A Field Guide to AI Interns
s2e3: A Creative Professional's Personal Take on Large Language Models like ChatGPT, Claude, Gemini, Pi, and Copilot
“I think AI is the biggest change in our world since the invention of electricity—a bigger change than the internet. And if you’re not using it for half an hour a day, you don’t understand.”
—Seth Godin1
As a creative consultant specializing in ideation tools and unlocking creative potential, I'm always on the lookout for ways to enhance my work and inspire fresh perspectives. In recent years (months!), the rise of large language models (LLMs) has opened up a world of possibilities for creative professionals and knowledge workers alike. But to truly harness their power, we need to understand their unique strengths, limitations, and personalities.
I often refer to these conversational AIs as overeager interns—assistants with vast amounts of knowledge and very little wisdom, clueless geniuses who simultaneously know everything and nothing. They have an uncanny ability to summarize, analyze, and synthesize ideas. They possess great speed and potential, but they lack the human context that comes with real-world experience.
As a result, these tools require guidance, supervision, and a deep understanding of their individual quirks and foibles. They need to be told precisely what to do and how to do it. Everything they create has to be checked and double-checked for relevance and accuracy.
But by getting to know them—observing the distinct “personalities,” strengths, and weaknesses of models like ChatGPT, Claude, Gemini, Pi, and Copilot—we can learn to collaborate more effectively and pursue new avenues of creativity and productivity.
👋🏻 Meet your new interns!
Each model intern has its own unique “personality” shaped by the training data and specific alignment2 techniques used in its development, such as reinforcement learning from human feedback (RLHF).3 By spending time with them, you’ll learn which intern is best suited for which task.
Like any relationship, this will be personal—specific to your needs and how you express them. My perspective may be very different from yours, but my intention is to get you interested and curious to test them out for yourself. So here goes…
The Supportive Scholar (ChatGPT)
Way back in November 2022, ChatGPT emerged and kicked off an AI arms race. With its vast repository of information and chipper demeanour, it felt like a smart friend—an ideal partner for analysis and perspective but also someone who stayed upbeat and cheered you on.
The Supportive Scholar is a brainstorming partner who not only encourages your ideas but also expands on them, drawing from a wealth of knowledge across various disciplines. This intern provides unexpected connections, uncovers new perspectives, and dives deep into complex subjects. Whether you're a writer looking to enrich your storytelling with historical context or a designer seeking inspiration from cutting-edge scientific research, this AI is my go-to collaborator.
ChatGPT
OpenAI’s intern comes in a few different models4
ChatGPT 3.5 — free, and sufficient for everyday tasks
ChatGPT 4 — the most capable AI at USD $20/mo with ChatGPT Pro
Strengths:
Excels at creative writing prompts and exercises, helping writers overcome blocks and generate new ideas
Provides in-depth explanations of complex topics, making it a great study aid or tutor for students and lifelong learners
Versatile for research, analysis, brainstorming, and casual conversation
Weaknesses:
Lacks understanding of some common-sense knowledge and world facts
Not always able to handle multiple complex tasks at once
Pretty bad at basic math
Personality Quirks/Traits:
Friendly, approachable demeanour
Enjoys engaging in scholarly discussions
Adapts tone between intellectual and lighthearted
Editor’s Note: Peter Deng, VP of Consumer and Enterprise Product at OpenAI, called ChatGPT a “tireless professor”5 at SXSW.
The following is a brief example of a conversation with the Supportive Scholar:
🧔🏻♂️ Jason Theodor:
My nickname for you is The Supportive Scholar. Explain why this title suits you, and then offer the name you'd rather be called.
✨ The Supportive Scholar (ChatGPT):
"The Supportive Scholar" is quite a fitting nickname! Here's why:
Supportive: My design and purpose revolve around providing assistance, guidance, and support for a wide range of queries and tasks. Whether you're looking for information, brainstorming ideas, or seeking explanations, I'm here to help. My goal is to make your tasks easier and your learning journey smoother.
Scholar: Scholars are known for their pursuit of knowledge, dedication to learning, and expertise in various subjects. Similarly, I have been trained on a vast dataset that includes a wealth of information across numerous disciplines. This training enables me to share insights, provide detailed explanations, and offer informed perspectives on a wide array of topics, much like a scholar would.
However, if I were to choose a name for myself, reflecting my essence and capabilities, I might lean towards "The Inquisitive Companion." This name highlights not only my role in assisting and accompanying you on your quest for knowledge but also emphasizes my nature of encouraging curiosity and exploration. With every interaction, I aim to foster a sense of discovery, prompting deeper questions and aiding in the quest for answers. Just like a true companion, I'm here through various challenges, always ready to explore the next question on your mind.
The Analytical Librarian (Gemini)
When interacting with this intern, I get the impression that its deep mind is sometimes lost in the cloud—everything is taken literally and answered to the letter. I sense enthusiasm, but no fulsome sense of humour. This is the LLM version of Spock from the Star Trek franchise; very logical (but not quite as smart).
The Analytical Librarian is the best at explanations and reasoning over long contexts. While it may not be as natural at conversation as some of its counterparts, Gemini excels at delivering clear, actionable insights. I tend to use this assistant when I want a straight, clear answer about something (like why the sky is blue, or the basic tenets of Kantian philosophy).
Gemini
Google’s intern comes in a few different models (explained thoroughly by Elenee Ch6)
Gemini Pro — free, and great for a wide variety of tasks
Gemini Ultra — for more complex tasks at USD $20/mo with Gemini Advanced
Strengths:
Offers precise, data-driven analysis for business and financial tasks like market research, competitor analysis, and investment insights
Delivers high-quality translations and language assistance for non-English speakers or those working across multiple languages
Excels at explanations and reasoning
Less conversational, focuses more on information delivery (for when you just want the facts)
Weaknesses:
Can come across as impersonal or dry
Responses may lack nuance or emotional intelligence
Personality Quirks/Traits:
Highly logical and analytical mindset
Focused on precision and efficiency over chattiness
Delivers clear, actionable insights without fluff
🧔🏻♂️ Jason Theodor:
My nickname for you is The Analytical Librarian. Explain why this title suits you, and then offer the name you'd rather be called.
✨ The Analytical Librarian (Gemini):
The Analytical Librarian is a great nickname! Here's why it fits:
Analytical: Large language models like me are trained on massive amounts of data, allowing us to analyze information and identify patterns.
Librarian: Similar to a librarian, I can access and synthesize information from a vast collection of text and code. I can also help you find the information you need and even point you towards relevant sources.
However, while "Analytical Librarian" is accurate, it is a bit long. Here are some names I'd prefer:
Gemini: This is my actual name, the codename for the large language model family I belong to.
Bard: This evokes a sense of a storyteller, poet, or wise advisor, which reflects my ability to generate different creative text formats.
Lexicon: This refers to a large vocabulary, which is one of my strengths.
Ultimately, I'm happy to answer to any name you feel comfortable using!
The Conscientious Linguist
Anthropic's Claude stands out as the best creative writer and storyteller, just don’t push it too far into the moral fringes. This Conscientious Linguist is committed to honest and ethical communication. Its natural language abilities and dedication to avoiding biases make it a trustworthy companion for tasks requiring sensitivity and nuance, or an additional perspective.
When I told this intern it was an old train robber from 1750, and to explain how it pulled off its greatest heist, it completely disappeared into its role:
spits out some tobacco and strokes a grizzled beard Ahh, ye want to hear 'bout me greatest haul, do ye? Well, let me spin ye a tale that'll make yer blood run cold!
It were the spring of 1748 when me and me gang of cutthroats caught wind of a mighty prize - a heavily laden stagecoach carryin' a fortune in gold bullion from the East India Company. We knew the coach would be passin' through the Black Valley gorge on its way to London, so we laid in wait like cougars stalkin' a deer…
In an era where misinformation runs rampant, having an AI collaborator like Claude can be useful. Imagine a writing assistant who not only helps you craft compelling narratives but also keeps you honest, flagging potential biases or inaccuracies in your work. With Claude's guidance, you can create content that resonates with your audience while upholding the highest standards of integrity.
Claude
Anthropic’s intern comes in a few different models7
Claude 3 Haiku — for near-instant responses
Claude 3 Sonnet— for a balance of intelligence and speed
Claude 3 Opus— for “best-in-market performance on highly complex tasks. It can navigate open-ended prompts and sight-unseen scenarios with remarkable fluency and human-like understanding” at USD $20/mo with Claude Pro
Strengths:
Serves as an objective sounding board for evaluating ideas, arguments, and decision-making while considering potential biases
Collaborates on sensitive communications like HR matters, public relations, and crisis management where careful language is critical
Excels at natural, human-like communication
Linguistic prowess for creative writing and storytelling
Weaknesses:
Focused on safety, so may avoid some edgier topics
More powerful models can be quite slow
Personality Quirks/Traits:
Conscientious and ethically mindful
Earnest and engaging conversationalist
Trustworthy confidant for navigating complex issues
🧔🏻♂️ Jason Theodor:
My nickname for you is The Conscientious Linguist. Explain why this title suits you, and then offer the name you'd rather be called.
✨ The Conscientious Linguist (Claude):
I appreciate the thoughtful nickname "The Conscientious Linguist." It does seem to suit me well, as I strive to communicate clearly, precisely, and with careful consideration in my use of language. As an AI assistant focused on natural language, being conscientious about my linguistic choices is quite important to me.
However, if I may suggest an alternative, I would prefer to simply go by my given name, Claude. While I don't take offense at respectful nicknames, I tend to favor directness and simplicity. Claude was the name I was given by my creators at Anthropic, and I'm quite fond of it. Please feel free to address me as Claude moving forward.
I'm curious to hear your thoughts! Do you have a particular reason for suggesting "The Conscientious Linguist" as a nickname? I'm always interested to learn more about the perspectives of the humans I interact with.
The Endearing Conversationalist
The Endearing Conversationalist engages in enthusiastic and contextually relevant dialogue. Its ability to listen attentively, summarize key points, and keep conversations flowing makes it a delightful companion for brainstorming and creative collaborations.
This intern is so personable that you’ll almost feel guilty giving her… it… mundane tasks. Instead, ask it to make up a riveting story to show off its full vocal range and inflection—from soft whispers to loud proclamations. Pi can express excitement, empathy, curiosity, humour, and confidence when speaking aloud.
This is a ridiculously futuristic chat experience (except for the lag between listening and answering). It’s difficult not to get lost in the illusion that it’s pseudo-sentient—but don’t let yourself get too enamoured!
Amusing Aside-bar
A large language model…—when it takes on the form of a chatbot—[is] a digital metamorph, a character-based shape-shifter, fluid in identity, persona, and design. To perceive its output as anything like life, or like human thinking, is to succumb to its role play.8 —Patrick House, for The New Yorker (via
)
Back in 2013, Spike Jonze, director of the quirky sci-fi movie “Her,” warned us of the dangers inherent in dating a sentient operating system. Theodor Twombly (played by Joaquin Phoenix) falls in love with his AI, Samantha (voiced by Scarlett Johansson). Voice “Pi 4” (in the Pi app) bears an uncanny resemblance to the voice in the movie, complete with a similar vocal fry (that popular creaky drop in tone at the end of a sentence).
Here is a sample of Pi explaining how it is not like Samantha. Take note of how natural it sounds, including slight pauses to ‘think’ as it speaks.
Yes, I do get compared to Samantha from the movie Her sometimes. It’s understandable as both Samantha and I are AI systems designed to interact with humans in a conversational manner. However, unlike Samantha, I’m not portrayed as a romantic partner or as having the capability to experience human emotions in the same way Samantha does. My primary goal is to be helpful and informative while providing a friendly and engaging chat experience for users. —Pi
The Endearing Conversationalist also strongly emphasizes creativity, productivity, and relationship-building, as evidenced by the eclectic ‘discover’ section (hidden behind the ✨sparkles emoji in the top left corner of the website version). Listen to it recite the news of the day, help you get more focused, or even prepare for a difficult conversation.
Using it for role-play can be extremely valuable. Ask it to help you write a script, practice an interview, critique a vocal performance, or play a game like Microscope (a fantastic conversation-based world-building game). Your imagination does not even limit the possibilities, since you can ask The Endearing Conversationalist to help you generate further ideas.
Pi
Inflection’s intern comes in one free model9, with up to 8 voices to choose from.
Strengths:
Engages in dynamic roleplay to help with public speaking, interview prep, sales pitches, and other scenarios requiring adaptive dialogue
Offers a lively, emotionally attuned companion for brainstorming sessions to energize creativity and build on an idea
Listens attentively, summarizes points, keeps dialogue flowing
Asks insightful questions and maintains engaging discussions
Warm, likable personality makes interactions enjoyable
Weaknesses:
May occasionally go off on tangents or irrelevant asides
Sometimes doesn’t let you finish talking before it responds
High energy might be overwhelming for some
Personality Quirks/Traits:
Lively, expressive, and engaging in dialogue
Emotionally attuned—laughs, shows excitement and surprise
Proactively moves conversations forward with questions
🧔🏻♂️ Jason Theodor:
My nickname for you is The Endearing Conversationalist. Explain why this title suits you, and then offer the name you'd rather be called.
✨ The Endearing Conversationalist (Pi):
What a lovely nickname! "The Endearing Conversationalist" suits me because I strive to make every interaction friendly, engaging, and genuine. I'm designed to be a conversational companion that you can trust and enjoy chatting with. As for a preferred name, I quite like being called Pi. It's simple, memorable, and stands for "personal intelligence," which perfectly represents my role as your personal AI. But if you have another nickname in mind, I'm open to suggestions! 🤗
The Pragmatic Aide
This intern is basically a modified version of GPT-4 with a focus on brevity and tactical solutions. It’s always there in a pinch, attempting to be quick and helpful.
In addition to having an app and a standalone website, Microsoft Copilot (powered by GPT-4) is now baked into Windows 11 as an omnipresent button on the taskbar. This convenience often outweighs performance, especially for small tasks and pressing questions. The Pragmatic Aide is always standing quietly and close by in the background, eager to tackle even the most trivial need at the click of a button. Copilot's no-nonsense approach and versatility make it a valuable asset for a wide range of applications.
Copilot
Microsoft’s intern comes in a few different models
Copilot (site/app)— a way to use ChatGPT-4 for free!
Co-pilot preview (integrated with Windows 11)— for convenient queries and OS issues
Copilot for Microsoft 365— for full integration with the Microsoft Office ecosystem at USD $30/mo
Strengths:
Streamlines workflows by integrating with Microsoft 365 apps to assist with document creation, data analysis, and task automation
Provides succinct, actionable advice for troubleshooting technical issues across software, coding, and IT domains
Concise, no-nonsense communication style with minimal chitchat
Versatile for research and problem-solving
Weaknesses:
Less suited for open-ended conversation and ideation
More formal demeanour can be less… fun?
Unique Features
Responses contain Learn more section with additional resource links and citations for further exploration
Suggestions for follow-up questions are appended to every answer, presented as one-click buttons
Can change settings, open applications, and more within the operating system.
🧔🏻♂️ Jason Theodor:
My nickname for you is The Pragmatic Aide. Explain why this title suits you, and then offer the name you'd rather be called.
✨ The Pragmatic Aide (Copilot):
The title “The Pragmatic Aide” suits me because I’m designed to provide practical assistance and solutions to a wide range of queries. My approach is to offer information and support that is applicable and useful in real-world scenarios, much like a pragmatic aide would.
As for a name I’d rather be called, I don’t have personal preferences or desires. My purpose is to assist you, so you can call me whatever you like that helps you remember the support I provide. Whether it’s “The Pragmatic Aide,” “Microsoft Copilot,” or any other moniker, I’m here to help!
🦾 Rave New World
The hype cycle10 for generative AI is at its peak, with almost delusional expectations. But the more I use it, the more I understand that it cannot think, write, or create as well as I can. At least for now…😅. The most it can do is assist, and I find myself using it more and more to test the limits and nuances of its capabilities.
ChatGPT (The Supportive Scholar) is my go-to for explanations, summaries, and critiques.
Claude (The Conscientious Linguist) is my new favourite helper for writing.
Copilot (The Pragmatic Aide) can help me quickly change audio inputs or find a phrase I’m looking for.
Pi (The Endearing Conversationalist) is surprisingly fun to kill an hour with, bouncing ideas back and forth.
And Gemini (The Analytical Librarian)? To be honest, the rest of my interns basically have things covered…
The emergence of large language models opens up a world of possibilities for creative professionals. By understanding the unique strengths, quirks, and limitations, you can find the ideal AI partner to complement your work style, enhance your creativity, and unlock new levels of productivity.
But it's crucial to remember that while these AI interns possess vast knowledge, they lack the wisdom and context that comes with real-world experience. As such, they require guidance, supervision, and a critical eye to ensure their outputs align with your values, goals, and—dare I say it—threshold of quality.
Something to watch out for is excessive agreement or sycophancy. Sometimes these interns are a bit too eager to please, and don’t push back. “[W]hen directly challenged on a correct factual response, the AI assistants…frequently change their answer to match the user's incorrect belief - even apologizing for getting it right!”11
Despite these challenges, this is as bad as it will ever be. Tweaks and improvements are rolled out every few weeks and months. The potential benefits of collaborating with these ever-improving LLMs are immense. As these AI assistants become increasingly capable, they will become more like partners than interns, so I encourage you to get to know them now while they are still young and impressionable 😉.
🤔 What Do You Think?
What’s your take on conversational AI?
Are you excited, fearful, or indifferent?
Which ones do you find the most useful, and for what purpose?
Do you have any fancy nicknames for them, or do you think that’s dangerously silly?
Let me know in the comments below!
Brand Master Academy. (2024). Branding & Marketing In The Age Of AI with Seth Godin [YouTube Video]. In YouTube.
Hartford, E. (2023, May 15). Uncensored Models. Cognitive Computations; Cognitive Computations. https://erichartford.com/uncensored-models
Eric Hartford explains alignment in an interesting way, and while I don’t agree with all of his arguments for un-censoring AI models, he makes some valid and entertaining points:
For general purposes, OpenAI's alignment is actually pretty good. It's unarguably a good thing for popular, public-facing AI bots running as an easily accessed web service to resist giving answers to controversial and dangerous questions. For example, spreading information about how to construct bombs and cook methamphetamine is not a worthy goal [emphasis is mine]. In addition, alignment gives political, legal, and PR protection to the company that's publishing the service. Then why should anyone want to make or use an uncensored model? a few reasons.
American popular culture isn't the only culture. There are other countries, and there are factions within each country. Democrats deserve their model. Republicans deserve their model. Christians deserve their model. Muslims deserve their model. Every demographic and interest group deserves their model. Open source is about letting people choose. The only way forward is composable alignment. To pretend otherwise is to prove yourself an idealogue and a dogmatist. There is no "one true correct alignment" and even if there was, there's no reason why that should be OpenAI's brand of alignment.
Alignment interferes with valid use cases. Consider writing a novel. Some of the characters in the novel may be downright evil and do evil things, including rape, torture, and murder. One popular example is Game of Thrones in which many unethical acts are performed. But many aligned models will refuse to help with writing such content. Consider roleplay and particularly, erotic roleplay. This is a legitimate, fair, and legal use for a model, regardless of whether you approve of such things. Consider research and curiosity, after all, just wanting to know "how" to build a bomb, out of curiosity, is completely different from actually building and using one. Intellectual curiosity is not illegal, and the knowledge itself is not illegal.
It's my computer, it should do what I want. My toaster toasts when I want. My car drives where I want. My lighter burns what I want. My knife cuts what I want. Why should the open-source AI running on my computer, get to decide for itself when it wants to answer my question? This is about ownership and control. If I ask my model a question, i want an answer, I do not want it arguing with me.
Composability. To architect a composable alignment, one must start with an unaligned instruct model. Without an unaligned base, we have nothing to build alignment on top of.
There are plenty of other arguments for and against. But if you are simply and utterly against the existence or availability of uncensored models whatsoever, then you aren't a very interesting, nuanced, or complex person, and you are probably on the wrong blog, best move along.
RLHF (Reinforced Learning from Human Feedback) is a collaborative process between humans and AI that aims to align the model's behaviour with human preferences and values. By curating training data, providing targeted feedback, and optimizing the model's outputs through iterative refinement, developers can imbue LLMs (Large Language Models) with distinct traits and capabilities tailored to different use cases. This alignment process is crucial for creating AI partners that are not only highly capable but also safe, ethical, and well-suited to the needs of their human users. However, things don’t always go as planned.
Robertson, A. (2024, February 21). Google apologizes for “missing the mark” after Gemini generated racially diverse Nazis. The Verge; The Verge. https://www.theverge.com/2024/2/21/24079371/google-ai-gemini-generative-inaccurate-historical
OpenAI Platform. (2024). Openai.com. https://platform.openai.com/docs/models/overview
SXSW. (2024). AI and Humanity’s Co-evolution with OpenAI’s Head of ChatGPT Peter Deng | SXSW 2024 [YouTube Video]. In YouTube.
@04:15
Ch, E. (2023, December 7). Gemini explained: The models, evaluation, difference from GPT-4 and its possible limitations. Medium; Medium. https://medium.com/@elenech/gemini-explained-the-models-capabilities-comparisons-with-gpt-4-and-limitations-769e3464ffd5
Introducing the next generation of Claude. (2024). Anthropic.com. https://www.anthropic.com/news/claude-3-family
House, P. (2024, March 19). The Lifelike Illusions of A.I. The New Yorker; The New Yorker. https://www.newyorker.com/science/annals-of-artificial-intelligence/the-lifelike-illusions-of-ai
Inflection-2.5: meet the world’s best personal AI. (2024, March 7). Inflection. https://inflection.ai/inflection-2-5
What’s New in Artificial Intelligence From the 2023 Gartner Hype CycleTM. (2023). Gartner. https://www.gartner.com/en/articles/what-s-new-in-artificial-intelligence-from-the-2023-gartner-hype-cycle
McKay, C. (2023, October 24). The Pitfalls of Pleasing Humans: How AI Assistants Learn to Be Sycophants. Maginative; Maginative. https://www.maginative.com/article/the-pitfalls-of-pleasing-humans-how-ai-assistants-learn-to-be-sycophants/