What makes for weak or strong AI? Weak AI is mimicking intelligence, while strong AI is actual intelligence.
- The philosopher John Searle pointed out that you can think of artificial intelligence in two different ways. There's strong AI, and weak AI. Strong AI is when you have a machine that displays all the behavior that you'd expect from a full fledged artificial person. This is usually what you see in science fiction movies. These are artificial beings that have emotions, a sense of purpose and can even tell jokes. Some computer scientists also refer to strong AI as general artificial intelligence.
General AI is a broad intelligence that doesn't just apply to one narrow task. Think of it this way, a strong AI machine might learn a new language just for the joy of learning a new language. On the flip side, there's weak AI, often called narrow AI. This is AI that is confined to a very narrow task. It might be like trying to process your natural language into text. Maybe it's trying to sort all the pictures on your computer.
A weak AI program is not trying to chat with you about the world. It's just trying to assist you with a very narrow task. Most AI experts believe that we're just starting down the path of weak AI. That means that there are many organizations already out there that are using weak AI to help with these narrow tasks. Thing about some of the newest AI assistants, like Apple's Siri. Each one is a good example of weak AI.
You can talk to them and even ask them questions. Then they'll convert your language into something that a computer will recognize. They'll also do pattern matching to connect your request to what's in their database, but if you think about it, that's not that much different from when you type something into Google or Bing. The difference is that it seems to behave much more like a human. These artificial personal assistants will search the web, your smartphone, and even your contacts and calendar. They might even give you the same answer that you'd get from a human assistant.
You can ask them to schedule your lunch, or even call your friend. It sticks to these narrow tasks and follows your voice commands. These personal assistants don't have general artificial intelligence. If they did, they'd certainly get tired of listening to your requests all day. Instead, they focus on a narrow task of listening to your input, and matching it to their database. Most of the energy in AI is around developing and expanding narrow AI.
As of right now strong or general AI is still just the subject of science fiction. John Searle was also quick to point out that any software that uses the symbolic systems approach, should be considered weak AI. Remember that the symbolic systems approach matches symbols in a way that seems intelligent. Think of the person in the Chinese room. They don't understand the language but can match two different patterns of Chinese characters.
In the 1970's and 80's, symbolic systems was used to create artificial intelligence software that could make expert decisions. These were commonly called expert systems. In an expert system, you have experts create a list of different steps that you could take to make some conclusion. If the long enough then it starts to feel like AI. These expert systems were sometimes used in the medical field. A nurse might input a bunch of symptoms into a computer.
If the patient has a cough, then checked if they have a temperature. If they have a cough and a temperature, then check if they're dehydrated. If they have a cough, a temperature, and are dehydrated, then it asks the nurse to check for Bronchitis. To a patient, it might look like they're being diagnosed by an intelligent computer. In reality the program is just going through a long list, just like they would in any medical textbook. It's matching the symbols and patterns that the expert created, to reach a possible diagnosis.
In the end, expert systems run into the same problems as any other symbolic system approach. They would lead to combinatorial explosions. There were simply too many symbols and too many matching patterns. There's too much information serving as input. Think about all the different questions a good doctor might ask to reach an accurate diagnosis. Yet the symbolic system was a key starting point for artificial intelligence. It's still used to this day and in fact many experts still refer to it as GOFAI, or good old fashioned AI.
This course will introduce you to some of the key concepts behind artificial intelligence, including the differences between "strong" and "weak" AI. You'll see how AI has created questions around what it means to be intelligent and how much trust we should put in machines. Instructor Doug Rose explains the different approaches to AI, including machine learning and deep learning, and the practical uses for new AI-enhanced technologies. Plus, learn how to integrate AI with other technology, such as big data, and avoid some common pitfalls associated with programming AI.
- The history of AI
- Machine learning
- Technical approaches to AI
- AI in robotics
- Integrating AI with big data
- Avoiding pitfalls