Reading · ajit’s Substack

This page did not load. You can still read it on the original site — the toolbar below keeps your place in the directory.

Background

Yesterday, we spoke at SIU in Carbondale about our plans for rethinking education in the LLM/ Generative AI era.

I have discussed these ideas before as the co pilot first approach and we have been working with these ideas in our teaching at the #universityofoxford.

Thanks for reading ajit’s Substack! Subscribe for free to receive new posts and support my work.

This is my second trip to Carbondale and we continue to extend our collaboration with Pinckney Benedict Joddy Murray and their team on LLMs

There are broadly three modalities of generative AI i.e. code, images and language.

We, at the #universityofoxford, are primarily involved in the code and language aspects of LLMs - especially in context of the Azure OpenAI services. For AI systems that can create realistic images and art from a description in natural language such as dall-e-2 - we collaborated with Pinckney's team at SIU.

Of course with GPT-4 (see the paper Sparks of Artificial General Intelligence: Early experiments with GPT-4), we are seeing multimodal AI applications

Yesterday, Ayse also presented about the potential of Github copilot X and Copilot 365

Reengineering business processes using LLMs and prompt engineering

I have also been thinking of extending this idea of rethinking processes more widely beyond education - and to the enterprise. More along the lines of the book Reengineering the corporation but for the GPT/LLM world. - specifically rethinking of end to end processes using prompt engineering and low code / power platform approach.

In practise that means,

a) understand a business process as it stands now

b) rethink it using the LLM / prompt strategy

c) design the workflow using power tools

d) implement this structure to demonstrate end user benefits

The basic structure of process reengineering is the mature i.e.

Source: https://run.unl.pt/bitstream/10362/150111/1/TGI829.pdf

The question is more:

How do we reimagine existing processes using LLMs and prompt engineering?

Power apps

We have been working with the Power apps suite and especially how the next generation ai copilot within power apps will transform low code development

From the above link

With the copilot control in Power Apps, app makers can give their users the ability to get intelligent insights about the data in their apps through a conversational chat experience.
Users can use natural language to create queries and refine their analysis, all with the help of an AI assistant embedded directly in the running app.
Whether a user needs to understand changes in their inventory, estimate average time to complete a task, or explore which of their campaigns drove the most revenue, the copilot control is ready to listen, analyze, and report.
Here are a few examples of natural language prompts your users can use:
“What are the most common reasons an inspection fails?”
“Who are the most active inspectors for that type of issue?”
“Which equipment is most often inspected?”

The potential

So, we could then look at the end to end process and rethink each process using prompt engineering as a mechanism for GPT/LLM. The tool set looks as below