A Neural Network Told Me to Do it: Why Behavior Analysis Represents a Gold Mine for AI Startups

The artificial intelligence(AI) space is evolving at an incredibly fast pace and that rapid growth has brought up all sorts of fears and speculations. recently, at a meeting of the National Governor’s Association, Tony Stark…I mean Elon Musk :) mentioned that AI is “ the greatest risk we face as a civilization” and urged the government to be proactive and forceful in regulation.

Needles to say that Musk’s remarks sparked quite a bit of controversy within the AI community. By enlarge, mainstream media and the general population tend to side with apocalyptic versions of AI while most AI experts believe Musk’s concerns agree overly exaggerated. Obviously, the Tesla-SpaceX CEO has a long history proving that he can think way ahead of everybody else but the reaction to his comments should not come as a surprise. While versions of AI taking over the world and building a modern Skynet are good for TV ratings, book deals and speaking engagements there are major challenges to overcome in order to get to a point in which we can even consider regulation. However, assuming that there is value in Musk’s call for the government to be proactive, the bigger question remains how can we regulate something that we don’t understand?

This long introduction brings me to today’s point; monitoring, analyzing and understanding the knowledge and behavior of AI agents represents one of the biggest opportunities in the AI market. Until know, we have made a tremendous amount of progress producing frameworks and platforms for the implementation of AI applications. Robust runtime for executing and scaling AI agents are starting to catch up. However, the tools for monitoring and understanding the behavior of AI systems are, by enlarge, missing from the ecosystem.

Behavioral analysis in AI agents is far from being an easy endeavor. To begin with, we are just barely starting to understand how humans make decisions. In cognitive psychology, the work started in the 1970s by scientists like Amos Tversky and Nobel laureate Daniel Kanehman sparked new field such as prospect theory or behavioral economics. However, the correlation of those theories with neuroscience patterns (aka how does the brain makes decisions) is still an active subject of research. In terms of AI, the market desperately needs the equivalent of behavioral economics for AI agents. Let’s put on our marketing hats and call this discipline Machine Intelligence Behavioral Analysis(MIBA). Is not very catchy but it will have to do for the purpose of this post ;)

Why MIBA is a Golden Ticket for AI Startups?

MIBA is one of those areas that has remained mostly untapped in the AI market. Companies such as OpenAI or DeepMind have started efforts in this space and Google recently announced some interesting technologies as part of its PAIR (people + AI research) initiative. However, all these efforts around MIBA are still in very early stages. Below I’ve listed a few reasons why I believe that MIBA represents a massive opportunity for AI startups and venture capitalists:

1 — Capitalizing on the AI Market Fragmentation: MIBA tools can provide a consistent model to visualize and understand AI models across the large number of AI frameworks that have been invading the market. Think about building the AI version of platforms such as NewRelic or AppDynamics.

2 — A Never Say No Market: There is a famous saying in enterprise software that organizations often have a hard time saying no to security and analytic tools. After all, nobody ever gets fired for investing in more security and better analytics. MIBA tools also fit that profile. If you are investing in AI solutions, why wouldn’t you have a strong toolset to understand the behavior of your AI systems.

3 — Bridge People-Machine Interfaces: MIBA tools are essential to improve the communication between human experts and AI agents. If we are able to understand the behavior of AI agents, we can train them better and improve their knowledge and decision making models.

4 — There are no Incumbents: Differently from other areas in the AI market, there are no incumbent platforms that dominate the MIBA space.

5 — Setup the Foundation for Smart Regulation: We are back full circle to Elon Musk’s remarks. MIBA tools will help us to better understand the behavior( good or bad) of AI agents and pave the way for smart government involvement and regulation.

I will have to leave this discussion as a purely theoretical exercise. In the next few days, I will try to deep dive into some of the capabilities that I believe should be included in the first generation of MIBA tools

Written by

CEO of IntoTheBlock, Chief Scientist at Invector Labs, Guest lecturer at Columbia University, Angel Investor, Author, Speaker.

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store