Artificial intelligence (AI) is a fixture of modern conversations. People are fascinated by what it does and what it could do. Many of the individuals who discuss it often specify weak vs. strong AI.
Once people know the difference between strong AI and weak AI, they’ll be well-equipped to learn from and contribute to the current and future conversations about the types of AI.
The Definitions Relate to Capabilities, Not Strength
The first thing to keep in mind is that these descriptors do not tell people how useful a particular kind of AI is in today’s society. They also don’t describe the quality of the programming or suggest it’s error-free.
Rather, when people use terms like “strong” or “weak” to describe AI, they’re elaborating on the extent of its capabilities.
How Does Weak AI Work?
The responses from weak AI all relate to information it received during programming.
For example, virtual assistants like Siri and Alexa are weak AI. They know to detect information that’s similar to what they already know, classify it and respond accordingly. If a person says, “Alexa, play some rock music,” the AI should choose some relevant tunes.
However, it is only reacting in a way consistent with its programming. Weak AI is also called narrow AI. It got that name because the AI only takes information from a dataset related to the requested task.
In other words, weak AI cannot think for itself and venture outside the relatively limited box containing its programmed information.
Weak AI performs impressively when asked to carry out its programmed tasks. When people try to get it to react beyond that, however, it falters. People who use smart assistants frequently and give them unfamiliar prompts that are not part of their programming receive responses such as “I’m sorry, I didn’t understand that.”
What About Strong AI?
When people discuss strong AI, they may also say “general AI” as a synonym for strong AI. This more advanced AI works more like the human brain. Instead of only pulling responses from what exists in its programming, strong AI functions through association and clustering.
Like having a conversation with a real human, people using strong AI cannot necessarily predict the results. If a person asks their friend, “Do you want to go to the park with me later?” they won’t know the definite answer before posing the question. In contrast, they know how weak AI will behave due to its programming.
Going back to the smart speaker example, a person might say to a virtual assistant, “Please turn on the kitchen lights.” They can then know that it’ll brighten up the kitchen, provided the technology understood the command.
Strong AI can think for itself, even outside its programming. It might learn, without prior human input, that when a person asks to turn the kitchen lights on, they’re likely getting ready to eat. Then, a smart speaker might also ask if they want music to accompany the meal.
This level of functionality is not the same as programming a smart speaker to go through certain routines at particular times of the day. Most speakers can do that now, despite their weak AI.
The difference between strong AI and weak AI is that the strong kind can demonstrate functions even if it never learns them from programming or humans first.
Strong AI Doesn’t Exist Yet
Tech experts agree people haven’t achieved genuine strong AI yet, but they’re getting closer.
For example, several years ago, scientists trained AI to master Atari video games. The technology didn’t know the rules of the games beforehand, but it learned the strategies needed to excel at them.
A survey of AI experts indicated a prevailing belief that general AI may not get developed in most people’s lifetimes. The average estimate of the guesses of when it would become a reality was more than 80 years away.
Unfortunately, that means people can only speculate about what strong AI will be like when it arrives. For many, the closest thing they can imagine now are the robots of science-fiction movies that seem to initially obey their programming, but eventually show they have minds of their own.
Analysts say general AI needs some crucial qualities to succeed. They include short and long-term memory, plus abstract reasoning.
Accurately Categorizing AI
Knowing the difference between strong AI and weak AI is essential for thinking about what the possibilities hold. The overview here helps tech enthusiasts understand the fundamental definitions for the two.
Now, they can better understand what people mean when they talk about what AI can do now and what it may achieve within the next century.