Why Reinforcement Learning became uncool and how it might come back.

A brief overview of the problems associated with RL

3 min readOct 18, 2023

Reinforcement Learning was one of the most hyped areas in AI. At one point, it was supposed to revolutionize the world.

Now it’s almost an after-thought to Supervised and Unsupervised Learning. AI Legend Yann LeCun had this to say about RL-

So what went wrong?

Reinforcement learning (RL) is a type of machine learning that enables an agent to learn how to behave in an environment by trial and error. The agent is rewarded for taking actions that lead to desired outcomes, and penalized for taking actions that lead to undesired outcomes. Over time, the agent learns to choose actions that maximize its expected reward.

RL is useful in situations where it is difficult or impossible to provide the agent with explicit instruction on how to behave. However it comes with three major flaws that held it back significantly-

Costs- No way around it, RL can be very expensive. High costs of development → Higher Barrier to entry → Lower Diversity of R&D. This creates a vicious loop, where most prominent RL research only comes from high-budget labs, restricting the discussion further.
Information Overload- The real world is infinitely more complicated than the environments RL agents are trained on. This leads to all kinds of complication and information overload for the RL agents. This is why we see fancy self-driving cars completely breakdown IRL.
Complexity- Both Supervised and Unsupervised Learning have conceptually simple use-cases: anyone can understand where those techniques can be helpful. Try coming up with something similar for RL. Most businesses aren’t interested in a go-playing bot.

Despite this, I believe that RL has great potential for testing applications in modern tech stacks. Since modern Tech Stacks rely on chains of API calls, letting an RL agent test the stability of the system can be a great augmentation.

A quick introduction to Reinforcement Learning [Math Mondays]

Some observations on one of Machine Learning's controversial fields.

codinginterviewsmadesimple.substack.com

What do you think? Is RL dead, or will it make a comeback? I’d love to hear your thoughts.

For more details, sign up for my free AI Newsletter, AI Made Simple. AI Made Simple- https://artificialintelligencemadesimple.substack.com/

If you liked this article and wish to share it, please refer to the following guidelines.

If you find my writing useful and would like to support my writing- please consider becoming a premium member of my cult by subscribing below. Subscribing gives you access to a lot more content and enables me to continue writing. This will cost you 400 INR (5 USD) monthly or 4000 INR (50 USD) per year and comes with a 60-day, complete refund policy. Understand the newest developments and develop your understanding of the most important ideas, all for the price of a cup of coffee.

Support AI Made Simple