12: AI - Hype Cycle or Real Revolution?

Hey there,

Welcome to where I write about my journey from a stable Big Tech Software Engineering job to the wild and volatile world of Venture Capital.

Last week, I attended an event where ten promising founders pitched their companies. Remarkably, all ten were AI startups, with ideas ranging from generative PowerPoint presentations to AI for corporate construction insurance. Their valuations and traction seemed astronomical. This week, I went to a dinner where we actually voted to treat AI like Voldemort during mealtime. It seems AI is dominating the zeitgeist, but is it all just hype?

Free Daily Trade Alerts: Expert Insights at Your Fingertips

• Master the market in 5 minutes per day

• Hot stock alerts sent directly to your phone

• 150,000+ active subscribers and growing fast!

Free Access Here

The Gartner hype cycle is frequently used to define the public’s expectations for a new product or technology.

A technological breakthrough, often without a product or market, triggers significant publicity despite no usable products or commercial viability.

As the media continues to tout the potential of the technology, we reach the Peak of Inflated Expectations.

Media hype can’t fuel a companies bottom line forever, and as P&Ls plunge into the red, the public reaches a trough of disillusionment.

Only the truly viable companies survive this Darwinian culling, and scale to become staples of our society with Mainstream Adoption. For example, Amazon survived the Dot Bomb era to become the behemoth that it is today.

The expectations for AI are at a fever pitch, with firms like PwC saying there will be 15.7 trillion dollars in global economic contribution from AI by 2030. NVIDIA, which provides the GPUs that power this AI boom, constitutes > 6% of the S&P 500 as of this writing and several others of the top 10 are wholeheartedly focused on AI innovation: Apple (7%), Microsoft (6.5%), Amazon (3.5%), Meta (2.4%), and Alphabet (3.7%).

For artificial intelligence, it’s hard to say where we actually are on the cycle because there are multiple cycle’s occurring in parallel and out of step. From the public perception, it could seem like AI is constantly reaching new heights on the peak of inflated expectations, but in actuality “AI” is multiple cycles of innovation happening in sequence:

Classical Machine Learning consists of techniques like supervised learning or unsupervised learning. In supervised learning a model is trained on pre-labelled data. For example, Netflix or Amazon’s recommendation algorithms work based on what other users with similar characteristics to you prefer. Unsupervised learning is where a model is trained on data without labels, ideally finding hidden patterns or structures in the data. For example, how Netflix or Amazon use massive troves of viewing or purchasing data to extract groups of similar users.

Deep learning is built on top of the multi-layered neural networks, with use-cases ranging image recognition, natural language processing, text & image prediction (and thus text & image generation). Just like neurons in our brain, neurons in a NN fire based on different input signals.

Most of the Generative AI hype cycle is built off of Transformers (for example: ChatGPT(ransformer) and other Large Language Models). A transformer is a specialized neural network that can process multiple words (or tokens) at once to establish a context and then predict the likely next sequence of words. To make these predictions, transformer based LLMs are trained on massive amounts of data (as much of the internet that AI companies can gobble up).

The summation of these technologies, viewed as “AI” is perceived by the public like this:

Will the hype subside into a traditional trough of disillusionment? We may be at a what feels like a Peak of Inflated Expectations currently, but a descent is dependent on now subsequent waves of innovation. If we continue to see independent innovations under the umbrella of AI, public sentiment could look like this:

OpenAI did just release a new model (O1) that’s based on a different paradigm (Reinforcement Learning). Could this be our Innovation X and what will be Innovations Y & Z?

Only time will tell, but it’s an exciting (and terrifying 😨) time to be alive.

Until next time!

Signing off and signing zero checks,

SWEdonym