Artificial Intelligence > Top-down vs Bottom-up Approaches to Strong AI

Top-down vs Bottom-up Approaches to Strong AI

Published: August 15, 2008

What are the major differences between top-down and bottom-up approaches to AI? The top-down approach is hinged on the belief that logic can be inferred from an existing intelligent system. The bottom-up approach, on the other hand, is concerned with creating basic elements and allowing a system to evolve to best suit its environment.

Both approaches already have applications in the real world. The top-down approach has shown amazing results in the form of semantic understanding. One common application is replicating the logic behind language to create speech and language interpretation systems. The bottom-up approach has produced genetic algorithms which have been created to help find optimal solutions to complex problems.

When looking at the future of Strong AI, top-down approaches are showing the most results by far. There are however, a few bottom-up approaches such as the creation of child machines with a willingness to learn that are worth noting.

A bottom-up approch would be the most ideal way to create human-like intelligence as we ourselves are part of a bottom-up design process (which occured in the form of evolution). Unfortunately the world is not going to wait for something to grow out the electronic ooze and instead, the top-down approach will slowly show signs of true intelligence as all things of understanding will be put into boxes of logic for digital machines to understand. In all likely-hood, Strong AI will be a combination of the two approaches with an amalgamation of different systems to produce a Frankenstein-like effect. Humankind itself is likely to go down this road in years to come with the prospect of becoming super-intelligent cyborgs able to conjure up anything you wanted to know. The direct brain interface is not quite there yet, but with the way people are clinging to their iPhones, one could be easly mistaken in thinking that this has already happened.


1. tkorrovi on February 16, 2009

The training program you mentioned in AI Forum may be whatever kind of, there are indeed training programs which directly control the system, and such are sure not proper for True AI. But the training program used for unrestricted systems are different, all they do is interracting with the system instead of human or environment, they have no further function for the system. Thus all they do is providing the system a minimal environment. A big environment like internet is good, but a system with even the highest potential must start from simple environment and be gradually be trained in the more complicted environments. Training program is necessary in any such system for the sole reason that it is too hard work to play the environment for the system all the time manually. Thank you for reading this.

2. Levi on May 29, 2013

Considering the evolution of bottom-up AI, the development of the AI that is greatly affected by external stimuli from the environment. One could readily infer that this approach is akin to children's development. However, it is unclear if AI would develop vivid imagination that is comparable to a human's. Of course, that won't be necessary when looking from a industrial perspective and not the aim of recreating a human personality out of inorganic materials.

Any Comments?


» Pied-a-Terre on Cambie Street

The Forever Web App Project

Published: November 15, 2009

The Forever Web App Project is an AI project to demonstrate a web app's ability to exist on the web unassisted (except by strangers) for as long as possible after a given date.

Sudoku & Artificial Intelligence

Published: March 16, 2009

Using AI on Sudoku could be considered overkill. Either way, any algorithm written to solve a Sudoku puzzle could be considered intelligent by understanding what it accomplishes.