Skip to main content
  1. đź“š Notes/

Impromptu - by Reid Hoffman

A nice quick read to gain deeper understanding of the current state of AI and how to best use it.

Notes #

GPT-4 is not a conscious being at the front of your own wondrously human mind. In my opinion, this awareness is key to understanding how, when, and where to use GPT-4 most productively and most responsibly. At its essence, GPT-4 predicts flows of language. Trained on massive amounts of text taken from publicly available internet sources to recognize the relationships that most commonly exist between individual units of meaning (including full or partial words, phrases, and sentences), LLMs can, with great frequency, generate replies to users’ prompts that are contextually appropriate, linguistically facile, and factually correct.

in your overall quest for authoritative information, GPT-4 helps you start somewhere much closer to the finish line than if you didn’t have it as a resource.

GPT-4 arranges vast, unstructured arrays of human knowledge and expression into a more connected and interoperable network, thus amplifying humanity’s ability to compound its collective ideas and impact.

Human beings should view a powerful large language model (LLM) as a tool, not as a source of truth, authority, or intelligence.

[If ChatGPT] can do a job as well as a person, then humans shouldn’t duplicate those abilities; they must surpass them.

humans could thrive alongside AI by: (1) specializing in asking the best questions, (2) learning insights or skills that are not available in the “training data” used by the deep learning networks, and (3) turning insights into actions.

large language models are also different from these technologies in that they can produce outputs that are not based on existing sources or inputs, but on their own learned patterns and probabilities. Students may need to verify, interpret, or modify the outputs of large language models, as well as understand the limitations, biases, or errors. Moreover, large language models can also learn from the feedback or interaction of students and teachers, which can create a dynamic and collaborative learning environment.

in early March, MIT researchers reported that two studies2 on generative AI’s effects on knowledge work found that writers and programmers both saw 50 percent gains in productivity with AI, and higher satisfaction to boot. Wharton professor Ethan Mollick called the results “completely unprecedented in modern history

But I do know enough about tech to be confident that GPT will be the next technology platform—on a scale like the internet, and very possibly bigger—on which the world will build applications, tools, and services of these sorts and literally a million others. And it’s going to happen fast, not least because GPT tools will themselves accelerate the development of all the rest of it.

When I graduated from college in 1990, jobs like “web designer,” “SEO strategist,” and “data scientist” didn’t exist.

In my opinion, ignoring AI is like ignoring blogging in the late 1990s, or social media circa 2004, or mobile in 2007. … Developing skills and competencies in it now will yield benefits for years to come.

Write a story of around 400 words about a manager and employee in the year 2035 using AI to work together to define a tour of duty for that employee, and leveraging AI to help that employee achieve greater productivity and accelerated career development. Illustrate the benefits to the company, the manager, and the employee of this approach.

When transformational technologies appear, most people are tempted to plug them in as a substitute for an existing technology or technique. … but this approach is actually a trap: seldom is the new technology an exact analogue for what came before.

Principle 1: Treat GPT-4 like an undergrad research assistant, not an omniscient oracle.

It also has many of the other drawbacks of a human research assistant: it’s not an expert, its grasp of any particular subject is fairly shallow, and it gets things wrong. In fact, when it’s wrong, it’s worse than a human research assistant, since a human will often have the good sense to warn you when he or she isn’t certain about the quality of their output.

Principle 2: Think of yourself as a director, not a carpenter.

You have to coax out your desired result, and you may need to try many variations on a prompt in the same way a director might need to film multiple takes of the same scene. It’s ultimately a collaborative process.

Principle 3: Just try it!

GPT-4 provides, with unprecedented speed and scale, each of us with probabilistic syntheses of the world’s existing words to consider as inputs, challenges, and inspirations for our own work.

Human beings are natural problem-solvers, and technology often becomes the vehicle through which we address challenges and make improvements to our lives.