Seminar Report on

What Chat-GPT can / cannot do

SUBMITTED BY
Shantanu Dhage, TEAIDB03

Under the guidance of

Prof. Nita Mahale

In partial fulfilment of the requirements for the

Bachelor’s Degree in Artificial Intelligence and Data Science of
SAVITRIBAI PHULE PUNE UNIVERSITY
[2020-2021]

Department of Artificial Intelligence and Data Science

D. Y. PATIL COLLEGE OF ENGINEERING

Akurdi, PUNE-411044.
Contents

i)Title Page

ii)Certificate

iii)Acknowledgement

iv)Abstract

1 Introduction 1
1.1 Problem Definition . . . . . . . . . . . . . . . . . . . . . 2
1.2 Motivation, Objective and Social Relevance…3
1.3 Planned Outcome . . . . . . . . . . . . . . . . . . . . . .4

2 Literature Survey of Topic………………………...5

3 Discussion of Base paper………………………...6

4 Algorithm……………………………………………7

5 Summary …………………………………………..8

6)References…………………………………………9
 D. Y. Patil College of Engineering
Akurdi, Pune-411044
Department of Artificial Intelligence and Data Science

CERTIFICATE

This is to certify that Shantanu Sunil Dhage has satisfactorily completed the seminar work
entitled “What Chat-GPT can / cannot do?” Which is a bonafide work carried out by him
under the supervision of Prof. Nita Mahale and it is approved for the partial fulfilment of
requirement of Savitribai Phule Pune University, for the award of the degree of Bachelors of
Engineering (Artificial Intelligence and Data Science) for the academic year 2020-21.

Dr Vinayak G. Kottawar
Nita Mahale
(Seminar Guide) (AI&DS HOD)
Place: Akurdi
Date:
 ACKNOWLEDGEMENT

With immense pleasure, I present the seminar report as part of the curriculum of the T.E.
Artificial Intelligence and Data Science. I wish to thank all the people who gave us an
unending support right from when the idea was conceived. I express sincere and profound
thanks to Prof. Nita Mahale, seminar Guide, and HOD Prof. Dr Vinayak G. Kottawar, who is
ready to help with the most diverse problems that I have encountered along the way. I express
sincere thanks to all staff and colleagues who have helped directly or indirectly in completing
this seminar successfully.

Name Seat No Signature

 ABSTRACT

Chat GPT is an advanced language model developed by OpenAI that possesses

impressive capabilities in generating human-like responses to text-based
prompts. This abstract provides an overview of what Chat GPT can and cannot
do. Chat GPT excels in language generation, offering coherent and contextually
relevant responses. It possesses a broad range of general knowledge, making it
capable of answering questions on various topics. Additionally, it can assist
with language translation, text summarization, and writing tasks by providing
suggestions and alternative word choices. However, Chat GPT has limitations.
It lacks context and memory of previous interactions, which can result in
inconsistent conversations. Factual accuracy is not guaranteed, and users should
verify critical information from reliable sources. Ethical considerations are
important, as the model can generate biased or inappropriate content. It lacks
true understanding of the world and may produce nonsensical or irrelevant
responses. Lastly, Chat GPT is limited to generating text-based responses and
cannot perform physical actions or interact with the physical world.
Understanding the capabilities and limitations of Chat GPT is crucial for users
to effectively and responsibly utilize the model
 1) Introduction

Chat GPT is an advanced language model developed by OpenAI that can generate human-
like responses to text-based prompts. While it possesses impressive capabilities, it also has
certain limitations. This report aims to outline what Chat GPT can and cannot do, providing a
comprehensive understanding of its strengths and weaknesses.
 1.1) Problem definition

What Chat GPT can / cannot do?

The problem at hand is to define the capabilities and limitations of Chat GPT, an advanced
language model developed by OpenAI. The objective is to clearly outline what Chat GPT can
and cannot do, providing a comprehensive understanding of its strengths and weaknesses.
The problem encompasses identifying the specific areas in which Chat GPT excels, such as
language generation, general knowledge, language translation, text summarization, and
writing assistance. It also involves recognizing the limitations of the model, including its lack
of context, potential for factual inaccuracies, ethical considerations, absence of real-world
understanding, and inability to perform physical actions. By defining the problem, we aim to
establish a clear understanding of the boundaries and capabilities of Chat GPT, enabling users
to make informed decisions about its usage and effectively leverage its strengths while being
mindful of its limitations.
 \

1.2) Motivation, Objective and SOCILA RELEVANCE

Motivation:
The motivation behind exploring the capabilities and limitations of Chat GPT is to
gain a comprehensive understanding of its functionalities. As an advanced language
model, Chat GPT has garnered significant attention and usage across various
domains. By understanding its strengths and weaknesses, users can make informed
decisions about its application, ensuring responsible and effective utilization.

Objective:
The objective of examining what Chat GPT can and cannot do is to provide a clear
and concise overview of its capabilities and limitations. By defining these aspects,
users can better understand the model's potential applications and its boundaries. This
knowledge enables users to leverage Chat GPT effectively, while also being aware of
its limitations to avoid potential pitfalls or misuse.

Social Relevance:
Understanding the capabilities and limitations of Chat GPT holds significant social
relevance. With the increasing integration of AI language models in various domains,
it is crucial to comprehend their capabilities to ensure responsible use. By being aware
of the model's limitations, users can critically evaluate its responses, fact-check
information, and avoid spreading misinformation. Additionally, recognizing ethical
considerations associated with Chat GPT's responses promotes responsible and
inclusive interactions, fostering a more positive and respectful online environment.
By exploring the social relevance of understanding Chat GPT's capabilities and
limitations, we can promote informed decision-making, responsible usage, and
contribute to a more knowledgeable and conscientious AI-driven society.
 1.3) Planned Outcome

The planned outcome of examining what Chat GPT can and cannot do is to provide users
with a clear understanding of the model's capabilities and limitations. By presenting a
comprehensive overview, users can make informed decisions about how to effectively
utilize Chat GPT while being mindful of its boundaries. The outcome aims to equip users
with the knowledge to leverage Chat GPT's strengths in language generation, general
knowledge, language translation, text summarization, and writing assistance. Users will
also gain an understanding of the limitations, including the lack of context, potential for
factual inaccuracies, ethical considerations, absence of real-world understanding, and
inability to perform physical actions. By providing this information, the planned outcome
is to empower users to engage with Chat GPT responsibly and effectively. Users will be
able to critically evaluate the model's responses, fact-check information, and exercise
caution when relying on its output. This outcome contributes to a more informed and
conscientious use of AI language models, fostering a positive and responsible AI-driven
society.
 2) Literature Survey of Topic

The literature survey on the topic "What Chat GPT can/cannot do?" reveals a
significant body of research and discussions surrounding the capabilities and
limitations of language models like Chat GPT. Several studies have explored the
potential applications and challenges associated with these models. Here are some key
findings from the literature:

1. Language Generation: Researchers have highlighted the impressive language

generation capabilities of models like Chat GPT. These models can generate coherent
and contextually relevant responses, making them useful for tasks such as dialogue
systems, chatbots, and content generation.
2. General Knowledge: Chat GPT demonstrates a broad range of general knowledge,
allowing it to answer questions on various topics. However, studies caution that the
model's responses should be verified from reliable sources, as it may occasionally
produce inaccurate or biased information.
3. Context and Memory: One limitation identified in the literature is the lack of context
and memory in language models like Chat GPT. These models do not retain
information from previous interactions, leading to inconsistent conversations and an
inability to maintain coherent dialogue over extended sessions.
4. Factual Accuracy: While Chat GPT can provide informative responses, it is important
to note that factual accuracy is not guaranteed. Studies emphasize the need for users
to fact-check critical information from trusted sources, as the model may occasionally
generate incorrect or outdated information.
5. Ethical Considerations: Researchers have highlighted the ethical considerations
associated with language models. Chat GPT can generate biased or inappropriate
content, reflecting the biases present in the training data. Efforts are being made to
address these biases and promote fairness and inclusivity in AI language models.
6. Lack of True Understanding: Despite their impressive language generation abilities,
models like Chat GPT lack true understanding of the world. They may produce
nonsensical or irrelevant responses, highlighting the need for cautious interpretation
of their output.
7. Physical Actions and Real-World Interaction: Language models like Chat GPT are
limited to generating text-based responses and cannot perform physical actions or
interact with the physical world. This limitation restricts their application in tasks that
require physical manipulation or real-time interaction.

The literature survey provides valuable insights into the capabilities and limitations of
Chat GPT and similar language models. It highlights the need for responsible usage,
critical evaluation of responses, and ongoing research to address challenges and
improve the performance of these models.
 3) Discussion of Base paper

The base paper on the topic "What Chat GPT can/cannot do?" provides a comprehensive
analysis of the capabilities and limitations of Chat GPT, shedding light on its potential
applications and challenges. The discussion in the base paper highlights the following key
points:
1. Language Generation: The base paper acknowledges the impressive language
generation capabilities of Chat GPT. It emphasizes the model's ability to generate
coherent and contextually relevant responses, making it suitable for tasks such as
chatbots, dialogue systems, and content generation.
2. General Knowledge: The base paper recognizes Chat GPT's broad range of general
knowledge, enabling it to answer questions on various topics. However, it cautions
that the model's responses should be verified from reliable sources, as it may
occasionally produce inaccurate or biased information.
3. Context and Memory: The base paper discusses the limitation of Chat GPT in terms
of context and memory. It highlights that the model does not retain information from
previous interactions, leading to challenges in maintaining coherent and consistent
dialogue over extended conversations.
4. Factual Accuracy: The base paper acknowledges the need for caution regarding
factual accuracy. While Chat GPT can provide informative responses, it may
occasionally generate incorrect or outdated information. The paper emphasizes the
importance of fact-checking critical information from trusted sources.
5. Ethical Considerations: The base paper addresses the ethical considerations associated
with language models like Chat GPT. It acknowledges that these models can generate
biased or inappropriate content, reflecting the biases present in the training data. The
paper emphasizes the ongoing efforts to mitigate biases and promote fairness and
inclusivity in AI language models.
6. Lack of True Understanding: The base paper discusses the limitation of Chat GPT in
terms of true understanding. It highlights that the model lacks genuine comprehension
of the world and may produce nonsensical or irrelevant responses. This limitation
underscores the need for cautious interpretation of the model's output.
7. Physical Actions and Real-World Interaction: The base paper acknowledges that Chat
GPT and similar language models are limited to generating text-based responses and
cannot perform physical actions or interact with the physical world. This limitation
restricts their application in tasks that require physical manipulation or real-time
interaction.
 4) Algorithm

1. Initialize the Chat GPT model and load the pre-trained weights.
2. Receive user input or query.
3. Preprocess the user input by tokenizing and encoding it to match the model's input
format.
4. Pass the pre-processed input through the Chat GPT model for inference.
5. Generate a response based on the input using the language generation capabilities of the
model.
6. Post-process the generated response by decoding and detokenizing it to make it
readable.
7. Display the response to the user.
8. If the user has follow-up questions or requests, repeat steps 2-7.
9. If the user input requires context or memory, store relevant information for future
reference.
10. Handle limitations and caveats:
a. Lack of context: Inform the user that the model does not retain
b. information from previous interactions, leading to potential inconsistencies in dialogue.
b. Factual accuracy: Advise the user to fact-check critical information from
reliable sources, as the model may occasionally produce inaccurate or
outdated information.
c. Ethical considerations: Highlight the potential for biased or
inappropriate content and encourage responsible usage and critical
evaluation of the model's responses.
d. Lack of true understanding: Remind the user that the model lacks
genuine comprehension of the world and may produce nonsensical or
irrelevant responses.
e. Physical actions and real-world interaction: Inform the user that the
model is limited to generating text-based responses and cannot perform
physical actions or interact with the physical world.
11. Continue the conversation with the user until the desired outcome or conclusion is
reached.
12. Terminate the algorithm when the conversation is complete or the user ends the
interaction.
 5)Summary

Chat GPT, a language model developed by OpenAI, possesses impressive language

generation capabilities and a broad range of general knowledge. It can generate coherent and
contextually relevant responses, making it suitable for tasks such as chatbots, dialogue
systems, and content generation. However, there are certain limitations to consider. Chat GPT
lacks context and memory, leading to challenges in maintaining coherent dialogue over
extended conversations. While it can provide informative responses, factual accuracy is not
guaranteed, necessitating fact-checking from reliable sources. Ethical considerations arise
due to the potential for biased or inappropriate content, reflecting the biases present in the
training data. It is important to note that Chat GPT lacks true understanding of the world and
may produce nonsensical or irrelevant responses. Additionally, it is limited to generating text-
based responses and cannot perform physical actions or interact with the physical world.
Responsible usage, critical evaluation of responses, and ongoing research are crucial to
address these limitations and enhance the performance of language models like Chat GPT.
Factoring in these considerations, Chat GPT can be a valuable tool for various applications,
but users should be mindful of its limitations and exercise caution when interpreting its
output.
 6)References

 Google

 YouTube

 Wikipedia

 Quora

 Geeks for geeks

 Chat GPT