Grammar
Kiedyś oglądałem telewizję codziennie, ale teraz oglądam głównie filmy w internecie. Przyzwyczaiłem się do tego, że sam wybieram, co oglądam.
[I used to watch TV every day, but now I mainly watch videos online. I’ve got used to choosing what I watch myself.]
A realistic video in which AI replaces one person’s face with another’s is called a ___. It can be so convincing that viewers cannot tell it is fake.
deepfake
Please ___ ___ the camera so that the software can recognise your face. The session will start in a minute.
turn on
W kinie nie wolno rozmawiać przez telefon, ale można zrobić sobie przerwę w połowie filmu. Kto chce wyjść, musi zrobić to po cichu.
[You mustn’t talk on the phone in the cinema, but you can take a break in the middle of the film. Anyone who wants to leave must do so quietly.]
I’m a program that you talk to by typing messages, and I answer instantly, as if I were a person. Many websites use me to answer customers’ questions at night.
chatbot
Zapytał mnie, czy kiedykolwiek korzystałem z tego programu. Odpowiedziałem, że nigdy o nim nie słyszałem.
[He asked me whether I had ever used that program. I replied that I had never heard of it.]
The quality of an AI-generated image depends largely on the instruction you type in. This instruction is called a ___, and writing a good one is a skill in itself.
prompt
With one click, you can ___ ___ the background noise from the recording. The sound is much cleaner afterwards.
filter out
smooth out
Jego telefon jest wyłączony, więc pewnie jest w samolocie. Może też po prostu rozładowała mu się bateria.
[His phone is off, so he must be on a plane. His battery may also simply have died.]
I’m a type of AI model inspired by the human brain, made of layers of connected nodes. The more layers I have, the “deeper” my learning is said to be.
neural network
Szkoda, że nie mamy większego mieszkania. Gdybyśmy mieli, moglibyśmy zaprosić całą rodzinę na święta.
[I wish we had a bigger flat. If we did, we could invite the whole family for Christmas.]
A recruitment tool kept rejecting female candidates because it had learned from past hiring data. This is an example of algorithmic ___, and it shows why training data must be checked carefully.
bias
I ___ ___ a strange photo while browsing, and I wasn’t sure whether it was real. Later I found out that it had been generated by AI.
came across
Mogłaś mi powiedzieć, że film jest tak długi! Wzięłabym wtedy coś do jedzenia.
[You could have told me the film was so long! I would have taken something to eat.]
I decide which videos appear on your home page by analysing what you have watched and liked. Because of me, no two people ever see exactly the same feed.
recommendation algorithm
Najwyższy czas, żebyś zaczął się uczyć do egzaminu. Wolałbym, żebyś nie zostawiał wszystkiego na ostatnią chwilę.
[It’s high time you started studying for the exam. I’d rather you didn’t leave everything until the last minute.]
The chatbot confidently quoted a book that does not exist, and the journalist almost published the quote. This kind of error, in which AI invents facts, is called a ___.
hallucination
Many platforms have started to ___ ___ on fake accounts that spread misinformation. Users who break the rules are banned permanently.
crack down
come down
To nie mógł być prawdziwy film, bo ten aktor od lat nie żyje. Ktoś musiał go wygenerować, a my nie powinniśmy byli tak szybko w to uwierzyć.
[It can’t have been a real film, because that actor has been dead for years. Someone must have generated it, and we shouldn’t have believed it so quickly.]
I’m trained on enormous amounts of text, and my basic task is to predict which word should come next. I can write articles, translate and answer questions, but I sometimes make things up.
Large Language Model
(LLM)
Nigdy wcześniej nie widziałem tak realistycznego filmu. Dopiero po godzinie dowiedziałem się, że został wygenerowany komputerowo.
[Never before had I seen such a realistic video. Only after an hour did I find out that it had been generated by a computer.]
The newsroom wants to show readers where every image came from and whether AI has edited it. Tracking this history, known as ___, helps the audience trust what they see.
provenance
The studio plans to ___ ___ its production with AI tools, so it can release twice as many videos. Some editors fear that the technology will ___ ___ their jobs altogether.
scale up/speed up
take over
Nie musiałeś zamawiać tylu pizz, bo goście prawie nic nie zjedli. Mogłeś przynajmniej zapytać, ile osób przyjdzie.
[You needn’t have ordered so many pizzas, because the guests hardly ate anything. You could at least have asked how many people were coming.]
I create images by starting with random noise and gradually removing it until a picture appears. Tools such as Stable Diffusion are built on me.
diffusion model