The Deep Learning Revolution: Understanding Concepts & Critical View

This essay provides a critical analysis of Terrence Sejnowski's 'The Deep Learning Revolution,' addressing key questions about the nature of deep learning, its resurgence, and its future implications. The essay explores Sejnowski's perspective on neural networks, the challenges in AI such as the cocktail party problem, and the potential of deep learning in cognitive computing, including self-driving cars and advancements in sensory intelligence. It also discusses Orgel's rules and Sejnowski's claims about generalization in deep learning, contrasting them with Chomsky's views. The essay highlights the advantages of deep learning in solving complex problems and analyzing large datasets, ultimately offering a comprehensive overview of the book's central themes and arguments. Desklib offers a platform to explore more such solved assignments and past papers.

Running head: DEEP LEARNING REVOLUTION
DEEP LEARNING REVOLUTION
Name of student:
Name of university:
Author’s note:

1
DEEP LEARNING REVOLUTION
Answer of question 1
1. According to Sejnowski, Deep learning is the learning of representations of data
as compared to the algorithms that are specific to the task. The author cited the
success of deep learning in the Google translate and several other natural
language.
2. The two things that led to the rebirth of deep learning in the recent times is the
requirement of advanced and faster solving of the problems of the computation. In
the present world, the companies generate huge amount of data that needs to be
processed and analysed quickly and efficiently. It has been observed that humans
are unable to perform these tasks efficiently.
3. The author learnt about the neural networks based on the intuition that is the
nature had solved the computational problems, then it can be learnt from nature
the process of solving these problems (Sejnowski, 2018). The artificial neural
networks uses the processing of brain as the base of developing the algorithms
that can be utilised for patterns of complex model and the prediction of problems.
4. The work on the visual systems is a major deal in the AI because it integrates the
information from various streams that includes the signals from reward system.
Answer of question 2
1. According to the authors, the cocktail party problem, there can be major
challenge in hearing the person that is in front of anyone when the room is filled
with the voices of people talking all over the place (Sejnowski, 2018). Imagining
any cocktail party that has over 100 people in any room and there are 100
nondirectional microphones that are spread around and each of the microphones
are receiving the sounds from everyone but with the various amplitude ratio for

2
DEEP LEARNING REVOLUTION
every person on each of the microphone. The author also provide a problem that is
it possible in devising any algorithm, which can differentiate each of the voices in
distinct separate output channels?
2. According to the authors, any Boltzmann machine is a machine that would break
down any logjam that was holding the learning in multilayer neural network
models for any generation.
3. According to the author, the Convolutional learning is commonly applied in the
analysing of the visual imagery and the reward learning is based on the
observation that is used by the agents in solving complex problems in uncertain
conditions by the exploration of the various options in environment and the
learning from the outcomes of the actions.
Answer of question 3
1. According to the author, the future of deep learning and the age of the cognitive
computing is dawning and soon the world would have the self-driving cars that
would drive better than the humans (Sejnowski, 2018). The current advances in
the artificial intelligence is majorly focussed on the cognitive and the sensory
sides of the intelligence, with the advances on the action and the motor
intelligence yet to arrive. It has been predicted that a core standard tools set would
adopt in the industry of deep learning. It is also predicted that the tools of deep
learning would incorporate the frameworks of simplified programming for
executing quick coding.
2. Orgel’s first rule is that whenever any spontaneous process is too slow or too
inefficient, a protein would evolve to make it more efficient or speed it up.
Orgel’s second rule states the evolution is cleverer than you are.

3
DEEP LEARNING REVOLUTION
Sejnowski claim that the deep learning are capable of generalisation and this
theory was dismissed by Chomsky as mysticism.
3. Deep intelligence is the artificial intelligence combined with the function of deep
learning which can help in solving the complex problems easily and provide the
solutions at a faster rate than the human brain (Sejnowski, 2018). This is the
branch of artificial intelligence where the machines can perform the tasks of
humans. The deep learning is the branch of machine learning where any neural
network of artificial intelligence and the algorithms are inspired by the human
brain and learn from huge amount of data. In the present work the virtual
assistants like Siri, Alexa and Cortana are extensively used in the recent times and
the machines understand the voices of the people and perform the tasks.
4. Deep learning is an advantageous in the recent times and it could help in the
discovery of the solutions of complicated problems. This era of digitalisation has
permitted the enrichment of technology, and the computers are able to analyse the
massive amounts for the complex data that is known as big data and it can offer
increased accurate results. This assists the companies in increasing the chance of
identifying opportunities that are profitable or it can avoid the unknown risks.

4
DEEP LEARNING REVOLUTION
References
Sejnowski, T. J. (2018). The deep learning revolution. MIT Press.