user‑friendly – удобный для пользователя
cutting‑edge – передовой
time‑saver – экономящий время
game changer – коренным образом меняет ситуацию
on the flip side – с другой стороны
intuitive – интуитивный
to enhance – улучшать
a double‑edged sword – палка о двух концах
b.Do you believe AI poses significant risks to society, or do the benefits outweigh the potential harm? Explain your position.
to be considered / thought / believed / beneficial – считаться полезным
seems to be detrimental – кажется вредным
transparent – прозрачный
to appear to be a necessary evil – оказаться неизбежным злом
to open a can of worms – породить множество новых проблем
outweigh – перевешивать
to tend to be skeptical about– быть склонным судить о чем-то скептически
to tip the scales – склонить чашу весов
unchecked – неконтролируемый
ethical – этичный
c. What measures do you think should be taken to ensure AI is used responsibly in sensitive areas like healthcare or security?
guideline – руководство (правило)
accountability – подотчетность
strict oversight – контроль
to lay down the law – установить жесткие правила
to draw the line – провести границу
an adequate /ˈædɪkwət/ safeguard – адекватная мера предосторожности
consent – согласие
a slippery slope – скользкий путь
robust – надежный
Use one of the following speaking-1 templates:
Basic Template:
Main Point (choose one):
- "I support the view that…"
- "I believe it is preferable to…"
- "I find it a great/poor idea to…"
Transition (optional):
- "I hold this opinion for several reasons."
First Reason:
- "To begin with…" + "For instance…" (followed by a personal example).
Second Reason:
- "Additionally…" + "To elaborate…" (and include more details).
- [To talk about a contrasting point] On the other hand,...
Advanced Template:
Personally, I firmly believe [MAIN POINT].
This is mainly due to [REASON 1].
By this, I mean [expand on the reason].
Furthermore, [REASON 2].
For example, [expand on the reason or provide an illustrative example].
3. Integrated Speaking_2
Active vocab
Step1. Read the following announcement:
Read: 1:00
From: Office of Campus Safety
Date: August 15, 2025
In order to enhance security and streamline campus access, the university will be installing AI‑enabled surveillance cameras at building entrances and common areas beginning next week. These devices use facial recognition technology to identify authorized individuals and automatically notify campus security of any guests or unusual activity. The goals of the system are to reduce unauthorized entry, ensure faster response times in emergencies and provide an extra layer of protection for our community. Footage will be stored securely in accordance with university privacy policies, and only authorized personnel will have access to the data. If you have questions or would like to opt out of facial recognition, please contact the Office of Campus Safety by September 1.
Step 2. Listen to the audio file below. During the real exam you will hear the audio only once. But now you can listen to it as many times s you want.
This announcement makes me feel really uneasy. It feels like we’re being watched all the time, like some sort of “Big Brother” situation. The announcement says the footage will be secure, but who actually knows where our images will endup? I don’t want third parties knowing who I spend time with or how late I leave the library. Being on campus should feel like a safe space, not a place where every move is tracked and analyzed. Instead of installing facial recognition the university could hire more human security staff or improve lighting around campus. This technology definitely crosses a line for me in terms of privacy!
Step 3. Prepare and record (in a messenger) an answer to the following question.
Preparation time - 30 seconds, speaking time - 60 seconds.
The woman expresses her opinion of the university’s announcement. State her opinion and the reasons she gives for holding that opinion
- "There are two reasons why the woman supports/objects to this ...."
- "The woman has negative feelings about this decision." (for mixed-opinion scenarios).
First Reason:
- "To start with, she states that…"
Second Reason:
- "In addition, she points out that…"
Advanced Template:
The announcement/article discusses [the proposed change].
This is partly due to [reason 1], as well as [reason 2].
The woman promptly voices her support/objection to this change.
Initially, she claims that [details], elaborating further that [additional details].
Moreover, she seems to believe that [details]. From her perspective, [more detailed reasoning].
Integrated speaking_3
Step1. Read the text
Reading time: 45 seconds
Read: 00:45
Challenges of new technology
Advances in artificial intelligence are often limited not only by algorithms but also by the hardware that runs them. Large neural network models require enormous amounts of computational power, specialized chips and electricity. As a result, training state of the art systems can cost millions of dollars and strain energy grids. According to policy researchers, modern AI is “data hungry” and “energy intensive,” and the demand for new data centers and larger training runs collides with already strained electrical infrastructure and geopolitical supply chain issues. Memory bandwidth and storage can also become bottlenecks, slowing down the processing of massive datasets. These hardware constraints have prompted researchers to explore more efficient architectures, such as spiking neural networks and neuromorphic chips, and to invest in better energy management. Understanding hardware limitations is therefore essential for evaluating the feasibility and impact of AI systems.
Source: cset.georgetown.edu
Step 2. Listen to the lecture
Note: Prepare a sheet of paper and a pen/pencil. Take notes of illustrations of the ideas given in the reading!
Let me give you an example to illustrate how hardware can constrain AI. A few semesters ago, our engineering department wanted to develop a deep learning model to predict extreme weather events. The idea was to train a neural network on decades of climate data. But when the team calculated the resources they would need, they discovered that a single training run would require dozens of high end GPUs running continuously for weeks. Not only would that have cost a small fortune in hardware, but the energy required would have been equivalent to powering several homes. In fact, a recent policy brief noted that the demand for new data centers to run large models is clashing with an already strained energy grid.
The researchers then tried to run a smaller version of the model on a lab server, only to hit another bottleneck: memory bandwidth. Feeding terabytes of climate data through the GPUs took so long that the model could not be updated in a reasonable time frame. To work around these limitations, the team explored more efficient algorithms and considered using neuromorphic chips, which mimic the brain’s spiking behavior and consume far less power. They also learned about “inference time” techniques—using a smaller model that thinks longer on each prediction—to reduce the need for massive training runs. The project ultimately moved forward, but only after the students realized that cutting edge AI isn’t just about clever code; it’s also about having the right hardware and managing the costs and energy demands that come with it. Source: cset.georgetown.edu
Summarize the points made in the lecture, and explain how they illustrate the ideas expressed in the reading.
Step 3. Prepare and record your answer (send in a messenger)
“The professor elaborates on this concept/these ideas providing illustrations from real life.”
First Example/First Part
“To begin with, he/she says/posits that…”
Second Example/Second Part
“Next, he/she says that…”
Advanced Template:
According to the reading, [title of reading] is [definition].
The professor challenges/elaborates on these statements.
First, (s)he explains/illustrates [details from example 1/ point 1]. Moreover, [more details from example 1/point 1].
The second thing/example (s)he draws is [details from example 2/point 2]. Finally, the professor goes on to say/mention... [more details from example 2 /point 2].
Get ready: 00:30
Integrated speaking_4
As of 2025 there is no longer any reading passage in this speaking task.
You should focus on defining the concept, then presenting the examples or perspectives given by the lecturer.
Step1.
Note: Prepare a sheet of paper and a pen/pencil. Take notes of illustrations of the ideas given in the reading!
Today I’d like to discuss Artificial General Intelligence, or AGI. AGI is the idea of a machine that can perform any intellectual task a human can, not just narrow tasks like playing chess or recognizing faces. Unlike today’s specialized systems, an AGI would need to learn and reason across many domains.
The first major hurdle is technical. Recent advances in large language models and deep learning might suggest we are close, but these systems still lack common sense reasoning and the ability to learn from a few examples. Simply making models bigger does not guarantee progress; some researchers now observe that “scaling laws” may offer diminishing returns. That means throwing more data and compute at a problem isn’t enough. To get closer to human like reasoning, scientists are exploring hybrid architectures that blend neural networks with symbolic reasoning or causal models.
A second challenge is alignment—ensuring that an AGI’s goals are compatible with human values. Current systems occasionally exhibit biases or pursue unintended objectives because they optimize only what they’re explicitly told to optimize. With an AGI, a misaligned goal could have far reaching consequences. Although researchers are working on value alignment techniques and control mechanisms, there is no consensus on how to encode abstract principles like fairness or respect into machine behavior.
The third obstacle is our incomplete understanding of human intelligence. Neuroscience and cognitive science have mapped brain regions and proposed models of memory and attention, but there is no unified theory explaining how consciousness and general reasoning emerge. Without a clear blueprint for intelligence, it is difficult to build an artificial version. Efforts to develop cognitive architectures and tackle the “symbol grounding” problem aim to bridge this gap, yet much remains unknown.
These three hurdles—technical limitations, alignment issues and gaps in cognitive theory—illustrate why AGI remains a long term research goal rather than an immediate prospect.
Source: cset.georgetown.edu
Explain how the examples of El Niño and La Niña illustrate the concept of large-scale ocean-atmosphere interactions. Use information from both the lecture and your own notes to support your response. You will have 20 seconds to prepare your answer and 60 seconds to speak.
Step 3. Prepare and record your answer (send in a messenger)