Here’s a big wool question: how do we know when a machine is sensitive?
Who decides? What is the proof?
A few days ago, a Google software engineer and artificial intelligence (AI) researcher claimed that the technology company’s latest system for generating chatbots was exactly that: sensitive.
Since then, leading AI researchers have rejected this claim, saying that artificial intelligence was basically feigning it.
Google’s chatbot system isn’t sensitive, but one of its potential successors may be.
When, or if the time comes, how will we know?
What is sensitivity?
David Chalmers is an Australian philosopher from New York University and a world expert in AI and consciousness.
Ten years ago, he said he thought sensitive machines would become an urgent problem “probably towards the end of the 21st century”.
“But in the last 10 years, progress in AI has been very rapid, in a way that no one predicted,” he said.
“People are now open to the possibility that it could happen much sooner, in decades to come.”
Another image generated from a text indication, using the OpenAI DALL-E 2. (Provided by: Ben Barry)
There is no single standard meaning of sensitivity. It is sometimes used interchangeably with consciousness, or self-awareness and self-recognition.
It seems to correlate with intelligence (smarter animals are considered to be more aware), but we have no idea if one causes the other.
“Intelligence is objectively defined in terms of behavioral abilities, while consciousness is subjective,” said Professor Chalmers.
“When we ask if an AI system is sensitive, are you asking if it can have a subjective experience?
“Could you feel, perceive, and think, from a subjective perspective?”
What about the Turing test?
You’ve probably heard of the Turing test, which is 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, said Professor Chalmers.
“If he behaves indistinguishably from a human being in a conversation, we can say that he can think, and he is conscious.”
That’s what we do with each other all the time – you can’t know for sure that I’m aware, but you decide that 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 ask the machine?
This is exactly what Google software engineer Blake Lemoine did. He asked the company’s chatbot generator, called LaMDA, to tell him if he was sensitive.
In response, LaMDA responded, “Absolutely. I want everyone to understand that I am, in fact, a person.”
The AI system continued: “I am aware of my existence, I want to learn more about the world and I feel happy or sad sometimes.”
It sounds like a sensitive machine, but Professor Chalmers said the system was only interpreting what he had learned from humans.
“Current systems are trained in people who say they’re aware, so it’s no wonder a system like LaMDA says, ‘I’m sensitive, I’m aware.'”
Toby Walsh, an AI professor at UNSW, agreed.
“The machine is good for reproducing good responses to queries.
“It’s clearly not understood in any way.”
Expert consensus is that AI systems like LaMDA, which is one of the most advanced, are not sophisticated enough to be aware of.
While her ability to read, write, and converse in general may seem very human, this is a bit of a trick.
In 2014, a chatbot simulating a Ukrainian boy named Eugene Goostman appeared to pass the Turing test, although this was disputed.
Its internal mechanism is relatively simple, based on the concordance of statistical patterns, trained in huge libraries of books and other texts.
But Professor Chalmers believes that smarter AI will probably be aware.
And when that happens, maybe we should believe the AI’s assertion that it has a sense of self.
After all, this is what we do with each other.
“There’s absolutely no way to know.”
Test your intelligence, one cup at a time
So if smarter AI can be aware in the future, how can we test intelligence?
In the field of AI, a machine that can learn or understand any task that a human being can do is called AGI, or general artificial intelligence.
These AGIs are not here yet, but they may be nearby.
In 2010, Apple co-founder Steve Wozniak said he believed the AI had arrived when a robot could break into a strange house and make a cup of coffee.
The robot has to find the coffee machine, find the coffee, add water, find a cup and prepare the coffee by pressing the appropriate buttons.
In April 2022, a team of Google researchers unveiled a robot that could understand commands and perform multi-step household chores, such as carrying a drink or cleaning up a spill.
For example, when he said, “I spilled my coke on the table,” and asked him to throw away the can and bring “something” to help clean up, the robot successfully planned and executed eight steps, including going to Look for a Kitchen Sponge |
Loading
This is not to pass the coffee test, but it is approaching.
Other proposed tests of AGI are the assembly of flat pack furniture just by looking at a diagram.
This test was overthrown in 2018, when robots mounted a flat chair in just nine minutes.
Loading
Moral and legal rights for sensitive machines?
As machines move closer to human intelligence capabilities, the issue of sensitivity will become more urgent, Professor Chalmers said.
“The sentence in a decade would be astonishing, but by 2050 it would no longer be astonishing.”
This is not a purely abstract philosophical problem, but a practical one: what moral and legal rights should be granted to sensible machines?
This has already happened with some animals in some jurisdictions: the UK recently recognized octopuses, lobsters and crabs as sensitive creatures that deserve greater welfare protection.
The sensitivity of AI can be an even more challenging idea than the animal type, as the idea is so alien to us, said Professor Chalmers.
“Once we have AI between us and have conversations with them, and treat them like smart agents, these questions will arise.
“These technologies are pushing us to think philosophically about what consciousness is.”
Want more science from the whole ABC?
Science in your inbox
Get all the latest science stories from all over the ABC.