CrowdAI, ChatGPT and GPT-3: Empowering the Business Analyst

For the last two decades, AI has been at the forefront of deep tech. But what has accelerated interest in the last few months, is the ability for the technology to be translated into an approachable and human readable interface.

AI is implemented in small doses throughout everyday experiences - from smarter email reply, to doorbell cameras, to weather forecasting. But what makes now different? The answer is more people are interacting directly with AI tools rather than seeing it as a function of a larger product. In the enterprise sector for domain specific models, the barrier to implementing artificial intelligence solutions has always been the interpretation and subjectivity of the quality of results. Objective metrics like F1, precision, recall always are swayed by the training data set and validation set. It's all about how you can interpret results for the question you are trying to answer. 

Let me take you through CrowdAI’s workflow:

At CrowdAI, we built a no-code platform that enables anyone to create a computer vision model. A user can upload data, label that data, build a solution that can automatically identify what you are looking for, and output the results.

The reason no-code solutions are so powerful is because it puts AI technology in the hands of a wide range of subject matter experts. For example, a person who spends a good portion of their day visually inspecting something does not always have the programming experience or time to build a model from scratch. With CrowdAI, they can turn their expertise into a computer vision model in a matter of days to automate their visual inspection. 

For example, CrowdAI ingests geospatial imagery data, and its associated metadata. Then, our platform builds a detection model that will locate whatever the model was trained to identify in a certain region (airplanes for this example). The model results in an output like this (GeoJSON, CSV, JSON):

{"meta": {"title": "gegd_featureb0a8a6a0e137e018f5affc3af31964ed_aoi61c991afd00be38e7522fc3cf4125d67.tif", "label_type": "segmentation", "media_type": "image", "media_id": 5615304, "project_id": 493, "media_size": [1533, 1468]}, "features": [{"geometry": {"type": "Polygon", "coordinates": [[[840.7375575412008, 681.3248341832327], [825.408460540938, 663.1077224408515], [844.3739397624753, 657.6553308070003], [859.0191059605844, 667.9611291983032], [840.7375575412008, 681.3248341832327]]]}, "properties": {"feature_type": ["commercial", "Plane"]}}]}

‍

‍

Great! We created a model that detected two airplanes and outputted a string of text.The challenge is that there is still work to be done to deliver the insights to a client in the best way possible, and this is where ChatGPT comes in.

Sending an output of a JSON, CSV, .xls, and/ or GeoJSON file may not support your business analyst needs, and this is where ChatGPT can help. ChatGPT outputs information in a very conversational way, and this helps make highly technical subjects more accessible to the everyday user. 

‍ChatGPT to Interpret Files types: 

After creating no-code AI solutions, ChatGPT can help make the results more approachable for the average user. For example, ChatGPT helped to break down key components of the GeoJSON above to help the user understand what it means:

“This GeoJSON (JavaScript Object Notation) file contains metadata and features of an image.

The "meta" section contains information about the image, such as its title, label type, media type, ID, project ID, and size.

The "features" section contains information about the different features of the image…”

ChatGPT to help Business Analysts with VBA and Macro:

Using CrowdAI’s model output JSON or CSV in conjunction with ChatGPT, can give users insight to what the findings mean in the real world; yet it can still be a massive amount of information to sort through. Leveraging Excel, a tool most business analysts are familiar with, can open up a whole new world of quick data inspection. 

ChatGPT makes writing macro in the Visual Basic Editor (VBE) more accessible. For example, once you build a model in CrowdAI, the output of the model can be in a file format that you can pull to Excel everyday. 

You can ask questions like: In Excel, how do you identify all planes of one type, for one date, across multiple spreadsheets in one excel workbook?

GPT3 for the Business Analyst Power User: 

ChatGPT (Chat Generative Pre-trained Transformer) is a chatbot launched by OpenAI. It is built on top of OpenAI’s GPT-3 family of large language models.

Combining GPT-3 with CrowdAI’s platform output CSVs supercharges data analysis, data queries & reasoning too (covers most of the types db/warehouse mySQL, BigQuery, etc). You can use GPT-3 in Excel by using the OpenAI API and the Excel WEBSERVICE function. The WEBSERVICE function allows you to call a web service from a cell in an Excel spreadsheet, and the OpenAI API allows you to access GPT-3 from a web service.

From there, you can use GPT-3 generated functions directly in Excel to search, summarize and uncover the size of the airplane, its longitude, and latitude coordinates, airport, city, country, time of day and date the aircraft was in this location, along with monitoring patterns of change. 

ChatGPT and GPT-3 enables Business Analysts to go further on CrowdAI’s Platform: 

This powerful combination allows our users to summarize data into easy to read reports via Excel. Our goal at CrowdAI is to make artificial intelligence more accessible, and providing model outputs in an accessible manner closes the gap between no-code AI solutions and delivering results for business analysts. Using ChatGPT and GPT-3 in conjunction with CrowdAI helps business analysts focus on deriving insights instead of interpreting the results.

What's next? Direct integration for analysts to have one seamless workflow from raw pixels to business analyst friendly reports.

‍