Language Models in Solopreneurship

For those willing to put in the work, understand the nuances, and approach AI with both caution and curiosity, the rewards can be huge—for business growth and career growth. This is an in-depth examination of AI language models, at length and with historical and contemporary context, informed by my experiences interacting with them nearly every day for the past year.

The aim of this piece isn't to persuade you to use AI tools. Your choices are yours alone. I'm documenting the value I've gotten from LLMs and offering insights into the mindset that, I believe, is beneficial—especially if you're business-oriented or career-driven. This is a glimpse into my thought process. I can't predict the future of this domain, but the tangible benefits I'm reaping are enough for me to write this post.

I work in the SaaS, B2B, and AI spaces, and also do freelance tech journalism on the side, so it's no surprise I'm deeply invested in this. I love new technology. I thrive on these discussions. You might not share my intensity, but there's no harm in being curious and exploratory. Because the goal is scale and revenue, and these tools can offer that.

For organizations, the low hanging fruit is insane

The number of poorly crafted emails that flood our inboxes is surprisingly high. Given the sheer volume of communication mistakes, there's a straightforward solution staring us right in the face: using AI for clearer emails. 

It's the simple things that often have the most profound impact. Just a touch of careful prompting and minor enhancements can drastically reduce miscommunication.

ChatGPT is designed precisely for this. It excels in producing clear, logical, and straightforward content—ideal for the B2B environment. Be it massive corporations aiming to clarify their bulk emails, reports, and training materials or small businesses keen to polish their client emails and pitches, ChatGPT has the potential to improve business communication. Imagine a world where even non-writers churn out clear writing.

I get questions about effective prompts, and a straightforward one that rarely misses the mark is:

PROMPT: 

Clean up this text for clarity and maintain my writing style.

[insert something you already wrote that’s already unique and differentiated and sources through primary sources / interviews / unique data you have access to or skills in analyzing / a rich experience you’ve had that only you can tell that story for.

Its strength lies in providing ample context with a concise directive. However, this shouldn't be misconstrued as an invitation to sidestep the heavy lifting and just toss tasks at the AI. You need to be well-prepared: furnish it with thorough data, establish the context, and line up the right inputs. Essentially, you're laying the groundwork for the AI's performance, not tasking it to set up the entire production. Instead of overwhelming these models with myriad tasks and scanty details, it's more fruitful to supply them with substantial material and a single, clear instruction.

Think of it as handing over a sizable block of clay with a direct sculpting directive, rather than a meager amount with an exhaustive list of demands. Text generators work by efficiently connecting Point A to B. Start with your foundational material and bring in the AI as a final touch for clarity. And a crucial reminder: don't delegate your cognitive responsibilities to a machine; its efficiency reflects the quality of your input. This should go without saying, but yes you have to put in the hard work.

‍Language models are not designed to recall perfectly-true facts, they’re designed to “think”

When you input a prompt into ChatGPT, it's not just searching a database or referencing a dictionary. We already have tools for that. 

To put this as simply as I can without dumbing it down too much, essentially the model analyzes a multi-dimensional space of linguistic embeddings—a 12,000-dimensional plane—to be precise. Every word, phrase, and sentence it has encountered during its training has a unique location in this space. By comparing and contrasting the distances and angles between these embeddings, ChatGPT identifies patterns and relationships, enabling it to craft coherent responses.

This is important: ChatGPT doesn’t necessarily "remember" specifics from its training. It doesn't have a repository of books, articles, or sources it can perfectly fetch. Instead, it recognizes patterns and associations based on the vast amount of words and context it has been trained on. So instead of getting mad at LLMs for not being infallible sources of truth, don’t even use them that way to begin with. Use them for tasks where you already have your primary data.

ChatGPT is backed by supercomputers from Microsoft. While its capabilities are impressive, it's essential to understand its limitations. ChatGPT excels at logical reasoning and pattern identification but falls short on the emotional spectrum. It’s a machine with no lived experiences.

It can simulate emotional language based on patterns it has observed, but it doesn’t genuinely "feel". It operates with mathematical precision, devoid of any heartbeat. Thus, while ChatGPT can churn out articulate sentences, expecting it to create deeply emotional literature would be much like asking a calculator to compose a love ballad. Don’t expect it to do that.

Use case: I upload Otter transcripts to ChatGPT with a very simple prompt

If you're using NLP for customer insights, you have valuable data. Feed this to AI models, and they'll format your insights (not give them to you). But only a human can identify emotional triggers that connect. Why? Because AI operates on data, not emotions. And that’s a good thing because that’s different from us. It gives us a new angle to explore and a new way to uncover insights – it’s a double whammy. I use Otter to transcribe interviews (there’s probably better ones out there, but Otter works well for me for now).

I then copy the whole Otter transcript from an hour or so long conversation and prompt ChatGPT:

PROMPT: 

Take the following interview transcript and give me the fully fleshed out bullet points of every important detail that was discussed. Don’t omit a single detail. It’s critical that you capture every single point that was discussed. 

If you’re a good interviewer, the model’s output will be good. If you’re a bad interviewer, the insights will be bad. And that has nothing to do with the model.

Being able to focus purely on ideation and communication, without getting bogged down by the intricacies of information processing, has been a game-changer for me. 

Use case: I tell my ideas to the ChatGPT iPhone app

One practical application is the use of AI to translate spoken words into text. Often, when I'm out and about, ideas flow. 

Instead of stopping whatever I’m doing to manually jot down fleeting thoughts, By simply activating the Whisper audio feature in ChatGPT, I can voice my thoughts, no matter how fragmented. A simple command such as, "Rewrite these thoughts clearly," transforms my ramblings into structured text right there on my ChatGPT thread. Alternatively, I have the option to request a verbatim transcription.

For creatives – writers, artists, videographers, etc. – the best ideas aren’t confined to the four walls of your office.

I can't count the number of times ideas have emerged during my workouts, whether I’m cycling, climbing or just walking. Having my phone within arm's reach ensures that I never miss out on capturing brainwaves. I’m in the business of ideas, so this matters a lot.

Use case: Save chat interaction threads as training material

For larger teams, it's beneficial to save ChatGPT threads as PDFs. This way, you retain control and can use them as training material. They can guide new hires on using LLMs for coding, analyzing data, or writing with AI context. Knowledgeable team members should spread their expertise internally. Sharing these threads is simple, but often overlooked.

‍If you do the hard work and apply language models intentionally, these tools can actually change your business and life

I approach these models as a solopreneur. My business is my future, my family's support, and my life's driving force.

While these tools will amplify tech giants like Google, Meta, Microsoft, and Amazon, they also hold promise for solo entrepreneurs, compact agencies, and independent professionals.

Even with their challenges, I'm embracing LLMs now. They may be overhyped, but they represent a significant business opportunity for those willing to experiment. Consider the long game—imagine the potential regret of missing out on being an early adopter of LLMs.

I hate to be that guy, but it’s true. These models are powerful and the truth is you’re missing out if you don’t even give them a chance.

We should critique AI models, understand their current limitations and debate their future potential. I consistently write their drawbacks. But recognizing their potential is equally important. These tools can be transformative when used correctly. So, rather than adopting prevailing opinions, dive in, be curious, and discover their power for yourself.

Consider the early days of the internet or the advent of mobile apps. Back then, it was a wide-open space—much like LLMs are now in the 2023-2025 period. Most people either didn't understand the potential or chose to be passive observers, waiting to see where the winds would blow. However, the few who actively leaned in, even without a clear roadmap, benefited immensely.

For a business-minded person, now's the time to get hands-on with LLMs. Whether you're in growth marketing, content strategy, data analytics, or even if you're just entrepreneurially inclined, the potential applications are vast. It might be about scaling content production, tailoring customer interactions, or even something as simple as sending more clearly-worded sales emails to potential clients, deriving insights from data, or some innovative use case that hasn't been popularized yet.

So, while the crowd speculates, hesitates, or makes assumptions, the proactive approach is to delve in, experiment, and integrate LLMs into one's arsenal. It's like investing in a bear market. It's not about predicting the future but about positioning yourself well for whatever comes next.

It doesn't mean you need to pigeonhole yourself as an "LLM consultant" or a "prompt engineer," but this is the moment to dive in. Embrace them, experiment, and integrate these tools into your business toolkit. With LLMs, you're not battling against entrenched traditions or established experts; you're pioneering alongside them. This is a chance to carve out your niche and potentially gain a significant edge in your field.

However, it's essential to recognize the rapidly evolving nature of these AI tools. By the time this post goes live, language models like ChatGPT may already have seen enhancements, introducing new features and functionalities, rendering some of my shared insights or screenshots dated. 

Yet, while the specifics may evolve, the foundational principles I've shared should guide your interactions with these models. Embrace the underlying concepts, and you'll be better positioned to harness the full potential of language models, no matter how they advance.