Glossary: AI, robotics, crypto and more

72 terms from AI, robotics, crypto, automation, security and data, explained in plain words.

Actuator Robotics: An actuator is the component that makes a robot or machine move. It converts energy, usually electrical, hydraulic or pneumatic, into motion such as turning a joint, opening a gripper or pushing a valve. The choice of actuator determines how strong, fast, precise and energy efficient a machine can be.
AI agent AI: An AI agent is a system that uses a language model to pursue a goal over several steps instead of answering a single question. It can plan, call tools such as a search engine or a calendar, check intermediate results and adjust its approach. Because agents act on their own, clear permissions, logging and human oversight matter more than with a simple chatbot.
Anonymization Data: Anonymization means altering personal data so that individuals can no longer be identified, not even by combining it with other information. Truly anonymized data falls outside the GDPR, unlike pseudonymized data, which can still be linked back to a person with additional information. In practice this is harder than it sounds, because seemingly harmless details such as postal code, age and gender can together point to a single person.
API (application programming interface) Automation: An API, or application programming interface, is a defined way for software systems to exchange data and trigger functions in each other. Instead of a person copying information by hand, one program sends a structured request and receives a structured response. APIs are the backbone of most automation, because they connect applications reliably without relying on screens or manual input.
Big data Data: Big data refers to data sets that are so large, fast-moving or varied that traditional software can no longer store or analyze them efficiently. Examples include sensor data from machines, transaction records or the clickstreams of online services. The value lies not in the sheer volume but in the questions that careful analysis can answer.
Blockchain Crypto: A blockchain is a shared digital ledger in which transactions are bundled into blocks and linked in sequence using cryptography. Copies are kept by many computers in a network, so no single party controls the record and past entries are very hard to change. Bitcoin, launched in January 2009, was the first widely used blockchain.
Cobot Robotics: A cobot, short for collaborative robot, is designed to work alongside people without a full safety fence. Sensors, limited force and speed, and rounded shapes reduce the risk of injury if contact occurs. Cobots are often used for tasks such as assembly, packaging or machine loading, where humans and robots share the same workspace.
Data governance Data: Data governance covers the rules, roles and processes an organization uses to manage its data responsibly. It defines who owns which data, who may access it, how quality is ensured and how long information is kept. Good data governance is a prerequisite for reliable analytics and AI, and it helps meet legal requirements such as data protection.
Data lake Data: A data lake is a central storage location that holds large amounts of raw data in its original format, whether structured tables, text, images or log files. The data is only prepared when it is needed for a specific analysis. Without clear organization and documentation, a data lake can quickly turn into an unusable collection, sometimes mockingly called a data swamp.
Data warehouse Data: A data warehouse is a central database that collects cleaned, structured data from different business systems and organizes it for analysis and reporting. Unlike in a data lake, the data is prepared before it is stored, which makes queries fast and consistent. It is the traditional foundation for dashboards, key figures and management reports.
DDoS attack Cybersecurity: A DDoS attack, short for distributed denial of service, floods a website or online service with so much traffic from many sources at once that it slows down or becomes unreachable. The traffic often comes from botnets, networks of hijacked computers and connected devices. Such attacks rarely steal data but can cause outages and lost revenue or distract from other intrusions.
Deepfake Cybersecurity: A deepfake is an image, video or audio recording created or altered with AI to make a person appear to say or do something they never said or did. Faked voices are used, for example, in fraud calls that impersonate executives or relatives. Verifying through a second channel, such as calling back on a known number, is often the most reliable defense.
DeFi (decentralized finance) Crypto: DeFi, short for decentralized finance, describes financial services such as lending, borrowing and trading that run on blockchains through smart contracts instead of through banks or brokers. Anyone with a wallet can use them, often without identity checks. The risks include software bugs, hacks, sharp price swings and unclear liability when something goes wrong.
Degrees of freedom Robotics: Degrees of freedom describe the number of independent ways a robot or joint can move, for example rotating around an axis or sliding along it. The more degrees of freedom an arm has, the more flexibly it can position and orient its tool. Many industrial robot arms have six, which is enough to reach a point within their range from any angle.
Digital twin Robotics: A digital twin is a virtual model of a physical object, machine, process or even an entire factory that is continuously updated with real data. Engineers use it to simulate changes, test robot programs or predict wear before anything happens in the real world. The value of a digital twin depends on how accurate and current its data is.
Edge computing Data: Edge computing means processing data close to where it is generated, for example on a machine, in a vehicle or on a local server, instead of sending everything to a distant data center. This reduces delays, saves bandwidth and can keep sensitive data on site. It matters wherever decisions have to be made in fractions of a second, such as in production or autonomous driving.
Embedding AI: An embedding is a list of numbers that represents a word, sentence, image or other item as a point in a mathematical space. Items with similar meaning end up close to each other, even if they share no words. Embeddings power semantic search, recommendations and retrieval systems that find relevant documents for an AI model.
Encryption Cybersecurity: Encryption converts readable data into a form that can only be decoded with the right key. It protects information both in transit, for example when loading a website or sending a message, and at rest on hard drives or in the cloud. With end-to-end encryption, only the sender and the recipient can read the content, not the service in between.
End effector Robotics: The end effector is the tool at the end of a robot arm, the part that actually interacts with objects. Depending on the task, it can be a gripper, a suction cup, a welding torch, a screwdriver or a paint nozzle. Many robots can switch end effectors automatically, so one arm can handle several different jobs.
ETL (extract, transform, load) Data: ETL stands for extract, transform, load, the classic three-step process for moving data between systems. Data is extracted from its sources, cleaned and converted into a consistent format, and then loaded into a target such as a data warehouse. In a newer variant called ELT, the raw data is loaded first and transformed later inside the target system.
EU AI Act AI: The EU AI Act, Regulation (EU) 2024/1689, is the European Union's law on artificial intelligence. It sorts AI systems by risk, bans certain practices outright and sets strict requirements for high-risk uses such as hiring or credit scoring. The law entered into force in August 2024 and applies in stages, with the first bans in effect since February 2025.
Exoskeleton Robotics: An exoskeleton is a wearable frame that supports or strengthens the body's movements. Passive models use springs or dampers, while active ones use motors and sensors. In industry they are meant to reduce strain during lifting or overhead work, and in medicine they help with rehabilitation, for example after a stroke or a spinal cord injury.
Fine-tuning AI: Fine-tuning means taking a model that has already been trained on general data and training it further on a smaller, specialized dataset. This adapts its style, vocabulary or behavior to a specific task, such as legal drafting or customer support. It is far cheaper than training from scratch, but the quality of the added data largely determines the result.
GDPR (General Data Protection Regulation) Data: The General Data Protection Regulation, Regulation (EU) 2016/679, has governed the processing of personal data in the European Union since May 2018. It requires a legal basis for every use of personal data, gives people rights such as access and deletion, and obliges organizations to protect the data. Violations can lead to substantial fines and claims for damages.
Hallucination AI: A hallucination is an output from an AI model that sounds plausible but is false or invented, such as a nonexistent source, a wrong date or a made-up quote. It happens because language models generate likely text rather than look up verified facts. Grounding answers in trusted documents and human review reduce the risk but do not remove it.
Halving Crypto: The halving is a built-in rule of Bitcoin that cuts the reward for creating new blocks in half roughly every four years, or every 210,000 blocks. It limits the total supply to a maximum of 21 million bitcoin. The most recent halving in April 2024 reduced the block subsidy from 6.25 to 3.125 BTC, and the next one is expected around 2028.
Human in the loop Automation: Human in the loop describes automated or AI-supported processes in which a person reviews, corrects or approves certain decisions before they take effect. It is common where errors are costly, such as in medical diagnostics, lending or content moderation. The approach only works if the people involved have enough time, information and authority to actually intervene.
Humanoid robot Robotics: A humanoid robot has a body modeled on the human form, typically with a torso, a head, two arms and often two legs. The idea is that such machines could use spaces, tools and equipment built for people. Walking, balance and dexterous hands remain demanding engineering problems, so most humanoids are still in research or pilot projects.
Industry 4.0 Automation: Industry 4.0 describes the networking of machines, products and IT systems in manufacturing, seen as a fourth stage of industrialization after mechanization, electrification and computer automation. Sensors collect data along the entire production chain so that processes can be monitored, analyzed and partly controlled automatically. The term emerged in Germany in the early 2010s and is now used internationally.
Inference AI: Inference is the phase in which a trained AI model is actually used: it receives an input and computes an output, such as an answer, an image or a classification. Training happens once or occasionally, while inference runs every time someone uses the model. For widely used services, the computing power and energy needed for inference therefore add up over time.
Intelligent document processing (IDP) Automation: Intelligent document processing combines text recognition, language models and business rules to read documents such as invoices, contracts or forms automatically. It extracts the relevant information, assigns it to the correct fields and passes it on to other systems. Unclear or unusual cases are typically forwarded to a person for review instead of being processed blindly.
Large language model AI: A large language model is an AI system trained on vast amounts of text to predict the next word or word fragment in a sequence. From that simple objective it learns to summarize, translate, answer questions and write code. Its answers sound fluent but are statistical predictions, so they can be wrong and need checking.
Layer 2 Crypto: Layer 2 refers to networks built on top of a main blockchain, the layer 1, to process transactions faster and more cheaply. They bundle many transactions off the main chain and post compressed results or proofs back to it for security. On Ethereum, the Dencun upgrade in March 2024 introduced so-called blobs, which lowered fees on layer 2 networks significantly.
Lidar Robotics: Lidar, short for light detection and ranging, measures distances by sending out laser pulses and timing how long their reflections take to return. From many such measurements it builds a detailed 3D picture of the surroundings. Robots, drones and self-driving vehicles use lidar to detect obstacles and determine their position, often combined with cameras and radar.
Low-code platform Automation: A low-code platform lets people build applications and automated workflows mainly through visual editors, drag and drop and prebuilt building blocks instead of writing code line by line. This allows business departments to create simple tools themselves. Without clear rules for security, data access and maintenance, however, many small and undocumented solutions can quickly emerge.
Machine vision Automation: Machine vision refers to systems in which cameras and image processing software help machines see their surroundings and make decisions. In factories it is used to inspect products for defects, read codes, measure parts or guide robot arms to the right position. Modern systems increasingly use AI methods, but they still depend on consistent lighting and clean training images.
Metadata Data: Metadata is data about data. It describes, for example, when a file was created, who wrote a document, where a photo was taken or how long a phone call lasted. Even without the content itself, metadata can reveal a great deal about people and their behavior, which is why it matters for data analysis and privacy alike.
MiCA Crypto: MiCA, the Markets in Crypto-Assets Regulation (EU) 2023/1114, is the European Union's common rulebook for crypto assets. Its rules for stablecoins have applied since June 30, 2024, and the full framework since December 30, 2024. Licensed crypto service providers may offer their services across the EU, and some member states granted existing providers transitional periods of up to July 2026.
Multi-factor authentication (MFA) Cybersecurity: Multi-factor authentication requires at least two independent proofs of identity at login, for example a password plus a code from an app or a hardware security key. Even with a stolen password, attackers cannot easily get in without the second factor. Codes sent by text message are better than no second factor, but app-based codes and security keys offer stronger protection.
Multimodal model AI: A multimodal model can process more than one type of data, such as text, images, audio or video, and often produce several of them as well. It might describe a photo, answer questions about a chart or transcribe and summarize a recording. Combining different inputs gives the model more context, but it also creates new sources of error.
NIS2 Directive Cybersecurity: The NIS2 Directive, Directive (EU) 2022/2555, sets common cybersecurity requirements for companies and public bodies in important sectors across the European Union. It covers far more organizations than its predecessor and requires risk management measures, the reporting of serious incidents and accountability at management level. Member states had to transpose it into national law by October 2024.
Open data Data: Open data is data that anyone may freely access, use and share, usually under an open license. Typical examples include government statistics, weather and geographic data, timetables or research results. Open data supports transparency, research and new services, provided it contains no personal information and is published in machine-readable formats.
Passkey Cybersecurity: A passkey is a login method that replaces passwords with a pair of cryptographic keys. The private key stays on the user's device and is released with a fingerprint, face scan or PIN, while the service only stores the public key. Because there is no password to type in, passkeys are highly resistant to phishing and to leaks from hacked servers.
Penetration testing Cybersecurity: Penetration testing is an authorized, simulated attack on a system, network or application to find vulnerabilities before criminals do. Testers use the same methods as real attackers, but within a clearly agreed scope. The result is a report listing the weaknesses found and how to fix them, which only creates value if the fixes are actually carried out.
Phishing Cybersecurity: Phishing is a type of fraud in which attackers pose as a trusted sender, such as a bank, a colleague or a delivery service, to trick people into revealing passwords or other data or opening malicious files. It usually arrives by email, text message or phone call. Because AI tools make such messages more convincing, checking the sender and the request itself matters more than spotting typos.
Predictive maintenance Automation: Predictive maintenance uses sensor data such as vibration, temperature or power consumption to estimate when a machine is likely to fail. Instead of servicing equipment on a fixed schedule or after a breakdown, technicians intervene when the data shows signs of wear. This can reduce unplanned downtime, but it requires good data and models that have been tested on the specific equipment.
Private key Crypto: A private key is a secret number that proves ownership of crypto assets and is used to sign transactions. Whoever holds the private key controls the associated funds, which is why it must never be shared. If it is lost and there is no backup, the assets usually cannot be recovered by anyone.
Process mining Automation: Process mining analyzes the digital traces that business software leaves behind, such as timestamps and status changes, to reconstruct how processes actually run. The result reveals bottlenecks, detours and deviations from the intended workflow. Companies often use it before automating anything, to find out which steps are worth tackling first.
Programmable logic controller (PLC) Automation: A programmable logic controller is a rugged industrial computer that controls machines and production lines. It continuously reads signals from sensors, processes them according to a stored program and switches motors, valves or conveyor belts. PLCs are built to run reliably for many years, which is why many older units are still in operation and can be difficult to secure.
Prompt AI: A prompt is the input a user gives an AI model: a question, an instruction, an example or a document to work with. How a prompt is phrased and what context it contains can change the quality of the answer considerably. Writing clear prompts with a defined goal, audience and format is often called prompt engineering.
Prompt injection Cybersecurity: Prompt injection is an attack on AI systems in which hidden or manipulative instructions are smuggled into the input a model processes, for example through a website, an email or a document. The model may then ignore its original instructions, leak data or carry out unwanted actions. The risk grows with AI agents that read external content and act on their own.
Proof of stake Crypto: Proof of stake is a method for securing a blockchain in which validators lock up, or stake, their own coins to earn the right to confirm transactions. Dishonest behavior can be punished by the loss of part of that stake. Ethereum switched to proof of stake in September 2022, and according to the Ethereum Foundation its energy use fell by more than 99.9 percent.
Proof of work Crypto: Proof of work is a method for securing a blockchain through computing effort. Miners compete to solve a mathematical puzzle, and whoever finds a solution first may add the next block and receives a reward. Bitcoin uses proof of work, which makes attacks extremely costly but also consumes a large amount of electricity.
Ransomware Cybersecurity: Ransomware is malicious software that encrypts data or locks systems and then demands a ransom for their release. Many groups also steal the data beforehand and threaten to publish it, which is known as double extortion. Regular offline backups, prompt security updates and a tested emergency plan are among the most important safeguards.
Retrieval-augmented generation (RAG) AI: Retrieval-augmented generation is a technique in which an AI system first searches a defined collection of documents and then passes the most relevant passages to a language model together with the question. The model bases its answer on that material instead of relying only on what it learned in training. This makes answers more current and traceable and reduces, but does not eliminate, hallucinations.
Robotic process automation (RPA) Automation: Robotic process automation uses software bots to perform repetitive computer tasks the way a person would, by clicking through user interfaces, copying data and filling in forms. It suits rule-based processes such as transferring invoice data between systems that lack a direct connection. Despite the name, no physical robots are involved, and the bots often break when screens or processes change.
ROS (Robot Operating System) Robotics: ROS, the Robot Operating System, is an open-source framework for developing robot software. Despite its name, it is not an operating system in the usual sense but a collection of libraries, tools and conventions that let the different parts of a robot exchange data. Its current generation, ROS 2, is widely used in research, education and industry.
Seed phrase Crypto: A seed phrase, also called a recovery phrase, is a sequence of ordinary words, usually 12 or 24, from which the private keys of a wallet can be restored. It is effectively the master backup for everything in that wallet. Anyone who learns it can take control of the funds, so it belongs offline and should never be entered on websites or shared with supposed support staff.
SLAM Robotics: SLAM stands for simultaneous localization and mapping. It describes methods that let a robot build a map of an unknown environment while at the same time working out where it is on that map. Robot vacuums, warehouse robots and drones rely on SLAM, using data from cameras, lidar or other sensors.
Smart contract Crypto: A smart contract is a program stored on a blockchain that runs automatically when predefined conditions are met. It can, for example, release a payment, swap tokens or record ownership without an intermediary. Because deployed code is often hard to change, programming errors can be costly, which is why independent code audits are common.
Stablecoin Crypto: A stablecoin is a crypto token designed to keep a stable value, usually pegged to a currency such as the US dollar or the euro. Most are backed by reserves such as cash and short-term government bonds, while algorithmic models try to hold the peg through code. The collapse of the algorithmic stablecoin UST in May 2022 showed that such a peg can fail.
Swarm robotics Robotics: Swarm robotics studies how many simple robots can solve tasks together without a central controller. Each unit follows a few local rules and reacts to its neighbors, and useful collective behavior emerges from these interactions, similar to ants or flocks of birds. Possible uses include search and rescue, environmental monitoring and agriculture, though most applications are still experimental.
Synthetic data Data: Synthetic data is artificially generated data that mimics the statistical properties of real data without containing actual records of real people or events. It is used to train AI models, test software or share data sets without exposing sensitive information. Its usefulness depends on how well it reflects reality, and biases in the original data can carry over.
Teleoperation Robotics: Teleoperation means controlling a robot or machine remotely, with a human operator steering it from a distance. Cameras and sensors transmit what the robot perceives, sometimes with force feedback so the operator can feel resistance. It is used in bomb disposal, deep sea and space missions, and increasingly to collect training data for robots that are later meant to act on their own.
Token (AI) AI: A token is the basic unit of text that a language model reads and writes. Depending on the language and the word, a token can be a whole word, part of a word or a single character. Model limits, processing speed and usage prices are usually measured in tokens, which is why the same text can cost different amounts in different languages.
Tokenization Crypto: Tokenization means representing ownership of an asset, such as a bond, a fund share or real estate, as a digital token on a blockchain. This can simplify settlement and record keeping and make it easier to divide assets into smaller units. Legal frameworks are emerging, and Germany, for example, has allowed electronic securities under its eWpG law since June 2021.
Transformer AI: The transformer is the neural network architecture behind most modern language models, introduced by researchers in 2017. Its core mechanism, called attention, lets the model weigh how strongly each part of an input relates to every other part. Because this work can run in parallel, transformers can be trained efficiently on very large datasets.
Vector database Data: A vector database stores content as embeddings, numerical vectors that represent the meaning of text, images or other data. Instead of searching for exact matches, it finds the entries that are most similar in meaning to a query. Vector databases are a core component of retrieval-augmented generation and other AI applications that need to search large document collections.
Webhook Automation: A webhook is an automatic message that one application sends to another as soon as a certain event occurs, such as a new order, a completed payment or an incoming form. Instead of repeatedly asking whether something has changed, the receiving system is notified right away. Webhooks are a simple way to trigger automated workflows in real time.
Workflow orchestration Automation: Workflow orchestration means coordinating many individual automated steps, systems and services so that they run in the right order as one overall process. An orchestration tool starts tasks, passes data between them, handles errors, retries failed steps and logs every run. It is used, for example, for data pipelines, employee onboarding or complex order handling.
Zero trust Cybersecurity: Zero trust is a security model built on the principle of never trusting automatically and always verifying. Every access request is checked based on identity, device and context, regardless of whether it comes from inside or outside the company network. This limits how far an attacker can move once a single account or device has been compromised.
Zero-day vulnerability Cybersecurity: A zero-day vulnerability is a security flaw that is unknown to the manufacturer or for which no fix exists yet. The name refers to the fact that developers have had zero days to address it. Attackers who exploit such a flaw can bypass protections that rely on known threats, which is why zero-days are highly valued on criminal markets.