Applied Generative AI Professional
1) Which Al capabilities are associated with Horizon 1 (H1) systems in the Al-First approach at Infosys?
Ans. Augmenting intelligence into existing systems
2) What is the primary purpose of a language model in natural language processing?
Ans. To predict and find patterns in natural language
3) What type of datasets are often involved when working with AI-First Software Engineer approach?
Ans. Structured, Unstructured
4) Which among the following is not a capability of GPT-3 ?
Ans. Responding to multiple problem at a time
6) What is the primary benefit of applying Al to software engineering?
Ans. Faster bug fixes
7) State True or False programming knowledge is required to interact with GPT-3?
Ans. False
8) Identify the primary principle that 'Al First' companies should keep in mind while designing their
products.
Ans. Product should be User Centric
9) What is the different between complete mode collapse and partial mode collapse?
Ans. In complete mode collapse, all generated samples are identical, while in partial mode collapse, only
a few modes are learned better than others
10) Prompt. "What are the different phases of software Development Life Cycle?" Identify which
category this example
Ans. Zero shot learning
11) What is the purpose of using multiple attention heads in GPT-3?
Ans. To improve the model's ability to capture different types of dependencies
12) State True or False: Large Language Models can accurately predict future events and outcomes
based on the given prompt
Ans. False
13) What is the difference between supervised and unsupervised prompt engineering?
Ans. Unsupervised prompt engieering creates prompts based on the structure of the input
data......................pre-specified tasks
14) Which of the following additional tactics can be used to make the prompting more effective?
Ans. Ask for structured output
Specify the intended tone of the text
Set model's temperature parameter
15) State True or False: ChatGPT can be asked to print daily news headlines
Ans. True
16) What is the best way for prompt engineering to lessen bias in language models?
Ans. By optimizing the language model for identifying and addressing bias in data
17) State True or False: In an Al-First Company, the process of collecting data, analyzing data and
data-driven decision- making by taking quantitative feedback from processes are automated.
Ans. True
18) State True or False: In OpenAI GPT framework, prompts should be minimum 64 characters long.
Ans. False
19) What is the role of transfer learning in prompt engineering?
Ans. To fine-tune a pre-trained language model for specific tasks
20) From what value should the 'Temperature' parameter be modified, in order to increase possibility
of consistent (same) output from the GPT-3 model?
Ans. 0.0
21) To obtain a more deterministic output the value of Temperature and Top_P should be
Ans. Lower value of Temperature and Lower value of Top_P
22) Which of the fallowing is NOT a common consideration in prompt engineering?
Ans. Font size of the prompt
23) During which of the following SDLC phases Generative Al capabilities can be utilized.
Ans. All (Requirement Elicitation, Design, Coding)
24) How can Al-based anomaly detection techniques be used in software engineering to increase
system security?
Ans. By conducting real-time monitoring and alerting for suspicious activities
25) Which of the following best describes Intent in prompt engineering?
Ans. It refers to the overall objective or goal of the conversation
26) State TRUE or FALSE: Pretrained models can suffer from bias present in the data they were trained
on.
Ans. True
27) True/False: AI-First Software Engineering eliminates the need for human involvement in software
development.
Ans. False
28) In the Software Development Life Cycle (SDLC), which phase is primarily concerned with defining
the overall system architecture, component design, and user interface layout?
Ans. Design and Planning
29) What is the most important factor while creating language model prompts?
Ans. Clarity of the prompt
30) In OpenAI GPT framework, TopP refers to
Ans. the cumulative probability of tokens that can be chosen by the language model
31) Which among the following is not a SDLC phase
Ans. Feedback collection
32) Which of the following statements about Large Language Models (LLMS) and their capabilities in
software engineering is
Ans. LLMs are an ideal tool for adopting an Al-First Software Engineering approach as they can perform a
wide range of tasks, including code generation refactoring error correction, code styling, documentation
generation, and translation between programming languages
33) What expertise is required in AI-First Software Engineering?
Ans. All (Software engineering principles, Data pre..., Statistical analysis, Machine Learn...)
34) Which kind of language models are typically used for prompt engineering?
Ans. Pre-trained languages models
35) Which of the following approaches can be employed to address the limitation of word
embedding’s in GPT-3 concerning context information?
Ans. Incorporating attention mechanisms, to enable the model to attend to relevant context
36) State True or False: GPT-3 can process all Text, Audio and image inputs.
Ans. False
37) Identify all the limitations of Large Language Models
Ans. .These models are stochastic in nature
These models are trained on finite data and may not contain recent updates
These models may be misled by an incorrect input prompt
38) State True or False: In the AI-First approach, Product designs do not becomes more user-centric
Ans. False
39) What kind of output can DALL-E2 produce?
Ans. Image
40) Identify which Prompt category below example belongs to? Red: Apple, Blood Blue : Sky,
Blueberries Green: Leaves, Grass Yellow;
Ans. Few shot learning
41) Prompt engineering can help in addressing the issue of
Ans. Sample quality
42) Which of the following is the most important consideration in AI-First Software Engineering?
Ans. Ethical implementation
43) How can AI contribute to the SDLC user acceptance testing (UAT) phase?
Ans. By analyzing user behavior and preferences
44) Which AI model can produce an image output?
Ans. DALLE-2
45) State True or False: Prompt Engineering requires domain expertise in the specific task at hand.
Ans. True
46) How does the positional encoding in GPT-3 handle longer input sequences?
Ans. It applies a fixed encoding.......all input lengths
47) When using GPT-3 for a question such as "What is the color of sky?", can be used.
Ans. Zero shot learning
48) State True or False: The value of the temperature parameter control the uncertainty in the output
Ans. True
49) What is the range of Temperature parameter?
Ans. [0, 1]
50) How are the attention weights computed in GPT-3?
Ans. By taking the dot product of the Query and Key ......a softmax function
51) State True or False: The OpenAI GPT framework can be used onl for Python projects in the
development phase
Ans. False
52) What is the different between a "prefix" and "prompt" in prompt engineering?
Ans. Prefixes are fixed word groups that always................................
53) In OpenAI GPT framework for getting more deterministic response. Identify the value for
temperature parameter
Ans. Zero
54) What can be the main objective of generative models?
Ans. generate new data that resembles the original training data.
55) Which of the following is a considerable challenge in AI-First Software Engineering?
Ans. Interpretation of decisions taken by AI systems
56) Choose all the valid attributes of AI-First experience?
Ans. Contextual com.., Seamless natural..., Multi-device...
57) In an AI-First company, which of the following tasks can be automated? A Data collection, B
Analyzing data, C Decision making
Ans. A,B,C
58) State True or False: Intent in prompt engineering is only applicable to natural language processing
(NLP)
Ans. False
59) How can prompt engineering be applied to the creation of content?
Ans. None of the above
60) Which generative model is known for its ability to generate realistic images?
Ans. OpenAI's DALL-E 2
61) What is the best way for prompt engineering to lessen bias in language models?
Ans. By removing bias from the prompt
62) Which among the following are the key patterns for horizon-2 systems.
Ans. AI Governance, Entity..., Classification
63) What is the main advantage of AI-first Software Engineering?
Ans. (Faster...., Enhanced...., Increased...)
64) Which statement correctly characterizes the architectural design of the GPT-3 model?
Ans. GPT-3 employs a transformer................
65) The input that user gives to GPT-3 model is termed as
Ans. Prompt
66) Which OpenAI model can be used for Requirement Elaboration?
Ans. text-davinci-003
67) What is the range of Temperature Parameter?
Ans. [0, infinity]
68) Select all the options that are applicable to "transfer learning" in the context of pretrained model?
Ans. Transfer the weight that......................
The utilization of knowledge learned.............
It is process of using the knowledge model.............
69) Which phase of AI-First Software Engineering involves training and fine-tuning AI models?
Ans. Implementation and coding
70) What does "fine-tuning" a pretrained model involve?
Ans. Training the model on a small domain-specific dataset
71) OpenAI GPT-3 models are based on which of the following architectures?
Ans. Transformers
72) Different users input a prompt as follows to GPT-3: "Print ant integer greater than 100" identify
the correct option
Ans. Every user MAY get a different output
73) Which of the following is not an example of Sensory AI ?
Ans. Multi-device application
74) Which of the following is not a problem when using Maximum Likelyhood estimation to obtain the
parameters in a language model?
Ans. Lemmatization
75) What is the different between a "prefix" and a "prompt " in prompt engineering?
Ans. A prompt is a constant group of words..................................................
76) Why are language models generally considered more complex ......machine learning models?
Ans. Language models require more computational resources...........................
77) GPT 3/4 models are restricted to answering on only few domains?
Ans. False
78) What should you do to make GPT-3 model provide a completion inspecifically formatted text
form?
Ans. you must give suitable example
79) Limitations of OpenAI GPT-3 are
Ans. All of the above
80) What is the purpose of the positional encoding in GPT-3?
Ans. to insert a notion of location.................
81) What is a large language model?
Ans. A model that can process and generate.....................
82) Select option which does not relate to AI First paradigm
Ans. Programming on mainframe computers
83) Identify name of the process in which AI systems that can take inputs via text, voice, images,
gestures, etc.
Ans. Sensory AI
84) What is the main focus of AI-First Software Engineering approach?
Ans. Integration of AI capabilities...............................
85) Which of the following is an example of an AI-First approach in software engineering?
Ans. Using AI to generate....................
5) Which of the following are key components of a transformer model architecture?
Ans. (Self-attention mechanism, Positional encoding)
----All the best