Technical essay on Artificial Intelligence & its application in the O & G industry
Weak
or narrow AI is applied to a specific domain, for example, language
translators, virtual assistants, self-driving cars, AI-powered web searches
recommendation engines, etc., but this type of AIs are not able to learn and
perform new tasks. Strong AI or generalized AI is AI that can interact and
operate a wide variety of independent and related tasks. It can learn new tasks
to solve new problems, and it does this by teaching itself new strategies. Super
AI or conscious AI is AI with human-level consciousness, which would require it
to be self-aware. Self-consciousness is not defined.
There
are many definitions based on AI brought by the technology-driven world? It is
defined as how machines can have cognitive capability, i.e., impart the
ability to think and learn new machines and how humans work. It is a set of
technology that identifies patterns of data and reproduces them on new
information. It is a set of mathematical algorithms that enables us to have
computers find intense patterns that we may not have known to have existed
without us being able to hardcode them manually. A study by PWC suggests a 16
trillion dollars increase in GDP by 2030 based on AI.
It
aggregates knowledge from different sources into one centralized cloud and
provides them in an accessible manner & perception. AI systems typically
demonstrate behaviours associated with human intelligence, learning, reasoning,
problem-solving, knowledge representation, and for data scientists, AI is a way
of exploring and classifying data to meet specific goals. It is like an assistant
to which we talk daily on our mobile phones or laptop.
Chatbots
have natural language processing capability, and it is used in healthcare to
question patients and run essential diagnoses like real doctors. In education,
it is used for providing students with easy to learn conversational interfaces and
on-demand online tutors. Thus, the primary use of AI speech to text technology.
Computer
vision is a form of AI used to provide the street vision for the car to
overcome obstructions on the road. It helps automate tasks such as detecting
cancerous moles in the skin, finding symptoms in X-ray, and MRI scans.
In
the case of a bank, it can be used for detecting fraudulent transactions,
identifying credit card fraud, and preventing financial crimes.
In
the medical field, it can help doctors arrive at more accurate preliminary
diagnoses, reading medical imaging, and find appropriate clinical trials for
patients.
It
has the potential to access an enormous amount of information, imitate humans,
make recommendations, and correlate data. In the oil and gas industry, it can
be used in developing petroleum exploration techniques and distinguish rock
samples. AI works on the concept of cognitive computing, which enables people
to create a profoundly new kind of value, finding answers and insights locked
away in volumes of data. Cognitive computing mirrors some of the critical
cognitive elements of human expertise systems that reason out problems as a human
does. They use similar processes as humans do to reason about information. They
read and can do this at massive speed & scale.
Unlike
conventional computing solutions that can only handle neatly organized
structured data such as what is stored in a database, cognitive computing
solutions can understand unstructured data, which is 80% of data today. They
rely on natural language, which is governed by rules of grammar, context, and
culture. It is implicit, ambiguous, complex, and a challenge to process.
Certain idioms can be particularly challenging in English, and it is difficult
to parse these languages. Cognitive systems read and interpret like a person.
They do this by the structurally discerning meaning of the semantics of the
written material. It is very different from simple speech recognition. These
systems try to understand the real intent of users' language and use that language
or understanding to draw interferences through a broad array of linguistic
models and algorithms. They do this by learning from their interactions with
us.
Machine learning is a subset of AI that uses
computer algorithms to analyze data and make intelligent decisions based on
what it has learned without being explicitly programmed. Their algorithms are
trained with large sets of data, and they learn from examples. Deep learning is
a specialized subset of machine learning (ML) that uses layered neural networks
to simulate human decision making. Their algorithms can label and categorize
information and identify patterns.
Artificial
neural networks often referred simply as neural networks. A neural network in
AI is a collection of small computing units called neurons that take incoming
data and learn to make decisions over time. They are often layered deep and are
the reason deep learning algorithms become more efficient as the data sets
increase in volume.
AI
is different from data science, which involves statistical analysis, data
visualization, Machine Learning (ML), and more. It can use many AI techniques
to derive insight from data and draw interferences from data using ML
algorithms. They both can handle significantly large volumes of data. Machine
learning relies on defining behavioural rules by examining and comparing large
data sets to find common patterns.
Supervised learning is a type of machine
learning where the algorithm is trained on human-labelled data. Unsupervised
learning is another type of machine language that relies on giving the
algorithm unlabeled data, and it finds patterns by itself. Input is provided
without providing the labels and let the machine infer qualities.
Clustered
data or grouped data includes providing the algorithm with a constant stream of
network traffic and let it independently learn the baseline network activity as
well as an outlier and possibly malicious behaviour happening on the network.
This
type of ML technology is known as reinforcement learning, which relies on
providing an ML algorithm with a set of rules and constraints and letting it
learn how to achieve goals. The desired goal is defined by is with allowed
actions and constraints. It is the model used to find patterns in the data
without the programmer having to program these patterns explicitly.
Supervised
learning can be classified into three categories regression, classification,
and neural networks. Regression estimates continuous values. It is built by
looking at the relationship between the features X and result Y, where Y is a
continuous variable. Neural networks are based on the discretized model.
Assigning discretized class labels Y based on many many input features X. Classification
models classify results with more than two categories, and it includes decision
trees, support vector machines, logistic regression, and random forests. Each
column is a feature; each row is a data point. Classification is a process of
predicting the class of given data points.
With
ML, data sets are typically split into training, validation, and test sets. The
training subset is the data used to train the algorithm. The validation subset
validates results and fine-tunes the algorithm parameters. Testing data is used
to evaluate how good our model is through some defined parameters such as
accuracy, precision, and recall.
Deep
learning algorithms learn from the unstructured data sets such as photos,
videos, and audio files. These algorithms do not directly map the input to
output; instead, they rely on several layers of processing units. Each layer
passes its output to the next layer, which processes and passes it to the next.
The
process of developing algorithms includes configuring the number of layers and
the type of functions that connect the outputs of each layer to the inputs of
the next. Then the model is trained with lots of annotated examples. These
algorithms improve as they are fed more data, unlike the ML algorithms, which
plateau as the data sets grow. It is used in facial recognition, medical
imaging, language translation, and driverless cars.
Neural
networks are a collection of small units called neurons. These neurons take
incoming data like a biological neural network and learn to make decisions over
time. They learn through a process called backpropagation. Backpropagation uses
a set of training data that match known inputs to desired outputs.
First,
the inputs are plugged into the network, and outputs are determined, then an
error function determines how far the given output is from the desired output.
The collection of neurons is called layers. They take in input and provide an
output. Hidden layers other than input & output layers take in a set of
weighted inputs and produce an output through an activation function.
Perceptrons
are the most straightforward and oldest type of neural networks that utilize
single-layered networks consisting of input nodes connected directly to output
nodes. Hidden and output nodes have a property called bias, which is a particular
type of parameter that applies to a node after the inputs are considered. The activation
function is run against the sum of the inputs and bias, and then the result is
forwarded as an output.
Convolutional
neural networks (CNN) are multi-layered networks that take inspiration from the
animal visual cortex; they are useful in image processing and video
recognition. Convolution is a mathematical operation where a function is
applied to another function is a mixture of two functions. They are good at
detecting simple structures in an image and putting those simple features
together to construct more complex features. It occurs in a series of layers,
each of which conducts a convolution on the output of the previous layer.
Recurrent
neural networks (RNN) perform the same task for every element of a sequence
with prior outputs feeding the subsequent stage inputs in a general network.
Input
is processed through several layers, and an output is produced with an
assumption that two successive inputs are independent of each other. It can be
used in making use of information in long sequences.
Google
uses AI-powered speech to text in their call screen feature to handle scam
calls and show the user the text of the person speaking in real-time. YouTube
uses this to provide automatic closed captioning. With the help of neural
network synthesizing, the human voice is possible, which is known as speech
synthesis.
The
field of computer vision focuses on replicating parts of the complexity of the
human system and enabling computers to identify & process projects in
images & videos, the same way humans do. It enables the digital world to interact with the
physical world. It plays a crucial role in an augmented and mixed reality that
allows smartphones, tablets, and smart glasses, etc. to overlay and embed
visual objects on real-world imagery.
There
are some ethical issues and concerns related to AI, which helps us in knowing
the negative impacts of AI on the human world? It can be used for nefarious
reasons, i.e., for a dictatorial government to enforce their will on people, on
arresting and suppressing democracy and others. Ethics is not a technological
problem, it is a human problem. In self-driving cars ethical question emerging
is the trolley problem, for example, if a car has to decide which accident to
cause, it has to pick between running into a sign and to hurt passengers in the
vehicle or running into pedestrians on the side of the road, potentially saving
the passengers of the vehicle. It opens lots of questions on whom to blame for
the accident, whether the car owner or the car company.
AI-powered
risk assessment systems in courts help in predicting the probability of a
person reoffending and hence provide guidelines for sentencing or granting
parole based on the calculated risk of recidivism. There is concern that these
systems can be biased against people of colour.
The main area of research in AI is solving the bias problem in ML. There is a technique of directly modifying the data we feed through techniques like data
augmentation to enable fewer bias data. But still, this technique of solving
the bias has aroused plenty of questions in the mind of researchers since it is
not an effective method of eliminating bias. AI systems experts must guard against
introducing bias, whether gender, social, or any other form of bias.
For
developers, there are 4 aspects of AI that help people perceive it as
trustworthy; Transparency: People should be aware of the fact of having some
sort of expectations while interacting with AI; Accountability: Any unexpected
results can be undone if required; Privacy: Personal information should always
be protected; Lack of bias: Developers
should use representative training data to avoid regular audits to detect any
kind of bias expecting in.
AI
can be of immense importance in the medical field for early detection of any
kind of diseases such as cancers, sight loss, and other problems for quick
treatment before the situation is aggravated. It can also be helpful in the agriculture
sector by keeping crops away from diseases in case of any kind of harm posed to
crops.
AI
in the oil and gas industry is centred around two fields, i.e., machine
learning & data science. British petroleum developed a cloud-based
geosciences platform is known as “Sandy” to interpret geology, geophysics,
historical, and reservoir project information. The national data repository
(NDR) of the UK has many terabytes of different wellbores, seismic surveys, and
pipelines, which is interpreted by AI.
Spark
cognition AI systems will be used in spark predict platform to monitor topside
and subsea installations and analyze sensor data to identify any kind of
failure before it occurs. Shell has also adopted AI software named as Azure C3
IoT (internet of things) platform for its offshore operations. It is a similar
kind of platform as compared with the spark predict platform.
To
develop and use AI systems responsibly, AI developers must consider the ethical
issues inherent in AI. They must have a realistic view of their systems and
their capabilities and be aware of different forms of bias potentially present
in their systems. With this awareness, developers can avoid unintentionally
creating AI systems that have negative rather than positive impacts.
To this issue, many researchers and scientists have shown their concerns by anticipating the future of AI. Professor Stephen Hawking said about the future of AI that “The rise of the powerful AI will be either be the best or worst thing to happen to humanity, we do not yet know which.” Elon Musk said that ”AI is more dangerous than nuclear weapons.” Hence Future of AI should be decided by us, whether it should be helpful for humanity or a threat.