Scientists still don't know the answer to this infamous question - Charles Wallace & Dan Kwartler

What is discussed, and when

1. 0:00 – 3:00: John Searle
2. 3:00 – 6:00: Artificial Consciousness Test

Notable quotes from "Scientists still don't know the answer to this infamous question - Charles Wallace & Dan Kwartler"

• Philosopher John Searle developed the original version of this premise in 1980, as a response to some of the AI work being done at the time.

  about John Searle · at 1:00

• For example, one lab’s artificial consciousness test probes AIs that have no data about consciousness for information they could only acquire from being conscious.

  about Artificial Consciousness Test · at 5:05

Who and what is mentioned in this episode?

People

• John Searle — Person · 1 mention · first at 1:00

Concepts & topics

• Artificial Consciousness Test — Concept · 1 mention · first at 5:05

Full transcript with timestamps

0:06 After waking up alone in a locked room, two documents are slipped under your door: a note in an alien language and a detailed instruction manual in your language. The manual explains that for each alien character in the note, you should write an indicated corresponding symbol. Following this chart, you write a response that you slip out the door. And for the next several days, this exchange continues. Outside the room, alien scientists are thrilled because they believe you’re conversing with them. But you still have no idea what these characters mean.

0:50 This scenario isn’t just a bizarre misunderstanding— it’s a valuable thought experiment for understanding artificial intelligence. Philosopher John Searle developed the original version of this premise in 1980, as a response to some of the AI work being done at the time. But while modern AI models don't work like those outdated machines or the prisoner in Searle’s hypothetical, the question motivating his thought experiment is still relevant. To quote Searle, he wanted to interrogate whether an “appropriately programmed computer literally has cognitive states.” In other words, if a computer looks like it understands something, does that mean it actually understands the way a human does?

1:44 Searle’s question falls into a long tradition of exploring whether or not AI could have a mind like ours. But answering these inquiries is incredibly difficult because, as philosophers and cognitive scientists will tell you, we still don’t know how our minds work. Even our fundamental definitions are slippery! Theorists generally agree that concepts like understanding, sentience, and consciousness are all different, but also that they’re related, and we don’t know how. Worse still, our usual scientific tools struggle to help us understand these experiences.

2:26 Consider drinking a cup of coffee. Scientists can observe the physical process of ingesting the coffee, and we can measure the chemical impacts of caffeine on your body. These things are objective realities. But your collective sensation of smelling, sipping, evaluating, and experiencing a morning routine is more than the sum of its parts. This is consciousness— your subjective experience of being alive. And despite major leaps in psychology, cognitive science, and neurology, researchers still don’t know how various firing neurons bring about this experience.

3:12 So if we can't define consciousness and understanding or identify what's uniquely human about them, how can we possibly test for these states in computers? Assessments like the Turing Test propose that if a human can't tell they're conversing with a computer, that computer could be seen as having some internal cognition. But this scenario is exactly what Searle was criticizing! This computer might just have the appearance of understanding. And just like cognitive scientists struggling to map consciousness onto brain activity, today’s AI researchers know how they trained their creations, but not how AIs reach their exact conclusions.

3:59 There are some ways in which modern machine learning models are less mysterious than their predecessors. Approaches like neural networks and deep learning are designed to mimic known elements of human cognition. Like us, these models excel at pattern recognition— they learn by becoming familiar with information and forming connections across data sets. This kind of processing arguably approaches Searle’s definition of understanding— but it also reveals a bias in his original question. Since humans learn through pattern recognition and we believe ourselves to be conscious, we might also be predisposed to think other beings who learn the same way are somehow closer to consciousness as well.

4:50 To combat this bias, some theorists have developed a different metric; specifically, that a fully conscious AI could draw connections beyond the information in its data set. For example, one lab’s “artificial consciousness test” probes AIs that have no data about consciousness for information they could only acquire from being conscious. This might involve asking an AI if it understands dreaming, or can report having had dreams itself. Can it understand a story about body swapping, where consciousnesses are shuffled between physical forms?

5:34 It’s unclear when or if an AI will be able to understand us the way we understand each other. But whatever happens, it’s up to us already conscious creatures to chart the path forward.

Questions about this episode

What is discussed in "Scientists still don't know the answer to this infamous question - Charles Wallace & Dan Kwartler"?

In order: 0:00 John Searle; 3:00 Artificial Consciousness Test.

What is mentioned in this TED-Ed episode?

2 people, works and ideas are mentioned, including John Searle, Artificial Consciousness Test. Each one on this page links to its own page with every episode that mentions it.

How long is "Scientists still don't know the answer to this infamous question - Charles Wallace & Dan Kwartler"?

The episode runs 6 min; the transcript is about 692 words in 8 timestamped sections.

More from TED-Ed

• All analyzed episodes
• Books mentioned on TED-Ed
• People mentioned on TED-Ed
• Topics mentioned on TED-Ed