• The Prompt
  • Posts
  • AI At The Helm: Navigating the Shift Towards Autonomous Decision-Making

AI At The Helm: Navigating the Shift Towards Autonomous Decision-Making

Breaking Information Silos, Mastering Code Execution, and Redesigning User Experience: The Unstoppable Evolution of AI

June 07, 2023

AI At The Helm: Navigating the Shift Towards Autonomous Decision-Making

Welcome back to another edition of The Prompt.

Here's your 3-2-1 summary:

🧠 3 eye-opening insights

💡 2 practical tips

❓ 1 thought-provoking question

Total read time: 3 minutes

Let's go!

3 Insights

I. Code Comprehension and Execution by AI

AI, specifically large language models like GPT-4, with the ability to understand and generate code has another ability not widely discussed. It is capable of executing code. Not only that it can execute or think through any logic, with a high degree of accuracy.

This means that it can blend logic and human decisionmaking.

Image a customer requesting money back after purchasing a blender, after the 30-day return policy. Why? Because it doesn’t blend the smoothie enough.

With a language model, it can process the request, compare it against the business rules and make a thoughtful decision based on the request.

Check out the full conversation with GPT-4, setting up the example and testing the rules.

Refund Request Evaluation Process - A GPT-4 conversation

This is a conversation between a human and a GPT-3 chatbot. The human first asks: Create a business rule about one customer and a product, that ends with a human decision. A request, where the customer writes a piece of text. There should be two numeric figures that should feature, but often the logic doens’t works if the reality is not that simple. Then a human steps in and evaluates. Write the business rule, the request text, and the decision..

II. The Dichotomy of UX in AI

The design of AI-first systems presents a unique challenge. On one hand, it involves simplifying and reducing decisions. On the other hand, it requires obtaining necessary user inputs at the right time.

This has been a persistent challenge in designing Aloy.so, where balancing user experience with decision-making is a constant tug of war. Because if the AI is doing all the work, there is no UX, right?

In an interview with a VC she asked: “So Aloy would help me write notes while taking a meeting?”

My response: “No, Aloy would participate, be invited as a member of the Zoom call, and would record all the things said in the meetings, then pull up key figures, and relevant people to connect to, from your second brain, effortlessly.”

What is the complete UX? No definitive solution exists yet. Our current UX principles are to simplify, trust the AI, or revert to human input. We do design this with rapid iterations, GPT-4's coding capabilities and feedback from beta testers are driving the development of next-generation AI user experiences.

Do you want to become a beta tester, building your digital twin, your AI second brain? Then join the beta program at Aloy.so.

III. Why raise money when AI does all the work?

In a few of the meetings with VCs I have had the past few weeks one topic has popped up:

“Do startups need as much cash, if GPT-4 is doing a large part of the work?”

I think the short answer is they get much further faster, being able to leverage the capital better to bottlenecks GPT-4 cannot help with.

In less advanced tech companies, I see a reluctance to use GPT-4 extensively, relying on a co-pilot to do the smaller task. Coding is where GPT-4 can really automate a lot of the work, but there is pushback. Who wants to lose their job right?

My response: “The question from a VC to a startup should not be whether they need less money, but rather whether the startup utilizes GPT-4 to automate the development process and build at scale.”

2 Actionable Tips

I. The best mentor in the world

The highest value use case I find with GPT-4 is its ability to take in my situation and apply its own knowledge to provide advice.

Sometimes it fails spectacularly, which might lead you on a wild goose chase if you are not critical.

To avoid this, I ask for advice on problems I am familiar with. Be it technology choice, business strategy, or even newsletter writing. More often than not I find it gives input and a spotlight on my blindspot. Making me more confident in my own decisions.

II. AI Notion creator

Using Notion? Planning to create something new? Just ask GPT-4:

Prompt:

I am creating a new notion page for my newsletter. Do the following:

1. Define the problem or task I am doing.

2. Describe a userstory of me using the notion page to solve my problem.

3. List the elements to use on the notion page.

4. Design the notion page.

5. Provide an example of how the page would be

The result?

A complete design with example text, and tables with columns for you to implement.

Maybe someone will even automate the setup of the Notion page and tables?

1 Thought-Provoking Question

In an AI-driven future where models can comprehend, generate, and execute code, how prepared are we, as individuals and society, to make the significant mental shift required to relinquish control and trust the decisions made by AI?

Let me know what you think of this issue

Login or Subscribe to participate in polls.

I consistently curate and incorporate top-tier ChatGPT prompts and resources into our collection.

What are your thoughts on this issue? Reply to this email and let me know!

— Alexander