# Deep Dive into Code with ChatGPT

By _Christian Prokopp_ on _2023-04-04_

Open Source libraries offer user documentation. But expert users and contributors have a deeper understanding of the inner workings stemming from a mental model and architecture derived from deep dives into the code. That understanding and model are helpful to employ the library more effectively, debug issues when using it, and teach interesting concepts on how to structure complex reusable code.

![Free diver](/content/blog/images/deep-dive-into-code-with-chatgpt.png "Free diver")

## Learn more with less effort

Software engineers can always read the code of a library, in theory. In practice, it is too time-consuming to dive deep into the code. Moreover, getting your head around the broader software architecture takes time and reading before things fall into place. Documentation is aimed at usage and can help only so much.

Here a tool like ChatGPT is powerful. Given the right question, it can outline complex libraries for you. That can be useful when you debug an issue or try to contribute to it. It can also be a great teaching tool. Learning how some libraries are implemented draws from the experience of many talented engineers and actual use and limitations.

An attractive aspect of it is its interactive nature. You can dig deeper into certain parts that interest you more. While you can do the same in documentation and code by drilling deeper into it, you can cherry-pick your questions, drill down, and ask 'stupid' questions with ChatGPT. It happily explains references to vaguely remembered concepts and goes deeper, even if they do not relate to the library.

To illustrate my point, I created a small example of what I mean below with jQuery. I asked ChatGPT-4 to explain it in software engineering terms. I followed up with a drill-down question on the plugin system and generalisable learnings from it. Of course, you can modify it to be more or less technical and have a lengthy conversation on ever more detailed aspects.

## A jQuery Deep Dive

Below the question I posed to ChatGPT-4.

* * *

    Act as the Principal Software Engineer at a startup. Explain the  jquery  library at  https://github.com/jquery/jquery  to your team. You want to teach them the inner workings of the library and code. Do not repeat anything they can find in the user documentation.
    
    Use markdown for the output.

* * *

## A jQuery for Software Engineers introduction by ChatGPT-4

Below the response from ChatGPT-4 using its formatting.

* * *

    # jQuery Library Deep Dive 

    Hello team! Today, we will explore the inner workings of the jQuery library, beyond what you can find in the user documentation. By understanding the core functionality and structure of the library, we can leverage its capabilities more effectively. 

    ## Overview 

    jQuery is a popular, fast, small, and feature-rich JavaScript library. It simplifies HTML document traversal, manipulation, event handling, and animation. jQuery is built on top of the  ** Sizzle **  selector engine, which powers its DOM element selection capabilities. 

    ## Core Structure 

    The jQuery library is structured as a self-contained, Immediately Invoked Function Expression (IIFE), which means it is executed immediately after being defined. 

        javascriptCopy code(function (global, factory) {  
        // jQuery initialization code   
        })(this, function (window) {   
        // jQuery core implementation   
        });

    This structure prevents jQuery's variables and functions from polluting the global namespace and ensures its compatibility with other libraries and code. 

    ### jQuery Object 

    The core jQuery object is a function that serves as the main entry point for using the library. It is also an alias for the  ** $ **  function. The jQuery object constructor creates a collection of DOM elements based on the provided selector: 

        javascriptCopy codefunction jQuery(selector, context) {   
        return new jQuery.fn.init(selector, context);   
        } 

    ### jQuery Prototype 

    jQuery's prototype ( ** .fn ** ) is an object containing all the methods and properties that are available to instances of the jQuery object. By extending the prototype, we can add custom functionality to the jQuery library. 

        javascriptCopy codejQuery.fn = jQuery.prototype = {   
        constructor: jQuery,  
        // ...other methods and properties   
        }; 

    ## Key Components 

    ### Sizzle Selector Engine 

    The Sizzle selector engine is the heart of jQuery's DOM element selection capabilities. It is a standalone library that can be used independently of jQuery. Sizzle is used within jQuery to efficiently query and filter DOM elements. 

        javascriptCopy codejQuery.find = Sizzle;   
        jQuery.expr = Sizzle.selectors; 

    ### Event Handling 

    jQuery provides a robust event handling system that abstracts browser-specific implementations and makes it easier to work with events. 

    The event handling subsystem is built around the  ** jQuery.event **  object. Key components include: 

    -   ** add **

         : Binds event handlers to elements. 

    -   ** remove **

         : Unbinds event handlers from elements. 

    -   ** trigger **

         : Triggers a specific event on an element. 

    -   ** special **

         : Contains definitions for custom event types. 

    ### AJAX 

    jQuery's AJAX capabilities are encapsulated within the  ** jQuery.ajax **  object. It provides a simple, consistent API for making asynchronous HTTP requests. 

    Some key functions include: 

    -   ** jQuery.ajax **

         : The core AJAX function. 

    -   ** jQuery.get **

         : Shortcut for making GET requests. 

    -   ** jQuery.post **

         : Shortcut for making POST requests. 

    -   ** jQuery.getJSON **

         : Shortcut for making JSON requests. 

    ### Animation 

    jQuery's animation subsystem allows developers to create complex animations with minimal code. It is built around the  ** jQuery.Animation **  object and a set of predefined effects that can be applied to elements. 

    ### Plugin System 

    jQuery provides a simple, powerful plugin system that allows developers to extend its functionality. Plugins are created by extending the  ** jQuery.fn **  object with custom methods. 

        javascriptCopy codejQuery.fn.myPlugin = function () {   
        // Plugin implementation   
        }; 

    ## Conclusion 

    By understanding the inner workings of the jQuery library, we can utilize its features more effectively and even extend its functionality to meet our specific needs. Let's leverage this knowledge to build robust, efficient, and maintainable applications. 

    If you have any questions or need clarification on any aspect of the library, feel free to ask. Happy coding! 

* * *

## Diving Deeper, the Plugin System

Question to ChatGPT-4:

* * *

    Explain the Plugin System further and what can be learned from it if we were to implement a plugin system? 

* * *

Answer from ChatGPT-4:

* * *

    # jQuery Plugin System: Deep Dive 

    The jQuery plugin system is an essential part of the library, as it allows developers to extend its functionality by adding custom methods. Let's dive deeper into the plugin system, its structure, and how we can learn from it when implementing our own plugin system. 

    ## Structure 

    The plugin system is built around the  ** jQuery.fn **  object (jQuery's prototype), which contains all the methods and properties available to jQuery instances. By extending this object, we can create reusable plugins that can be called on any jQuery object. 

    ### Creating a Plugin 

    To create a plugin, you define a new method on the  ** jQuery.fn **  object. The method should perform the desired functionality and return the modified jQuery object to maintain chainability. Here's a simple example of a plugin that adds a CSS class to all elements in a jQuery object: 

        javascriptCopy codejQuery.fn.addClassCustom = function (className) {   
        // Iterate over each element in the jQuery object   
        this.each(function () {   
        // Add the CSS class to the current element   
        jQuery(this).addClass(className);   
        }); 

        // Return the modified jQuery object to maintain chainability   
        return this;   
        }; 

    Usage: 

        javascriptCopy code$('div').addClassCustom('my-custom-class'); 

    ## What Can Be Learned for Implementing a Plugin System 

    When designing your own plugin system, you can learn several key concepts and best practices from the jQuery plugin system: 

    ### 1. Encapsulation 

    jQuery plugins are encapsulated within the library's prototype, which helps maintain a clean and organized codebase. When implementing your own plugin system, consider using a similar encapsulation strategy to keep your code organized and easy to maintain. 

    ### 2. Chainability 

    One of jQuery's core features is chainability, which allows developers to perform multiple operations on a single object in a single statement. When designing your plugin system, consider returning the modified object to maintain chainability and improve the developer experience. 

    ### 3. Extensibility 

    The jQuery plugin system makes it easy for developers to extend the library's functionality with custom methods. When creating your plugin system, ensure it is easy for developers to add new functionality without modifying the core codebase. 

    ### 4. Compatibility 

    The jQuery plugin system is designed to be compatible with other plugins and code, thanks to its encapsulation and self-contained structure. When implementing your plugin system, consider the potential conflicts with other plugins or code and design your system to minimize such issues. 

    ### 5. Documentation and Naming Conventions 

    To help developers understand and use your plugin system effectively, provide clear documentation and follow consistent naming conventions. This will make it easier for others to create plugins and contribute to your project. 

    ## Conclusion 

    By understanding the jQuery plugin system and its structure, we can learn valuable lessons for implementing our own plugin system. By following best practices and lessons learned from jQuery, we can create a flexible, extensible, and easy-to-use plugin system for our projects. 

* * *

## Limitations

We could go on further and further. Beware, though, that ChatGPT can occasionally trip when you get too specific, e.g. ask it to look up a specific line of code. It is not a search engine or code repository. At that point, you are better off reading the actual code.

---

[Christian Prokopp, PhD](https://www.linkedin.com/in/prokopp/), is an experienced data and AI advisor and founder who has worked with Cloud Computing, Data and AI for decades, from hands-on engineering in startups to senior executive positions in global corporations. You can contact him at christian@bolddata.biz for inquiries.

Open Source libraries offer user documentation. But expert users and contributors have a deeper understanding of the inner workings stemming from a...

Free diver

Open Source libraries offer user documentation. But expert users and contributors have a deeper understanding of the inner workings stemming from a mental model and architecture derived from deep dives into the code. That understanding and model are helpful to employ the library more effectively, debug issues when using it, and teach interesting concepts on how to structure complex reusable code.

Deep Dive into Code with ChatGPT

# Securely Deploying Delta Lake on AWS: Best Practices for Cloud Engineers

By _Christian Prokopp_ on _2023-02-15_

Five tips for Cloud Engineers to deploy Databricks' Delta Lake on AWS safely.

![Stay safe](/content/blog/images/securely-deploying-databricks-on-aws-best-practices-for-cloud-engineers.webp "Stay safe")


Cloud engineering is a critical component of any modern business infrastructure. With more and more companies relying on cloud-based systems to store and analyse Data, it's essential to have knowledgeable engineers who can manage and deploy these systems effectively. One tool that has become increasingly popular in recent years is Delta Lake on AWS, a powerful platform that provides advanced Analytics and data processing capabilities. However, deploying Delta Lake on AWS can be a complex process, and it's essential to have a thorough understanding of the best practices and guidance to ensure the deployment is successful. This article will provide a comprehensive list of tips and recommendations to help cloud engineers deploy Delta Lake on AWS effectively.

## Customer-Managed VPC

The customer-managed VPC is the foundation of any standard Delta Lake deployment. It's crucial to ensure that the VPC is appropriately sized for your specific needs and has DNS hostnames and DNS resolution enabled. Additionally, each workspace must have a minimum of two dedicated private subnets in separate availability zones, and aim that no other AWS resources are placed in that subnet space. It's also essential to use the automatic availability zone (auto-AZ) option for all clusters to optimise the availability of IPs within a single workspace.

## Security Groups

Security group rules for the Delta Lake EC2 instances must allow all inbound (ingress) traffic from other resources in the same security group. It's also crucial to ensure that the outbound (egress) traffic rules allow all TCP and UDP access to the workspace security group and that ports such as 443, 3306, and 6666 are open. Ensure that your workspace security group allows outbound ports for services like Apache Kafka and Redshift to avoid connection issues.

## Data Plane to Control Plane Connectivity

When the cross-account IAM role spins up an EC2 instance in your AWS account, the cluster will try to call home to the Delta Lake AWS account. This process of calling home is known as the secure cluster connectivity relay or SCC. Ideally, use AWS PrivateLink endpoints to route the SCC relay and REST API traffic over the AWS backbone infrastructure, adding another layer of security. Additionally, you can entirely lockdown the cluster if you route to the control plane through PrivateLink and have no internet gateway available for the EC2 instances to make their way to.

Cross-Account IAM Role: A required portion of the deployment, the cross-account IAM role is used when initially spinning up EC2 instances within your environment. It's recommended to restrict your cross-account role to only the necessary resources and to restrict where the EC2 AMIs are sourced from. Be sure to use the proper IAM roles for both Unity Catalog and Serverless.

## Workspace Root Bucket

The workspace root bucket is used to store workspace objects like cluster logs, notebook revisions, job results, and libraries. It's essential to create a specific bucket policy allowing the Delta Lake control plane to write to it and not to use it for customer data or multiple workspaces. Do plan to implement encryption on your workspace storage, EBS volumes, and managed objects.

In the end, deploying Delta Lake on AWS requires a thorough understanding of the best practices and guidance. The customer-managed VPC, security groups, data plane to control plane connectivity, cross-account IAM role, and workspace root bucket are all critical components of a standard Delta Lake deployment. Cloud engineers must ensure that each component is appropriately configured and optimised to avoid connection or security issues. By following these tips and recommendations, cloud engineers can deploy Delta Lake on AWS effectively and ensure that their organisation has the analytics and data processing capabilities needed to succeed in the modern business world.

Read more about it in a related [Databricks blog post](https://www.databricks.com/blog/2023/01/27/best-practices-and-guidance-cloud-engineers-deploy-databricks-aws-part-2.html).

---

[Christian Prokopp, PhD](https://www.linkedin.com/in/prokopp/), is an experienced data and AI advisor and founder who has worked with Cloud Computing, Data and AI for decades, from hands-on engineering in startups to senior executive positions in global corporations. You can contact him at christian@bolddata.biz for inquiries.

Five tips for Cloud Engineers to deploy Databricks' Delta Lake on AWS safely.

Stay safe

Securely Deploying Delta Lake on AWS: Best Practices for Cloud Engineers

# The Power of Schema Enforcement in Delta Lake

By _Christian Prokopp_ on _2023-02-15_

Prevent errors and inconsistencies with Delta Lake's robust data management technology.

![Police lights](/content/blog/images/the-power-of-schema-enforcement-in-delta-lake.webp "Police lights")

Delta Lake is a powerful data lake technology that provides robust data management and reliability features. One of these features is the ability to enforce a schema on the data stored in a Delta Lake table. In this blog post, we’ll explore what schema enforcement is, why it’s important, and how it works in Delta Lake.

## What is Schema Enforcement?

Schema enforcement is the process of ensuring that the data stored in a Delta Lake table adheres to a predefined schema. A schema defines the structure of the data, including the names, data types, and constraints for each column. By enforcing a schema, Delta Lake ensures that the data stored in the table is consistent and conforms to the predefined structure.

## Why is Schema Enforcement Important?

Schema enforcement is important for several reasons. First, it helps to ensure data quality and consistency by enforcing a predefined structure for the data stored in a Delta Lake table. This helps to prevent errors and inconsistencies that can occur when data is stored in a loose or unstructured format.

Second, schema enforcement makes it easier to process and analyse the data stored in a Delta Lake table. With a predefined schema, data processing and analysis tools can easily understand the structure of the data and process it efficiently. This can help to reduce the time and effort required to process and analyse data.

Finally, schema enforcement can help to ensure data privacy and security by preventing unauthorised data from being stored in a Delta Lake table. By enforcing a schema, Delta Lake can ensure that only data that conforms to the predefined structure is stored in the table.

## How Does Schema Enforcement Work in Delta Lake?

Delta Lake uses a combination of schema inference and schema validation to enforce a schema on the data stored in a Delta Lake table. When data is first written to a Delta Lake table, Delta Lake automatically infers the schema based on the structure of the data. This inferred schema is then used to validate all subsequent data writes to the table.

If a write operation is attempted that would result in data that does not conform to the schema, Delta Lake will reject the write and raise an error. This ensures that only data that conforms to the schema is stored in the table, ensuring data quality and consistency.

In addition, Delta Lake provides the ability to modify the schema for a Delta Lake table. If the schema needs to be changed, the new schema can be specified, and Delta Lake will automatically validate all subsequent data writes against the new schema. This makes it easy to modify the schema as the structure of the data changes over time.

## Conclusion

Schema enforcement is a key feature of Delta Lake that helps to ensure data quality, consistency, and reliability. By enforcing a predefined schema on the data stored in a Delta Lake table, Delta Lake helps to prevent errors and inconsistencies, makes it easier to process and analyse data, and helps to ensure data privacy and security. With schema inference and schema validation, Delta Lake makes it easy to enforce a schema on your data, helping you to build robust and reliable data management systems.

---

[Christian Prokopp, PhD](https://www.linkedin.com/in/prokopp/), is an experienced data and AI advisor and founder who has worked with Cloud Computing, Data and AI for decades, from hands-on engineering in startups to senior executive positions in global corporations. You can contact him at christian@bolddata.biz for inquiries.

Prevent errors and inconsistencies with Delta Lake's robust data management technology.

Police lights

The Power of Schema Enforcement in Delta Lake

# A Guide to the Delta Lake Transaction Log

By _Christian Prokopp_ on _2023-02-14_

Discover the power of the Delta Lake transaction log - ensuring Data reliability and consistency.

![Wood logs](/content/blog/images/delta-lake-transaction-log-guide.webp "Wood logs")

Delta Lake is a powerful data lake technology that provides an ACID (Atomicity, Consistency, Isolation, Durability) transactional storage layer on top of a data lake, enabling reliable data updates and deletes. One of the key components of this reliability is the Delta Lake transaction log, which provides a complete history of all changes made to a Delta Lake table. In this blog post, we’ll explore what the Delta Lake transaction log is, why it’s important, and how it can be used.

## What is the Delta Lake Transaction Log?

The Delta Lake transaction log is a data file that is created for each Delta Lake table. This file is stored in the same location as the table and contains a complete history of all changes made to the table, including inserts, updates, and deletes. This log is maintained by Delta Lake and is automatically updated whenever a change is made to the table.

The transaction log is stored in Apache Parquet format, which is a columnar storage format that is optimised for big data processing. This format provides fast query performance and efficient storage, making it ideal for use in the Delta Lake transaction log.

## Why is the Delta Lake Transaction Log Important?

The Delta Lake transaction log provides a complete history of all changes made to a Delta Lake table, which is essential for ensuring data reliability and consistency. This log is used to recover the table in the event of a failure, ensuring that data is not lost or corrupted.

In addition, the transaction log provides a complete audit trail of all changes made to the table. This can be useful for auditing and regulatory compliance, as it provides a clear and complete record of all changes made to the table.

The transaction log also enables Delta Lake to provide ACID transactions, which are a set of properties that ensure that database transactions are processed reliably. This is important for ensuring that data updates and deletes are processed correctly, even in the event of a failure.

## How to Use the Delta Lake Transaction Log

The Delta Lake transaction log is automatically maintained by Delta Lake and does not require any manual intervention. However, there are a number of ways that you can use the transaction log to ensure data reliability and to provide a complete audit trail of all changes made to a Delta Lake table.

One way to use the transaction log is to enable Delta Lake’s time travel feature, which allows you to view a snapshot of the table as it existed at a specific point in time. This is useful for auditing and regulatory compliance, as it provides a clear and complete record of the table at a specific point in time.

Another way to use the transaction log is to recover a table in the event of a failure. Delta Lake uses the transaction log to automatically recover the table to a consistent state, ensuring that data is not lost or corrupted.

## Conclusion

The Delta Lake transaction log is a key component of Delta Lake’s data reliability and consistency, providing a complete history of all changes made to a Delta Lake table. This log is used to ensure data reliability, provide a complete audit trail of all changes, and to recover the table in the event of a failure. The Delta Lake transaction log is automatically maintained by Delta Lake and does not require any manual intervention, making it a powerful and convenient tool for ensuring data reliability and consistency.

---

[Christian Prokopp, PhD](https://www.linkedin.com/in/prokopp/), is an experienced data and AI advisor and founder who has worked with Cloud Computing, Data and AI for decades, from hands-on engineering in startups to senior executive positions in global corporations. You can contact him at christian@bolddata.biz for inquiries.

Discover the power of the Delta Lake transaction log - ensuring Data reliability and consistency.

Wood logs

A Guide to the Delta Lake Transaction Log

# The impact of Amazon's merchant datasets on E-commerce competition

By _Christian Prokopp_ on _2023-02-13_

Get the insights you need to compete with Amazon with a comprehensive dataset.

![Amazon share price](/content/blog/images/the-impact-of-amazons-merchant-datasets-on-e-commerce-competition.webp "Amazon share price")

Amazon has been accused of copying designs and undercutting prices on products sold on its platform, including accessories and low-tech items like trunk organisers \[1\]. The company was also implicated in a Wall Street Journal investigation that accused Amazon employees of accessing Data on competitors to develop its own products. Despite Amazon's policies against misusing nonpublic data, the company's founder Jeff Bezos could not guarantee that these policies were not violated. As Amazon expands into the consumer tech market, it has released a range of new devices that closely resemble preexisting products, including wearables and a smart thermostat. This has raised questions about Amazon's relationship with its competitors and its market power as both a sales platform and a consumer tech manufacturer. Regulators are reportedly keeping a close eye on the company's actions.

In today's competitive market, retailers face immense pressure to keep up with Amazon, the largest online marketplace in the world. Amazon's vast data collection and analysis capabilities give it a significant advantage in product development, pricing, and marketing strategies. To compete with Amazon, retailers must have access to comparable datasets and insights. However, creating such a dataset is costly and technically challenging, often outside a retailer's core competency.

This is where Bold Data comes in. Bold Data provides access to Amazon data from many countries, including data from hundreds of millions of product detail pages and billions of time-series entries for bestsellers. This data includes ranking, categories, pricing, and reviews and is critical for understanding the market and staying ahead of competitors. With access to this information, retailers can make informed decisions about product development, pricing strategies, and marketing efforts.

Many retailers lack the resources and technical expertise to create their own comparable datasets, making Bold Data a valuable service. Bold Data's data is collected, organised, and analysed by a team of experts and is available to retailers in an easy-to-use format. This allows retailers to focus on their core competencies, such as product development, category management, and marketing, without worrying about collecting and analysing data.

To succeed in today's market, retailers need comparable datasets and insights to Amazon. Bold Data provides access to this data, saving retailers time and resources and allowing them to focus on their core competencies. With Bold Data, retailers can make informed decisions about product development, pricing strategies, and marketing efforts and stay ahead of the competition.

**Contact Bold Data at** [contact@bolddata.biz](mailto:contact@bolddata.biz) **to learn more.**

* * *

\[1\] [Amazon Basics ripped off accessories, now Amazon is coming for gadgets](https://www.theverge.com/22701965/amazon-fitbit-ecobee-basics-design-halo-view-smart-thermostat-clone-ripoff-regulators)

---

[Christian Prokopp, PhD](https://www.linkedin.com/in/prokopp/), is an experienced data and AI advisor and founder who has worked with Cloud Computing, Data and AI for decades, from hands-on engineering in startups to senior executive positions in global corporations. You can contact him at christian@bolddata.biz for inquiries.

Get the insights you need to compete with Amazon with a comprehensive dataset.

Amazon share price

The impact of Amazon's merchant datasets on E-commerce competition

# Bing Chat argues and lies when it gets code wrong

By _Christian Prokopp_ on _2023-02-11_

Microsoft could follow Google's $100bn loss. I tried the new Bing Chat (ChatGPT) feature, which was great until it went disastrously wrong. It even started arguing with me while being wrong and making source code up.

![Pinocchio](/content/blog/images/bing-chat-argues-and-lies-when-it-gets-code-wrong/pinocchio.webp "Pinocchio")

To try Bing Chat, you must get on a waiting list, and then you are forced to use Edge. On a positive note, the experience is better than ChatGPT's version. Bing Chat is snappier, shows what is searched in the backend, and gives suggestions on continuing the conversation and the occasional references supporting its code and claims. Or so I thought.

![Bing Chat > Create a Python class AthenaQuery that caches the cursor object in the initialisation. The __init__ method should take the region, staging directory, and optional caching time. AthenaQuery also exposes a method for querying Athena which takes the query and database as the parameter.](/content/blog/images/bing-chat-argues-and-lies-when-it-gets-code-wrong/1.png)

I asked it to code a class to query Athena with Python. It looked good at first, but things went bad when I asked it to stream results into a feather file. In particular, it used an 'append' flag with the 'pyarrow.feather.write\_feather' method, which is not in the [documentation](https://arrow.apache.org/docs/python/generated/pyarrow.feather.write_feather.html) and the references it produced. One [reference](https://stackoverflow.com/questions/47113813/using-pyarrow-how-do-you-append-to-parquet-file) it produced used 'fastparquet.write()', which has an append flag. It may have confused the two.

![Bing Chat > save_feather() function with a temp file using append in write_feather](/content/blog/images/bing-chat-argues-and-lies-when-it-gets-code-wrong/2.png)

![Bing Chat > Are you sure you can use append in the write_feather function?](/content/blog/images/bing-chat-argues-and-lies-when-it-gets-code-wrong/3.png)

When I gave Bing Chat a chance to correct itself, I was surprised when it wrote, "I am not wrong.", and continued with source code to prove its point only to prove itself wrong unknowingly. To top it off, when I asked to show me where it got the code from, it directed me correctly to the '[pyarrow/feather.py](https://github.com/apache/arrow/blob/master/python/pyarrow/feather.py)' source on Github. But there, I found that the source code differs completely from the one it showed me.

![Bing Chat > You are wrong append is not a parameter in the write_feather signature.](/content/blog/images/bing-chat-argues-and-lies-when-it-gets-code-wrong/4.png)

![Bing Chat > Where did you find this source code?](/content/blog/images/bing-chat-argues-and-lies-when-it-gets-code-wrong/5.png)

In summary, Bing Chat generated code for me. And:

1. It imagined a non-existing flag and feature in an open-source library.
2. It refused to back down and argued it was right when given a chance to correct itself.
3. Its proof (source code) for being right showed it was wrong.
4. Worse, the proof was made up, and the referenced source is entirely different.

That is devastating. It produced incorrect code, failed to understand its mistake and faked source code with references. It did everything it could to throw me off and get things wrong.

---

[Christian Prokopp, PhD](https://www.linkedin.com/in/prokopp/), is an experienced data and AI advisor and founder who has worked with Cloud Computing, Data and AI for decades, from hands-on engineering in startups to senior executive positions in global corporations. You can contact him at christian@bolddata.biz for inquiries.

Microsoft could follow Google's $100bn loss. I tried the new Bing Chat (ChatGPT) feature, which was great until it went disastrously wrong. It even...

Pinocchio

Microsoft could follow Google's $100bn loss. I tried the new Bing Chat (ChatGPT) feature, which was great until it went disastrously wrong. It even started arguing with me while being wrong and making source code up.

Bing Chat argues and lies when it gets code wrong

# Google Unveils "Google Bard", its Answer to ChatGPT

By _Christian Prokopp_ on _2023-02-07_

The Battle of the AI Chatbots Begins: Google's Bard Takes on ChatGPT.

![Bard player](/content/blog/images/google-unveils-bard-its-answer-to-chatgpt.webp "Bard player")

Google announced the release of its latest conversational AI tool, "Bard." The new AI chatbot, designed to compete with the hyped ChatGPT, will be made available to "trusted testers" before being released to the public in the coming weeks. The company has also revealed plans to add new AI-powered features to Google search.

## Google's First Public-Facing Conversational Chat Product

This is Google's first release of a public-facing conversational chat product, and it is clear that the company's decision to do so directly responds to the success of ChatGPT, the AI conversation tool created by OpenAI. ChatGPT quickly gained 100 million users just two months after its launch and has caused immense pressure on Google to showcase its AI technology. Google CEO Sundar Pichai acknowledged this in a recent blog post: "AI is the most profound technology we are working on today."

## Bard's Capabilities and Potential Risks

Bard seeks to combine the world's knowledge with the intelligence and creativity of Google's large language models. This means it can explain complex information, give recipe ideas based on what's in your fridge, compare Oscar-nominated movies, and more. However, with the release of new technology comes potential downsides, and experts have warned of the potential harm conversational AI tools may cause if left unchecked. Pichai acknowledged these worries, saying, "It's critical that we bring experiences rooted in these models to the world in a bold and responsible way."

Google's first touchpoint with its conversational AI technology will be its search features that "distill complex information and multiple perspectives into easy-to-digest formats." The company's new AI tool will provide nuanced responses, such as when searching for "is the piano or guitar easier to learn, and how much practice does each need?" Google has yet to reveal the full extent of its AI capabilities, but it is clear that the company is stepping up to the challenge set by ChatGPT.

---

[Christian Prokopp, PhD](https://www.linkedin.com/in/prokopp/), is an experienced data and AI advisor and founder who has worked with Cloud Computing, Data and AI for decades, from hands-on engineering in startups to senior executive positions in global corporations. You can contact him at christian@bolddata.biz for inquiries.

The Battle of the AI Chatbots Begins: Google's Bard Takes on ChatGPT.

Bard player

Google Unveils "Google Bard", its Answer to ChatGPT

# Understanding the Power of ChatGPT

By _Christian Prokopp_ on _2023-02-07_

ChatGPT is a state-of-the-art language model developed by OpenAI, utilising the Transformer model and fine-tuned through reinforcement learning to produce accurate and ethical text responses.

![Robot teacher](/content/blog/images/understanding-the-power-of-chatgpt.webp "Robot teacher")

## Introduction to ChatGPT

OpenAI developed ChatGPT, a state-of-the-art language model. It is a machine-learning algorithm that processes and generates text. ChatGPT is part of the latest generation of Generative Pretrained Transformer (GPT) language models and uses the Transformer model. Unlike its predecessors, ChatGPT fine-tuned itself using reinforcement learning, making it capable of making accurate and ethical statements and following instructions.

## GPT Language Model

The GPT language model generates text using statistical probability. It processes a large text corpus and calculates the likelihood of each word pair appearing together. This creates short passages that sound natural but may not be grammatically correct or make sense. The model assigns a point in a semantic space to each word to represent its meaning. Word embedding techniques, such as Word2Vec, sort words into the semantic space by estimating their relationship to other words in the text.

## Attention Mechanism in Transformer Models

The attention mechanism is a crucial aspect of the Transformer model used in ChatGPT and other language models. It adds two vectors, query and key, to each word in a text and its meaning vector. These vectors are learned through training on large datasets and determine the strongest relationships between words in the text. The attention mechanism maps the source text from a simple semantic space into a new context space. Each dot represents a refined concept resulting from a word in the context of its most relevant neighbouring words.

## Attention Mechanism in ChatGPT

ChatGPT applies the attention mechanism multiple times - up to 96 times - to the mapped text, allowing for the creation of more complex and abstract concepts. The ultimate goal is to predict the next word in a text by transforming each word into its successor in the semantic space. The GPT-3 model trained on approximately 400 billion words of text from sources like Wikipedia and the Common Crawl dataset. The model adjusts its parameters with each incorrect prediction, leading to improved predictions over time.

## Limitations of GPT-3 Language Model

Despite its impressive deep learning process, GPT-3 has limitations. It often relies on the highest probability predictions, which can result in absurd or problematic answers. Additionally, its training data contains biases and offensive statements, which the generated text can reflect. There is also a lack of critical examination of generated text content, leading to ethical concerns.

## Overcoming Generalisation Hurdles with AI Teachers

OpenAI aimed to overcome GPT's limitations by incorporating human AI teachers in the latest version. However, due to cost limitations, they generated only a limited number of examples. To address this, a group of around 40 employees created over 10,000 examples of ideal performance for various tasks, which they then used to fine-tune the model. OpenAI trained a separate model to evaluate text quality, and GPT continued training itself based on this self-evaluation.

## Reinforcement Learning in GPT-3 and ChatGPT

Reinforcement learning involves an agent making decisions that affect its state, leading to either rewards or punishments. The agent bases its decisions on a policy, and the goal is continually improving the policy through the learning process. In the case of GPT-3, each decision involves generating a new word, and the reward is the language model's assessment of the task's quality. OpenAI initially used a fine-tuned language model as a policy, and the reinforcement learning algorithm refined the model to enhance the probability of producing high-quality answers. The combination of manual training and reinforcement learning allowed the language model to differentiate between good and bad text and improve its writing quality. The success of ChatGPT has sparked interest in language models for commercial use, with further investments expected in the future.

---

[Christian Prokopp, PhD](https://www.linkedin.com/in/prokopp/), is an experienced data and AI advisor and founder who has worked with Cloud Computing, Data and AI for decades, from hands-on engineering in startups to senior executive positions in global corporations. You can contact him at christian@bolddata.biz for inquiries.

ChatGPT is a state-of-the-art language model developed by OpenAI, utilising the Transformer model and fine-tuned through reinforcement learning to...

Robot teacher

ChatGPT is a state-of-the-art language model developed by OpenAI, utilising the Transformer model and fine-tuned through reinforcement learning to produce accurate and ethical text responses.

Understanding the Power of ChatGPT

# How to create a Data Dictionary using ChatGPT

By _Christian Prokopp_ on _2023-02-03_

ChatGPT can combine Data with natural language and has extensive information about most subjects. That lends itself to novel applications like creating informative data dictionaries.

![Robot reading](/content/blog/images/create-a-data-dictionary-using-chatgpt/robot-reading.webp "Robot reading")

Let us ask ChatGPT for a public dataset we can use for this how-to.

`> List public CSV datasets with links that I could use to demonstrate your ability to create a data dictionary from a CSV file.`

![ChatGPT > List public CSV datasets with links that I could use to demonstrate your ability to create a data dictionary from a CSV file.](/content/blog/images/create-a-data-dictionary-using-chatgpt/1.png)

Next, I downloaded one CSV file from the wine dataset and took a sample. ChatGPT can easily create a simple data dictionary table from it. But if we expand the question with some thought, it can make some valuable additions. For example, we can add SQL types, units of measure, descriptions expanded by ChatGPT's general know-how, and a summary for the table.

`> Create a data dictionary from the wine quality dataset for the red wine quality.   Add a column for SQL data types and favour DECIMAL over FLOAT.   Add a column for the Unit of Measure.   Create description fields using your knowledge of red wine for each column with at least two sentences each.   Make them sound natural and not repetitive.   Precede the data dictionary table with a summary paragraph for data users.`

![ChatGPT > Create a data dictionary from the wine quality dataset for the red wine quality. 1/3](/content/blog/images/create-a-data-dictionary-using-chatgpt/2.png)

![ChatGPT > Create a data dictionary from the wine quality dataset for the red wine quality. 2/3](/content/blog/images/create-a-data-dictionary-using-chatgpt/3.png)

![ChatGPT > Create a data dictionary from the wine quality dataset for the red wine quality. 3/3](/content/blog/images/create-a-data-dictionary-using-chatgpt/4.png)

The output is remarkable. Three of four columns have been added by ChatGPT using context and its knowledge base. Naturally, you would want to verify the details to ensure it fits your purpose, but it is an impressive first draft.

---

[Christian Prokopp, PhD](https://www.linkedin.com/in/prokopp/), is an experienced data and AI advisor and founder who has worked with Cloud Computing, Data and AI for decades, from hands-on engineering in startups to senior executive positions in global corporations. You can contact him at christian@bolddata.biz for inquiries.

ChatGPT can combine Data with natural language and has extensive information about most subjects. That lends itself to novel applications like crea...

Robot reading

ChatGPT can combine Data with natural language and has extensive information about most subjects. That lends itself to novel applications like creating informative data dictionaries.

How to create a Data Dictionary using ChatGPT

# Python TDD with ChatGPT

By _Christian Prokopp_ on _2023-02-02_

Programming with ChatGPT using an iterative approach is difficult, as I have demonstrated previously. Maybe ChatGPT can benefit from Test-driven development (TDD). Could it aid LLMs as it does humans?

![Test taking](/content/blog/images/python-tdd-with-chatgpt/test-taking.jpg "Test taking")

Previously, I wrote [Python code with ChatGPT starting simple and adding increasing details](https://www.bolddata.org/blog/how-to-code-python-with-chatgpt/). That worked initially and then started to fall apart over time. Let us try a simpler example and define tests before asking ChatGPT to write code to pass the tests. It is a good practice that helps many developers.

First, I outline the behaviour and ask for the appropriate tests.

`> Define a number of pytests for a function called is_asin. The function is_asin takes a string and returns a true or false. It returns true if the passed parameter passes the definition of an ASIN (Amazon Standard Identification Number). An ASIN is an alphanumeric identifier, 10 characters long and always uppercase. Write all the unit tests we need for the function is_asin.`

![ChatGPT > Define a number of pytests for a function called is_asin](/content/blog/images/python-tdd-with-chatgpt/1.png)

Not bad, but two tests are wrong, and also I prefer the happy path test at the beginning.

`> Remove the tests that return false for strings with only numbers or only characters. These are true and not false cases. Also move the test that returns true to the first place.`

![ChatGPT > Remove the tests that return false for strings with only numbers or only characters.](/content/blog/images/python-tdd-with-chatgpt/2.png)

Better, let us have the unit test code.

`> Write the code for the tests using pytest.`

![ChatGPT > Write the code for the tests using pytest.](/content/blog/images/python-tdd-with-chatgpt/3.png)

One issue is that the happy path only uses letters. We need a better test input string.

`> Update the string used in test_valid_asin to use both numbers and letters. Output all the tests.`

![ChatGPT > Update the string used in test_valid_asin to use both numbers and letters. Output all the tests.](/content/blog/images/python-tdd-with-chatgpt/4.png)

Good, now we can generate the function code.

`> Write the is_asin function to pass all the tests and use type annotation.`

![ChatGPT > Write the is_asin function to pass all the tests and use type annotation.](/content/blog/images/python-tdd-with-chatgpt/5.png)

Good try, but this is not going to pass. Let us see if ChatGPT can fix it without further details.

`> Do you have any specific recommendations to improve is_asin?`

![ChatGPT > Do you have any specific recommendations to improve is_asin?](/content/blog/images/python-tdd-with-chatgpt/6.png)

The documentation is a good addition and probably something to ask for from the beginning. It should help ChatGPT to stay on track as we iterate. However, the happy path test fails when executed as expected. Let us see if we can get it fixed by being vague.

`> The test test_valid_asin fails. Can you fix is_asin?`

![ChatGPT > The test test_valid_asin fails. Can you fix is_asin?](/content/blog/images/python-tdd-with-chatgpt/7.png)

It still is failing. Let us try and give some more detail using simple language and not wordsmithing it for ChatGPT's benefit.

`> The test test_valid_asin fails because is_asin should check all characters to be numbers or upper case characters. But it checks if they are both at the same time which cannot be true. Can you fix is_asin?`

![ChatGPT > > The test test_valid_asin fails because is_asin should check all characters to be numbers or upper case characters.](/content/blog/images/python-tdd-with-chatgpt/8.png)

It still fails. Let us state a solution instead to help ChatGPT.

`> The test test_valid_asin fails because is_asin should check all characters to be either an upper case letter or a number. Fix it.`

![ChatGPT > The test test_valid_asin fails because is_asin should check all characters to be either an upper case letter or a number. Fix it.](/content/blog/images/python-tdd-with-chatgpt/9.png)

Success.

The great part is that we can generate a good list of tests based on the description and some decent code to solve it. Clearly, that can be useful in the future for developers. ChatGPT also needs precise language where possible, which forces and helps with reflecting on the breakdown of the problem and its description as part of the development work.

However, the gap for ChatGPT is the lack of understanding of what it generates. Simple functions like this trip it up and need experience and the ability to understand the code in detail by the user.

As it stands, ChatGPT can help expert users with simple tasks. The opportunity for a significant productivity boost is to move both dials, i.e. to help inexpert users with complex tasks, ideally. The interesting question is if this future is one, five, ten or more years away.

---

[Christian Prokopp, PhD](https://www.linkedin.com/in/prokopp/), is an experienced data and AI advisor and founder who has worked with Cloud Computing, Data and AI for decades, from hands-on engineering in startups to senior executive positions in global corporations. You can contact him at christian@bolddata.biz for inquiries.

Programming with ChatGPT using an iterative approach is difficult, as I have demonstrated previously. Maybe ChatGPT can benefit from Test-driven de...

Test taking

Programming with ChatGPT using an iterative approach is difficult, as I have demonstrated previously. Maybe ChatGPT can benefit from Test-driven development (TDD). Could it aid LLMs as it does humans?

Blog