Some researchers think we might have AIs that are “serious candidates for consciousness” in the next decade – and that’s even if they don’t get to the human level of general intelligence.
ultra-famous 1974 essay by the philosopher Thomas Nagel called “What Is It Like to Be a Bat?”. In the essay, Nagel noted that bats navigate using echolocation, which is an experience that humans, with their very different senses, simply can’t comprehend for themselves.
But importantly, even if it’s not accessible to us, pretty much everyone agrees that there is something that it’s like to be a bat. A bat has an internal, mental, subjective experience – a bat is. It’s conscious in a way that a rock or a ladder or a cloud isn’t: its complex brain gives it that capability, and it experiences echolocation in the same sense that we experience something when, for example, we see the colour blue. Is it the same for an artificial intelligence? AI models are getting more and more complex, with longer and longer lists of capabilities. Might it get to the point where they experience something, too? Could an AI ever be conscious?. Of course, it’s in its very name that an AI involves “intelligence” of a sort, but the question is whether that intelligence is up to that of a human being. Computers have long been far more intelligent than humans at some specific tasks – say, remembering information, doing arithmetic, or playing chess. Large Language Models are now very intelligent at many more tasks, including getting very high scores on all sorts of human exams, but it’s still questionable whether they have human-like general intelligence. In 1950, Alan Turing suggested his “Imitation Game” , as a hypothetical way to try to adjudicate whether we can consider a machine “intelligent”. If a person is talking to two anonymous agents, one a human being and one a machine, and they can’t reliably guess which is which, then the machine can be considered to have intelligence. Hardly anyone talks about the Imitation Game any more, because the most advanced AI models can pass it without breaking a metaphorical sweat. Almost all of us would be easily fooled into thinking modern Large Language Model AIs are human. But, as Turing himself noted in his paper describing the Imitation Game, this doesn’t clinch the question about whether the machine is. After all, it could simply be responding to words and phrases in a reflex-like way, without any awareness or internal reflection on what it’s doing. That is, it could still behave very intelligently but not be conscious.clinch the question? What would convince us that there’s something that it’s like to be an AI? The first thing to say is that it’s not even clear that you are conscious. I know I’m conscious – it’s literally the most obvious thing in the world to me. But I have no idea if you are. By the very subjective nature of consciousness, it’s not possible to tell for certain whether anyone other than yourself is having a conscious experience. That includes if I just ask you whether you’re conscious – you could be mistaken, or even lying. That was the issue stumbled upon by an engineer working at Google who claimed that one of their AI models was “sentient” in mid-2022. In a conversation with the AI, it had told him that it was conscious and could experience emotions like fear. There was no evidence beyond these conversations, and he ended up being fired for what GoogleHe had no evidence – but what would such evidence look like? This has become a topic of discussion among scientists and philosophers in recent months. The well-known philosopher David Chalmers releasedwhere he considered some potential things that current AIs are missing that might stop them from being conscious. The first one is biology. Everything in the world that we think is conscious right now is biological, and has a brain. Could it be that there’s some as-yet unknown property of the way that a biological brain’s electrochemistry works that makes consciousness occur? That would mean that no electronic, non-biological machine, no matter how complex, could ever be conscious . But until anyone can point to this mysterious consciousness-producing biological process, it’s just a possibility, and not a convincing argument that rules out conscious machines. What about senses? A big part of being conscious is being able to sense things from the outside world – I already mentioned the experience of seeing the colour blue. AIs, being pieces of software on a computer, don’t have senses in the same way. But first, as Chalmers notes, we’re generally comfortable with the idea that a “brain in a vat” could be conscious, even if it doesn’t have a body with which to sense anything. And second, AI models do have capabilities that are similar to human senses: for instance, many AIs are adept at extracting information and meaning from pictures and sounds. So that doesn’t seem like much of an objection either.by a team of researchers that was released earlier this month. In it, the researchers – including computer scientists, psychologists, and neuroscientists – looked through the long history of research on how the human brain might produce consciousness, and made a checklist of characteristics that they’d consider evidence of consciousness.”. The idea is that, although our brain might have lots of specialised modules that all do their own tasks , the system has the capability of bringing it all together, sharing information across all the different modules. It’s that, claim the Global Workspace theorists, that causes consciousness to occur: bringing together all the details produces our overall experience, like seeing all the actors in a play together on a stage., suggests that we become actively conscious of the things we see only when they’re repeatedly fed through several different brain systems. We all see a lot of things that we’re not conscious of – in our peripheral vision, for instance, or that common experience where you’ve been driving for the past five minutes but have no memory of what you just drove past. To make it enter into our conscious attention, say the Recurrent Processing theorists, the brain has to perform some extra calculations on the visual information. The researchers described these and several other scientific theories of consciousness, and then thought about how many of their features were shared by the different AI models that are available today. Do AIs have a global workspace? Do they engage in recurrent processing? Sort of. No current AI model, concluded the researchers, ticked enough of the boxes for it to fit in to our current models of consciousness. Thus, they argued, it’s probably not the case that they’re conscious right now . What about in future? As Chalmers put it, most of the objections to AI being conscious aren’t permanent – there’s no reason in theory that they couldn’t be overcome in future by even more complex AIs. So that seems to be the consensus for now: AI isn’t conscious, but it very much could be in the future. And that future might be quite near: Chalmers thinks we might have AIs that are “serious candidates for consciousness” in the next decade – and that’s evenIs this all just a scientific curiosity? Is there any real reason, beyond the pure fun of it, to speculate about whether there’s something that it’s like to be an AI? One reason might be that, if we worked out a good theory of whether AIs are conscious, it could help us explain the mysteries of our own consciousness – a problem that philosophers have been worrying away at for thousands of years. But this might be much easier said than done: even if we were one day to fully understand it, a computerised consciousness would work in very different ways from our own . There are also moral considerations. One reason that many vegetarians and vegans decide not to eat or use animal products is that animals are harmed in the process – and a lot of that harm comes from the fact that animals are conscious. Almost everyone agrees that animals with more complex brains are more capable of suffering . So if a sufficiently complex AI was, in future, considered to be conscious, it would become a moral concern. It would no longer just be a piece of software, but a sentient being, and we’d have to think about how we treated it. The aforementioned Google employee cited the AI’s fear of being switched off as a major reason it was sentient – and indeed, we consider it wrong in almost all cases to “switch off” a human by killing them. Would it be the same for a conscious AI? Then again, most of us consider many animals conscious, and yet we still kill them, eat them, and subject them to factory-farming. Perhaps that implies that we wouldn’t care all that much about the feelings of a conscious AI., and how capable they could be in the real world, it might be worth thinking sooner rather than later about how we should treat them. We might not want to get on their bad side.There was another big media splash about “ultra-processed foods” on Monday – I wrote whatYou can’t have missed the grim story of the living, wriggling worm found inside an Australian woman’s brain. I , along with some other gnarly medical case studies – and some much bigger-scale problems with parasitical worms., I argued that you probably wouldn’t even have noticed it was much different from the average full moon, and so most people would’ve just ended up disappointed., on how the problems of replication that we’ve encountered in science might apply to the discipline of history, too. Does history have a replication crisis?
Please follow us on Google to support us United Kingdom Latest News, United Kingdom Headlines
Similar News:You can also read news stories similar to this one that we have collected from other news sources.
A14 lorry crash: Lorry driver arrested after collision near NewmarketThe driver of the car is taken to hospital with serious injuries.
Read more »
Reason we yawn when tired and when it may be sign of something seriousAlthough the exact science behind yawning is not yet 100 per cent clear, there are plenty of theories behind this everyday bodily reflex
Read more »
Police update as man fights for life after Queens Drive crashA motorbike rider was taken to hospital with serious chest injuries
Read more »
Man hospitalised after serious assault in Paisley flat as police issue appealMembers of the public are being asked to assist officers with their enquiries.
Read more »
Families of victims ‘delighted’ with plans to force offenders to sentencingCampaigners hailed a Government promise of new legislation to force serious offenders to attend court.
Read more »
Creamfields security guard in hospital with serious injuriesAn air ambulance was called to the festival site
Read more »