how AI systems collect data and address bias.
AI Safety: An interdisciplinary field that relates to the long-term impact of AI and how it suddenly moves into superintelligence that could be hostile to humans.
algorithm: A set of instructions that allow computer programs to learn and analyze data in a specific way, such as patterns, will then learn and accomplish their own tasks.
Alignment: Finely tune your AI to better produce the desired results. This can be referenced anywhere from moderated content to maintaining positive interactions with humans.
Personification: When humans tend to impart human properties to non-human objects. In AI, this involves believing that chatbots are more human and perceived than they are.
Artificial Intelligence or AI: The use of technology to simulate human intelligence, either in computer programs or in robotics. The field of computer science, which aims to build systems that can perform human tasks.
Autonomous Agent: An AI model with features, programming and other tools to accomplish a specific task. For example, self-driving cars are autonomous agents, as they have sensory input, GPS, and driving algorithms that navigate the roads on their own. Stanford University researchers show that autonomous agents can develop their own cultures, traditions and shared languages.
bias: For large-scale language models, errors arise from training data. This can cause certain traits to be misaligned to a particular race or group based on stereotypes.
Chatbot: A program that communicates with humans via text that simulates human language.
chatgpt: An AI chatbot developed by OpenAI using large-scale language modeling technology.
Cognitive Computing: Another term for artificial intelligence.
Data Enhancement: Train your AI by remixing existing data or adding more diverse data sets.
Dataset: A collection of digital information used to train, test and validate AI models.
Deep Learning: AI methods and machine learning subfields use multiple parameters to recognize complex patterns of photographs, sounds and text. This process is inspired by the human brain and uses artificial neural networks to create patterns.
diffusion: A machine learning method that retrieves existing data like photos and adds random noise. The diffusion model trains the network to redesign or recover its photos.
Emergency action: When the AI model shows unintended capabilities.
End-to-end learning, or E2E: A deep learning process in which the model is instructed to perform tasks from start to finish. They are not trained to accomplish tasks in sequence, but instead learn from input and resolve them at once.
Ethical considerations: The ethical implications of AI and recognition of issues related to privacy, data use, fairness, misuse and other safety issues.
Foom: It is also known as fast takeoff or hard takeoff. The concept that if someone constructed an AGI, it might already be too late to save humanity.
Generic adversary networks, or GANs: A generated AI model consisting of two neural networks for generating new data. Generators and discriminators. The generator creates new content and the discriminators check if it is authentic.
Generated AI: Content generation technology that uses AI to create text, video, computer code, or images. AI is fed with a large amount of training data and find patterns that generate new unique responses. This may be similar to the source material.
Google Gemini: The AI chatbot by Google works similarly to ChatGPT, but pulls information from other Google services such as search and maps.
guardrail: Policies and restrictions placed in AI models so that data is processed responsibly and the model does not create disturbing content.
Hallucinations: Incorrect response from AI. You can include answers that generate AI that are incorrect but state with confidence as if they were correct. The reason for this is not entirely understood. For example, you might respond to an AI chatbot with a false statement saying, “When did Leonardo da Vinci paint the Mona Lisa?” and “Leonardo da Vinci painted the Mona Lisa in 1815.” This is 300 years after it was actually painted.
inference: The AI model process is used to generate text, images, and other content about new data. Speculation From their training data.
Large language models, or LLM: An AI model that trains large amounts of text data to understand languages and generate new content in human-like languages.
Waiting time: A time delay from when an AI system receives inputs or prompts and generates output.
Machine Learning, or ML: The AI components allow computers to learn and produce better predictions without explicit programming. It can be combined with a training set to generate new content.
Microsoft Bing: Microsoft's search engine can now use ChatGpt on power supplies to provide AI-powered search results. It's similar to Google Gemini that has an internet connection.
Multimodal AI: A type of AI that can handle multiple types of input, including text, images, video, audio, etc.
Natural Language Processing: A branch of AI that provides computers with the ability to understand human language using machine learning and deep learning.
Neural Network: A computational model that resembles the structure of the human brain and aims to recognize patterns of data. It consists of interconnected nodes or neurons that can recognize patterns and learn over time.
Open weight: When a company releases an open weight model, it exposes the final weight of the model – how to interpret information from training data containing bias. Typically, open weight models can be downloaded for running locally on the device.
Overfitting: It works closely with machine learning error training data, allowing you to identify only specific examples of the data above, but may not be new.
paper clip: The greatest theory of paperclips, created by Oxford University philosopher Nick Bostrom, is a hypothetical scenario in which an AI system creates as many literal documents as possible. With the goal of generating the largest amount of paper clips, the AI system hypothetically consumes or transforms all materials to achieve the goal. This includes dismantling other machines to produce more strings, machines that are beneficial to humans. The unintended consequence of this AI system is that it can destroy humanity with its goal of document creation.
parameter: Numerical values that give the structure and behavior of LLMS allow predictions to be made.
Confusing: Names of AI-powered chatbots and search engines owned by Perplexity AI. It uses a large language model like the one found in other AI chatbots, but is connected to the open internet for up-to-date results.
prompt: Suggestions or questions to enter into the AI chatbot to get a response.
Prompt Chain: The ability of AI to use information from previous interactions to color future responses.
Quantization: A process in which large-scale learning models of AI process smaller and more efficiently (slightly less accurate) by decreasing from higher to lower forms. A good way to think about this is to compare a 16 megapixel image with an 8 megapixel image. Both are still clear and visible, but zooming in gives a higher resolution image more detailed.
Slop: Low-quality online content, produced in large quantities by AI, has gained views with almost all effort and effort. The goal of AI Slop in the realm of Google Search and social media is to flood the amount of content that captures as much advertising revenue as possible, usually in order to undermine real publishers and creators. Some social media sites have accepted the influx of AI slops, but others have been pushed back.
Probabilistic Parrot: Software is an analogy from LLMS that shows that you don't really understand the meaning behind the language and the surrounding world, regardless of how the output is convinced. This phrase refers to the way in which parrots mimic human language without understanding the meaning behind them.
Style Transfer: The ability to adapt the style of one image to the content of another image allows AI to interpret the visual attributes of one image and use it in another image. For example, take a Rembrandt self-portrait and recreate it in Picasso style.
Composite data: Data created by generated AI, trained with real data rather than from the real world. Used for training mathematics, ML and deep learning models.
temperature: Parameters set to control random output of the language model. High temperature means that the model takes more risk.
Generate images from text: Create images based on text descriptions.
token: A small bit of written text that AI language models the process to formulate a response to a prompt. A token can be about four letters of English, or about a quarter of a word.
Training data: Datasets are used to aid in learning that AI models contain text, images, code, or data.
Transformer model: A neural network architecture and deep learning model that learns context by tracking data relationships, such as sentences and images. So instead of analyzing the sentences one word at a time, you can look at the entire sentence and understand the context.
Turing Test: Named after Alan Turing, a well-known mathematician and computer scientist, it tests the capabilities of machines to act like humans. If a person cannot distinguish a machine's response from another person, the machine passes.
Unsupervised learning: A form of machine learning in which no labeled training data is provided to the model and instead the model must identify the pattern of the data on its own.
Weak AI, also known as narrow AI: AI focuses on specific tasks and cannot learn beyond skill sets. Most AI today is weak AI.
Zero Shot Learning: A test in which the model must complete a task without being given the necessary training data. An example is to recognize a lion while being trained with the Tigers.
