Vocabulary
What is it called when a model confidently gives out false or fictitious information that does not have real facts?
AI Hallucinations
In 1997, the Deep Blue supercomputer defeated the world champion in this game for the first time. What kind of game are we talking about?
Chess
What was the name of the AI that Arnold Schwarzenegger was fighting?
Skynet
What is the main fuel for modern neural networks?
Data/Databases
Explain the meaning of the word "query".
This word refers to any user's request to a language model (text, audio, or visual) that serves as an instruction for generating a response.
What is the name of the test that allows you to determine whether an AI is capable of thinking like a human?
The Turing test
How does Nazarbayev University approach artificial intelligence?
It is allowed to use AI but only with transparency and academic integrity
Neural networks are in great need of these two components, which could be found in any computer. Name these components.
Graphics cards (GPUs) Or Random-access memory (RAM).
What does the word "discrepancy" mean?
Difference between the predicted result of the model and the actual data
When was the chat gpt created?
In 2022 year
Who formulated the "Three Laws of Robotics" that were supposed to guarantee the safety of AI for humans?
Isaac Asimov
In traditional programming, it is written by a human. In machine learning, AI builds this sequence of actions itself by finding patterns in data. What mathematical “instruction” is being referred to?
Algorithm
Try to explain what "deep-learning" means.
A type of machine learning that uses large artificial neural networks with many layers to learn patterns from data.
Free Ride!
Name 5 names of AI (chatbots or models) that are currently popular all over the world.
ChatGPT, Claude, Gemini, Midjourney, Llama and etc.
There is a "Dead Internet Theory". What is it?
The theory goes that most of the content on the web is already created by bots for bots, and there are almost no real people left on the Internet.
What is a "meta-learning" in terms of AI?
An approach where a model is trained not just to solve one specific task, but to quickly adapt to new tasks using prior experience.