Saying AI three times will not get you funded
We have seen a lot of AI pitches lately, particularly from non-technical founders.
Artificial Intelligence (AI), particularly deep learning popularized by Google, has done and will continue to do amazing things. Still, there are certain constraints and requirements of d eep learning that make it very difficult for early-stage startups to take advantage of. Let us take a brief look at what deep learning is, and then we will discuss those constraints.
Deep learning is the basis for technologies that are fundamentally reshaping the ways we live, work, love, and war. Current applications of deep learning include machine vision, media curation, facial recognition, medical diagnosis, dating sites, self-driving cars, and target acquisition for robot war machines.
Of course, this being the technology industry, we have hyped deep learning as a solution for everything while ignoring the complexities and limitations of this complex and essential technology. Before you base your pitch on your AI models changing the world, make sure to consider the following:
Data is Queen
Deep learning is a subset of machine learning where neural networks — algorithms inspired by the human brain — learn from large amounts of data. algorithms perform a task repeatedly and gradually improve the outcome, thanks to deep layers that enable progressive learning. It’s part of a broader family of machine learning methods based on neural networks.
Deep learning models are data-driven unless your startup has a large unique dataset; your company does not have a sustainable AI advantage! Given the same data, a competitor can create a competitive deep learning model if they can afford the hardware to do it.
But if you have a unique dataset like our portfolio company Boost Sports , we can talk. Still, even with their unique dataset, if it were not for Boost’s CEO and CTO proven abilities to recruit high-level ML talent, it is unlikely that we would have invested, which leads to our next warning.
Real Deep Learning experts are expensive
With Google, Facebook, and other F100s fighting for still rare talent, you will be competing for talent in a profession where salaries of 350,000 dollars or more are not uncommon, and predatory hiring practices are standard. I know of at least two AI startups that failed because a FANG company hired their key AI engineers. AI development is becoming more accessible, but I think it will be a couple of years before it easy enough to see the prices drop for AI developers.
Fake AI makes founders look naïve
It is not uncommon for us to meet with a founding team and have a CEO proclaim that they are an AI company, and when we dig into the technology, there is no AI, and usually, no one on staff that would be capable of creating an AI model. Behavior like this will not raise your valuation; it will just keep you from getting funded.
You do not need to be an AI company to take advantage of AI
If your startup operations generate a unique dataset, you can potentially sell that data to AI companies. You can use publicly available models to improve your products with minimal effort and a much lower level of expertise.
If you are a non-technical founder and you are basing your business on developing an AI model, you will have to fund high capital needs due to the costs of deep learning talent and the costs of data acquisition. If you do not already have ownership of a unique dataset, it will not be easy to get funded. Instead, focus on using available models to enhance your core business and look for ways to monetize the data your business generates.
Originally published at https://www.impactseat.com on September 21, 2020.