Dear Reader,

An interesting AI model caught my eye recently. It offers a boost in AI utility.

Known simply as “Gorilla,” it’s the result of collaboration between researchers at Microsoft and the University of California, Berkeley.

Some of the most popular AI models are being criticized for a lack of real-time utility.

For example, ChatGPT has a data cut-off point from September 2021. If you are a paying “Plus” user, you can get some real-time data through plug-ins for $20 per month.

In ChatGPT’s case, the developers support a limited number of plug-ins. Right now, that includes plug-ins like Expedia, OpenTable, and Spotify.

If you’re a Plus user, you can ask ChatGPT to help you find your next vacation destination… reserve a table at a nice restaurant… and create a playlist to match your trip.

That’s a step in the right direction…

But it’s limited to the plug-ins that OpenAI allows. Obviously, its developers have focused on getting the most popular plug-ins first. But that means that ChatGPT – while useful – is not an “everything AI” just yet.

Gorilla, however, could change that.

Gorilla Works Differently

Gorilla doesn’t need a developer to hardcode access to a plug-in. It can automatically find the right API for the job and use it.

API stands for “application programming interface.” It sounds complicated, but it’s not. APIs are used to connect apps together. It’s how the AI tells Expedia you want to book a flight to Spain.

Gorilla creates a seamless AI experience by giving users access to hundreds of apps and websites. In the initial research paper, developers gave it access to 1,645 APIs.

What’s so handy about Gorilla is that you don’t need to tell it which API to use. It’s designed to figure that out on its own by using its AI to crawl through API providers for the right tool.

That’s a huge advantage. With something like ChatGPT, an app maker has to request OpenAI to greenlight a plug-in. Then, both sides need to write code to make it work. And finally, an end user might be able to make use of it.

For as fast as AI has been developing, this is a surprisingly slow, manual process.

With Gorilla, it handles the whole affair automatically.

Here’s a chart that compares Gorilla to other AI models like GPT-4 and Claude.


Zero-shot retrieval means that the AIs were not trained on which API to select for a given task. The AI had to select the right API only using the natural language prompt given to it. And hallucinations refer to “made-up” answers.

Gorilla had about twice the accuracy as the next best AI model… with about a third as many hallucinations.

The researchers weren’t shy about stating the implications of Gorilla’s impressive results.

They stated that an AI like Gorilla “could transform large language models (LLMs) into the primary interface to computing infrastructure and the web.”

The Real Motivation

But as great as Gorilla is, I don’t expect it to ever become a household name.

Instead, I think developers at the likes of Google, Apple, and Microsoft are going to learn from the success of the Gorilla model and build those capabilities into their own models.

On Alphabet’s most recent earnings call, CEO Sundar Pichai said that the next generation of Google search, search generative experience (SGE), is going to make searching even more natural and intuitive.

That means you’ll be able to have a conversation with an AI to find what you’re looking for instead of typing in a string of keywords.

Along with that comes improved targeted ads.

Advertisers bid up the cost of ad placements to get their ads at the top of a user’s search results.

With AI and SGE, Google will be able to serve up a smaller but more curated list of options for buyers.

That’s going to reduce wasted ad spending that doesn’t drive results. But it will also drive up the cost for ads with a higher probability of success.

I completely understand why these companies – especially Google – would be interested in using this technology to drive more targeted ads. But my hope is that this won’t be the only use case for Gorilla.

Generative AI and LLMs are arguably the most important technological breakthrough of the last decade. They have the potential to make our lives easier, healthier, and more efficient. And yet, the first instinct among many executives is to use them to deliver more advertisements…

This instinct is understandable, but it lacks imagination.

I’m more interested in the companies using this new technology to create products and services that completely redefine their target markets. Or, I’m looking for AI companies that are creating new markets from scratch.

That’s where the truly disruptive potential will be. And that’s where we will focus our investment research.


Colin Tedards
Editor, The Bleeding Edge