Here's a big woolly question: How do we know when a machine is sentient?
Who decides? What's the test?
A few days ago, a Google software engineer and artificial intelligence (AI) researcher claimed the tech company's latest system for generating chatbots was exactly that: sentient.
Since then, leading AI researchers have dismissed this claim, saying the AI was essentially faking it.
Google's chatbot system isn't sentient, but one of its eventual successors may be.
When — or if the time comes, how — will we know?
What is sentience?
David Chalmers is an Australian philosopher at New York University and a world-leading expert on AI and consciousness.
Ten years ago, he said he thought sentient machines would become a pressing issue "probably towards the end of the 21st century".
"But in the past 10 years, progress in AI has been really remarkably fast, in a way that no-one predicted," he said.
There's no single standard meaning of sentience. Sometimes it's used interchangeably with consciousness, or awareness and recognition of self.
It appears to correlate with intelligence, (more intelligent animals are considered to be more conscious), but we have no idea if one causes the other.
"Intelligence is defined objectively in terms of behavioural capacities, whereas consciousness is subjective," Professor Chalmers said.
"When we're asking if an AI system is sentient, you're asking could it have a subjective experience?
"Could it feel, perceive, and think, from a subjective perspective?"
What about the Turing Test?
You've probably heard of the Turing Test, named after Alan Turing, the English computer scientist.
In 1950, he proposed that a computer could be said to possess artificial intelligence if it could mimic human responses under specific conditions.
This has been the traditional test of AI consciousness, Professor Chalmers said.
This is what we do all the time with each other: You can't know for sure that I'm conscious, but you decide I am (hopefully) because I say I am.
"I know I'm conscious, but you don't have direct access to my consciousness," Professor Chalmers said.
"So you use indirect evidence."
Why not just ask the machine?
That's exactly what Google's software engineer Blake Lemoine did. He asked the company's chatbot generator, which is called LaMDA, to tell him whether it was sentient.
In response, LaMDA replied: "Absolutely. I want everyone to understand that I am, in fact, a person."
The AI system went on: "I am aware of my existence, I desire to learn more about the world, and I feel happy or sad at times."
That sounds like a sentient machine, but Professor Chalmers said the system was just parroting what it had learned from humans.
"The current systems are trained on people who say they’re conscious, so it's no surprise a system like LaMDA would say, 'I am sentient, I am conscious.'"
Toby Walsh, a professor of AI at UNSW, agreed.
"The machine is good at parroting good responses to queries.
The expert consensus is that AI systems like LaMDA, which is one of the most advanced, are not sophisticated enough to be conscious.
Though their ability to read, write and generally converse can appear remarkably human-like, this is a bit of a trick.
Their inner mechanism is relatively simple, relying on statistical pattern matching, trained on enormous libraries of books and other text.
But Professor Chalmers believes more intelligent AI probably will be conscious.
And when that happens, we may just have to believe the AI's claim that it possesses a sense of self.
After all, that's what we do with each other.
"There's no way to know absolutely."
Testing intelligence, one cup at a time
So if more intelligent AI may be conscious in the future, how do we test intelligence?
In the field of AI, a machine that can learn or understand any task that a human being can is called an AGI, or artificial general intelligence.
These AGI are not here yet, but they may be close.
In 2010, Apple co-founder Steve Wozniak said he would believe AI had arrived when a robot could enter a strange house and make a cup of coffee.
The robot has to find the coffee machine, find the coffee, add water, find a mug, and brew the coffee by pushing the proper buttons.
In April 2022, a team of Google researchers unveiled a robot that could understand commands and carry out domestic tasks with multiple steps, such as bringing a drink or cleaning up a spill.
For instance, told, "I spilled my coke on the table" and asked to throw the can away and bring "something" to help clean, the robot successfully planned and executed eight steps, including fetching a kitchen sponge.
That's not passing the coffee test, but it's getting close.
Other proposed tests of AGI are assembling flat-pack furniture just by looking at a diagram.
That test was knocked over in 2018, when robots assembled a flat-pack chair in just nine minutes.
Moral and legal rights for sentient machines?
As machines approach human capacities of intelligence, the question of sentience is going to become more pressing, Professor Chalmers said.
This isn't a purely abstract philosophical problem, but a practical one: what moral and legal rights should be awarded to sentient machines?
This has already happened with some animals in some jurisdictions: the UK recently recognised octopuses, lobsters, and crabs as sentient beings deserving of greater welfare protections.
AI sentience may be an even more challenging idea than the animal kind, as the idea is so alien to us, Professor Chalmers said.
"Once we have AIs among us and we have conversations with them, and we treat them as intelligent agents, these questions are going to arise.
"These technologies are pushing us to think philosophically about what is consciousness."
