AI Script 2

Download as doc, pdf, or txt
Download as doc, pdf, or txt
You are on page 1of 6

[Opening scene: Animated visuals of futuristic technology and AI-related imagery]

Narrator: In recent years, the media has fueled our imaginations with tales of superintelligent machines,
depicting a dystopian future where terminators roam the Earth, raining havoc upon humanity. But is this
portrayal of AI grounded in reality, or are we falling victim to an exaggerated narrative? And also what
exactly is AI, how dangerous is it and how does it work?

[Background music fades in]

Narrator: "To understand AI, we must first comprehend the foundation upon which it stands: machine
learning. Currently there are three branches to teaching computers to become extraordinary learners, we’ll do
this through the lens of recognising our furry friends: cats and dogs”

[Scene of a teacher, with a baby robot with a nappy showing examples of labelled cats and dogs]

Narrator: “Supervised learning. Supervised learning is like having a helpful teacher guiding the learning
process. We show the computer countless pictures of cats and dogs, labelling each picture with the correct
answer. The computer learns to associate specific features such as cats have pointy ears and dogs have
floppy ears. Now when we give it a new picture it confidently predicts wether it is a dog or a cat.”

[Scene of robot baby sifting through stack of images, seeing similarities and differences]

Narrator: “Unsupervised learning. Unsupervised learning, is like a curious explorer discovering hidden
patterns on its own. To do this we provide the computer with a massive collection of unlabelled cat and dog
pictures. It sifts through the data autonomously finding its own patterns, such as cats have whiskers, while
dogs have broader snouts. By clustering similar images together, the computer can separate cats from dogs
without any labels. Unsupervised learning uncovers hidden structures and can help us understand the world
in a new light.

[Scene of a game showing baby computer get things right with a nice tune (happy face) and getting things
wrong with a incorrect tune (sad face).]

Narrator: “Reinforcement learning. Reinforcement learning, is like a playful adventurer, our computer begins
by guessing randomly, sometimes hitting the mark, sometimes missing.Through trial and error, guided by
rewards and punishments, its neural network evolves. With each correct identification, the computer receives
positive feedback. Over time, it discerns crucial features, mastering the art of image recognition.

[Visual representation of the different machine learning algorithms described above in a row ]

Narrator: "Overall machine learning algorithms analyse vast amounts of data, looking for patterns and
relationships. By identifying these patterns, the algorithms can make accurate predictions or take actions
based on new, unseen data."
[Visual of neurons communicating with each other and an image of electrical neurons doing the same, image
below from Wikipedia (Artificial Neural Network)]

Narrator: "In the human brain, each neuron interacts with other neurons
receiving, processing and transmitting information to each other. In a similar
way an artificial neural network receives information from an input layer
processes it through at least one hidden layer, and passes the result to an output
layer. In the simplest form, a neural network is a machine learning model that
consists of an input layer, a hidden layer and an output layer.

Side note, the first implementation of an artificial neural network was the Perceptron Mark I in
1958"

[Visual representation of deep neural network using a tower with bricks compared to a shorter tower which is
the traditional neural networks]

Narrator: "Deep learning is where things get really interesting, by creating neural network architectures (or
stacking neural networks) we can better simulate how human brains work. The term deep comes from the
number of hidden layers which is up to 150, as compared to 2-3 hidden layers in a traditional neural
network"

[Visual: Use extracted image below from Convolutional Neural Networks, to create a visual. Then pan to a
baby computer (representing machine learning) and adult computer (representing Deep learning), baby
computer gets overwhelmed smoke comes out of his head and the adult computer gets a graduation cap
dropped on his head)
Narrator: " There are many of Deep learning algorithms, but we cannot cover all of them in this video. But
in short Deep Learning algorithms use these networks to process and analyse complex data, much like
unsupervised learning, this is referred to as end to end learning. Through multiple layers of computation,
they can extract high-level features and learn intricate representations of the input data. As you add more
data, deep learning techniques can continue to improve their performance, this is different from traditional
machine learning that reaches its peak in performance and doesn’t improve much with more data "

[Transition to visuals of massive amount of data, like the matrix, then transition to a huge stack of dvd’s]

Narrator: "One of the recent breakthroughs in the field of AI is the availability of large and complex datasets.
In the last two years alone, we have generated a staggering 90% of the world's data! That equates to 2.5
million Terabytes of data created every day, that’s the equivalent of about 125,000 HD 4K Movies."

[Show evolution of the adult computer, from graduate, nurse to doctor]

Narrator: "This wealth of data fuels the advancement of AI algorithms, allowing them to learn and make
more accurate predictions. The more data available, the better equipped these algorithms are to identify
patterns and make informed decisions."

[Transition to visuals explaining Narrow/Weak AI and Artificial General Intelligence/Strong AI]

Narrator: “It is no secret that the media has a knack for sensationalism. The allure of captivating stories,
coupled with the unknown potential of AI, has given rise to a captivating narrative, one that stokes our fears
and enthrals our collective consciousness. We are bombarded with images of menacing robots, apocalyptic
scenarios, and cautionary tales of AI gone rogue. But amidst the spectacle, the true nature of the challenge at
hand often remains obscured.”

[Visual representation of Narrow/Weak AI]

Narrator: "The development of General Artificial Intelligence is an intricate puzzle, an intricate puzzle that
the brightest minds of our time tirelessly endeavour to solve. It involves creating a machine capable of
reasoning, understanding, and learning across a wide range of tasks—an entity that mimics human-like
intelligence. Yet, despite our remarkable progress, we are still a long way from achieving this elusive goal.”

Narrator: “What we have today is far from the all-powerful General Artificial Intelligence of science fiction.
Instead, we are working with what is known as Narrow AI — specialised systems designed to excel at
specific tasks. These remarkable creations are certainly impressive, they have revolutionised industries, from
medicine to transportation, but they are far from achieving the capabilities that would allow them ability to
reason, adapt to new situations or learn new skills outside of their programming.”

Narrator: "Narrow AI refers to AI systems that excel in specific tasks, such as facial recognition, chat bots
(such as chat GPT), picture / art generation, or autonomous driving. These systems are highly proficient in
their domain but lack the ability to adapt to new situations or learn new skills outside of their programming."
[Visual representation of Artificial General Intelligence/Strong AI]

Narrator: "To create General Intelligence, we must not only understand the intricacies of intelligence itself
but also grapple with the profound puzzle of consciousness. This phenomenon, the subjective experience of
being aware and self-aware, remains one of the greatest mysteries of our existence. It is a puzzle with no
clear solution—a problem that has stumped even the brightest minds in human history. To assume we are on
the verge of solving consciousness is to undermine the gravity of the problem"

Narrator: “One such phenomenon that highlights the challenge, is the propensity for AI systems to
experience hallucinations—unintended and often bizarre outputs that arise from the complex interactions
within their neural networks. In their pursuit of pattern recognition, they can sometimes over-interpret or
misinterpret the information presented to them.“

[Transition to visuals of automation and technology]

Narrator: "Over the past few decades, automation and technology have rapidly advanced, revolutionising
various industries. This progress has brought many benefits, such as increased productivity and efficiency.
However, it has also presented significant challenges."

[Visual representation of changing job market]

Narrator: “The job market is undergoing a profound transformation. Current predictions indicate that around
300 million jobs could be lost or degraded due to Artificial Intelligence. A key factor influencing job
susceptibility is the level of routine and predictable tasks involved. Occupations that primarily rely on
repetitive activities, such as those in transportation or administrative support, are particularly vulnerable to
automation.
[Visual representation of job displacement]

Narrator: "Technological advancements have empowered machines to replicate many human capabilities.
Tasks that once required human intelligence, perception, or manipulation can now be performed by AI
systems and robots."

[Conclusion scene: Animated visuals of a diverse workforce and collaborative technology]

Narrator: “While that all sounds very bleak, it's important to note that automation doesn't have to be a threat.
Instead, we can embrace the collaboration between humans and machines. This means fostering lifelong
learning, cultivating skills that complement technological advancements, and preparing for the jobs of
tomorrow.”

“The impact of AI on the job market cannot be ignored. However we believe a more present danger is AI
bias, AI bias refers to the systematic errors or unfairness that can occur when machine learning algorithms
make decisions. These biases can emerge from the training data used to train the AI models, reflecting and
amplifying the biases present in society. The consequences can be far-reaching, as biased algorithms can
perpetuate discrimination, reinforce stereotypes, and disadvantage certain groups of people.

“Imagine a scenario where an AI algorithm used for diagnosing a particular disease exhibits bias towards
certain demographics. If the algorithm is trained on a dataset that is not diverse and representative, it may not
accurately identify symptoms or make accurate diagnoses for individuals from underrepresented groups.”

Narrator: “This bias can have devastating consequences. Patients from marginalised communities might
receive incorrect diagnoses, leading to delayed or inappropriate treatment. Their conditions may worsen, and
lives could be lost due to algorithmic bias.”

“There are ways we could mitigate this bias. Firstly AI algorithms should be trained on data representing
various demographics and backgrounds. This helps to mitigate biases rooted in data imbalances and
promotes fairness. “

“Secondly randomisation in decision-making can reduce bias. By introducing random elements into the
decision process, algorithms can avoid favouring certain groups or perpetuating discriminatory patterns.”

“Finally, blinded testing of algorithms ensures unbiased evaluation. Similar to blind testing in clinical trials,
where researchers are unaware of which participants received the active drug, algorithms should be tested
without knowing the protected attributes of individuals.”

However it’s essential to consider the challenges and limitations of applying these solutions. Ensuring
diverse and representative data, balancing the trade-offs between randomness and predictability, and
addressing potential pitfalls are all critical aspects that require further exploration.

Therefore it’s essential that we approach AI development with ethical considerations in mind. As AI systems
become more autonomous, we need to ensure they adhere to principles of fairness, transparency, and
accountability. By implementing robust regulations and ethical frameworks, we can mitigate the risks
associated with AI and create a future that benefits all of humanity.

So, as we embark on this journey into the realm of AI, let's remember that the true power lies in our ability to
harness these technologies for the greater good. By embracing the opportunities, addressing the challenges,
and prioritising the well-being of individuals and society, we can shape an AI-driven world that is both
prosperous and equitable.

[Closing scene: Animated visuals of futuristic AI applications and positive human-AI collaboration]

Narrator: “As we traverse the vast cosmos of knowledge together, remember: "I may entertain, but
knowledge I can't guarantee. So, delve into the depths, explore the unknown, for wisdom awaits beyond my
video spree. Stay curious!" But most importantly… let's hope I didn't mispronounce anything!

[Background music fades out]


Papers used:

The future of employment:


https://www.oxfordmartin.ox.ac.uk/downloads/academic/The_Future_of_Employment.pdf
AI bias:
https://link.springer.com/content/pdf/10.1007/s43681-022-00138-8.pdf

Book: The Second Machine Age

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy