How to learn a lot about generative AI: Start a new company.
Recently, NfX reached out to early founders who seek to build generative AI companies; they invited them to apply for funding.
We fit NfX’s ‘Why’, so I re-engaged five, long-time, wonderful, reliable colleagues to sprint toward an application. Here’s what we learned.
What did we expect?
We expected to:
Lead: Participate in NfX’s Generative Tech FAST; describe our team, the company, a market, a business model, timing, traction, fundraising, and fit.
Self-assess: Do we really understand the nature of generative technology? How might we refine our product-market fit?
Identify: Select parts of Knowledge Harvesting, Inc. and permutate as [Generative Problem-Solving + Interconnecting App + Enterprise-delivery = Funding for a new company].
What did we experience?
We didn’t submit an application.
We learned a lot of useful things about VCs and generative AI.
NfX’s videos are really good! Fundraising Advice videos where top VCs share insights on the most important elements of fundraising pitches
Here are two articles about generative tech: Generative Tech Market Map and 5-Layer Tech Stack and Generative Tech Begins.
Here are two articles about network effects: Why Network Effects Are Mission Critical and The Network Effects Manual: 16 Different Network Effects – 2022. Here’s the full array of Articles about Network Effects.
The end of this article provides a few of our learnings.
Why was there a difference between our expectations and experiences?
We had dozens of options to evaluate and it was challenging to select a single, concise product/market fit. Also, we did not figure out how to align valuation with a verifiable, humongous market.
We learned a lot because NfX elegantly presented the essential questions and delivered effective guidance for answering them.
How might our learnings influence future actions?
Continue this exercise. Challenge assumptions and address gaps, especially those described in the Ladder of Proof.
What about generative tech? We’re in! Now is the time to figure out how to meld Knowledge Harvesting technology with generative AI and ensure that the product functions activate network effects.
NfX, thank you for your well-designed FAST process and for sharing your deep smarts. The application process was an exceptional way to close 2022.
Best regards,
LT
A few learnings
There’s a pattern for that.
Here’s an example of what we learned in regard to startup patterns.
Great Startup Idea: Novelty is a risk.
Novelty: the quality of being new, original, or unusual
Why is novelty a risk?!
Change only one element from what came before
Element: a part or aspect of something abstract, especially one that is essential or characteristic
For a complex solution, there are lots of elements. It’s time to figure out how to deliver a complex product rather than reducing an idea to a few elements, then changing one element at a time.
Be aware of what parts of your idea you’re testing that are new, and what parts of your idea you can count on working
Definitely!
A product has components. Start graphing; link components and markets.
The biggest ideas often strike a balance between what has been proven to work and something totally new.
This sounds similar to the previous attribute/criterion.
Components have/deliver functions. Add to the graph.
Pitch as “solving a problem (pain)” or “creating an opportunity (gain)”?
Pitch the idea as a gain. It’s time to lead with connection, not fear. If we have to peddle fear, then the whole idea feels wrong. After leading with connection, then it’s OK to mention a pain or two.
Add all of the messages to the graph.
Messages must sync and change with functions.
Predict: Will we have market risk or execution risk?
In the past, we’ve experienced execution risk. We need a partner to lead this.
Which market?!
We are going for All Layers All Models (ALAM): Our product should thread OS functions, APIs, and access to proprietary data.
All Layers All Models (ALAM) consists of:
Applications layer
OS or API layer
Hyperlocal AI models
Specific AI models
General AI models
NfX: About the Applications Layer:
These applications are interfaces where humans and machines collaborate. These are the workflow tools that make the AI models accessible in a way that enables business customers or consumer entertainment.
In this layer, it’s easy to envision network effects or embedding defensibilities.
There will be 10,000’s of these applications built for various needs in the next 2 years. Incumbent software providers will add generative features. New companies will create competitors to the old, emphasizing generative as a wedge. New companies will create brand new applications people will use with generative AI as a starting point.
How do we dial it down to a product-market fit (PMF) that the VC will judge as reasonable, lucrative, and generalizable? Maybe, we go from specific to general.
Hyperlocal: How to do this for a culture/geography/sector (HMI/HITL)?
API: For capability or process X, which API connections and interfaces correspond with which sub-processes?
Base = AI models: How do we blast it with graphs? All it takes is graph theory 101 and instantiation of the lowest common denominator, soon to arrive from the upcoming ISO standard. How does an AI model cohabitate with knowledge graphs? It must be neruo-symbolic paradigmed. The hidden growth factor and an ocean-sized market are due to our product’s universality which is operationalized with the “base.”
What’s changing?
NfX: What has changed in technology, platforms, customer behavior, laws, etc so that what you are doing is newly possible?
It was helpful to specify the changes, then answer: “What hasn’t changed?!” For Knowledge Harvesting, Inc., here’s what has not changed:
The invisibility of expertise
The value of expertise
The imperative for expertise
The predictable ebb and flow of divergent and convergent thinking in all forms of problem-solving and decision-making
Traction metrics
NfX says: If you don’t see the metric that best shows your traction, please write it in the open text field at the bottom.
Here’s what we measure and monitor:
Types (varieties) and amounts of information captured = InQ = Information Quotient
Knowledge Harvesting and Interconnecting process improvements/year
Number of new domains/disciplines
Number of problem-solving methods generated and implemented
Total number of codified cognitive tasks
Number of client projects
Longevity of client relationships
Recurring revenue
Number of trained Knowledge Harvesting professionals
Total number of organizations to institutionalize Knowledge Harvesting
Demand for Knowledge Harvesting-like services
Supply of knowledge capture services
Volume of messages about brain drain risks
Volume of messages about the need for “interconnecting”
One minute
The NfX application required a one-minute video. Here’s our first draft script.
Now is the time to integrate knowledge harvesting with generative AI.
Yes, the black box seems really good at statistics and endlessly producing digital items, but the current generative AI investments don’t holistically account for the full essence of problem-solving and decision-making. The complete cycle weaves divergent and convergent cognition.
Please consider: The most valuable resource on the planet is people with expertise. However, it’s challenging to make expertise explicit.
Our company knows how to make (human) tacit knowledge explicit. We do this by providing experts with 1) a tangible, living copy of their mental model and 2) actionable tools for amplifying their creative and critical thinking.
Keywords: #knowledgeharvesting, #generativetech, #generativeai, #nfx, #ai