Category
A region is a geographic location containing cloud infrastructure; an availability zone is one or more physically separate data centres within that region, with independent power, cooling and
Because cloud pricing charges for many dimensions simultaneously, most of which scale with usage that engineers do not directly control — so cost becomes an emergent property of system behaviour
A situation where the cost of leaving a supplier exceeds the benefit of doing so — not because a contract forbids it, but because migration is expensive, risky or technically impractical. It is
Increasingly, the use rather than the technology — regulators have largely converged on assessing what a system is used for and how much harm it could cause, rather than attempting to regulate the
Data that is generated rather than collected — created by simulation, by statistical modelling of a real dataset, or increasingly by a machine learning model — and used where real data is
Against standardised test sets with published scores — and the honest position is that benchmarks measure something real, measure it narrowly, and are increasingly unreliable as headline comparisons
Because every single response requires substantial computation on expensive, power-hungry hardware — and unlike most software, the cost does not fall as usage grows, which breaks the economics people
By training a model to map one person's appearance or voice onto another's performance — and detection is possible but increasingly unreliable, which is why the response has shifted toward provenance
An agent pursues a goal over multiple steps, deciding what to do next and taking actions in the world; a chatbot responds to messages. The difference is autonomy and the ability to act rather than
No. Open weights means the trained model parameters are downloadable; open source, in its established meaning, requires considerably more — and the distinction is actively contested because a great
Where the computation happens — and the trade-off is between privacy, latency and offline capability on one side, and capability, scale and updatability on the other. Cloud AI. Your input is sent to
A way of recording where a piece of media came from and what was done to it, cryptographically attached to the file — an approach to the synthetic media problem that works in the opposite direction
An attack in which instructions hidden in content the model reads are followed as though they came from the user — and it is a genuinely unsolved problem, not a bug awaiting a patch. Why it exists. A
Three different ways of getting a model to work with particular information, differing enormously in cost, and frequently confused when people ask about "training it on our data". Pre-training.
A token is the unit a language model processes — roughly a word fragment — and the context window is how many tokens it can consider at once. Both determine cost, capability and the limits you run
Because a language model is trained to produce plausible continuations of text, not to state truths — and fluent plausible text and accurate text are different targets that usually coincide and
Obtain valid consent before storing or accessing non-essential information on a device — and most cookie banners do not meet the standard, which is why regulators across Europe have issued repeated
A personal data breach is any security incident leading to accidental or unlawful destruction, loss, alteration, unauthorised disclosure of, or access to personal data — which is considerably broader
An API key identifies an application; an OAuth token represents a user's authorisation for an application to act on their behalf. They answer different questions, and using one where the other is
What you are permitted to do with the software, and — critically for the last of them — what you must do if you distribute your own version. Proprietary. The copyright holder grants a limited licence
Telemetry is automatically collected data about how software and hardware are behaving, sent back to the developer. It ranges from genuinely necessary diagnostics to detailed behavioural tracking,
Single sign-on (SSO) means one authentication grants access to many applications, with the applications never seeing your credentials. A password manager stores many separate passwords and enters
How much of the stack someone else operates. Each model draws the line in a different place, and what falls on your side determines both your workload and your responsibility. Infrastructure as a
Whether the data can still be linked back to a person. Anonymised data cannot, and falls outside data protection law entirely. Pseudonymised data can, with additional information, and remains fully