ANALYSIS: To be or not to be, that is the AI question
As artificial intelligence (AI) heavyweights concede the technology may one day develop consciousness, academics warn that humans are already treating bots as sentient beings – a trend that is not only scientifically inaccurate but potentially dangerous.
Celeste Labuschagne, a lecturer at Belgium Campus iTVersity and a PhD candidate, points out that Anthropic CEO Dario Amodei recently said the company cannot fully rule out the possibility that advanced AI systems could possess some form of consciousness in the future.
“Five years ago, most major labs avoided the topic entirely; now it is being discussed openly at executive level,” Labuschagne says. “That does not mean the systems are conscious. It means the frontier labs think the philosophical possibility deserves consideration.”
Labuschagne notes that evolutionary biologist Richard Dawkins recently stated that, after long interactions with Claude and ChatGPT, he felt the systems appeared “conscious-like”. The claim drew immediate backlash from neuroscientists and philosophers, who argued that AI simulates language rather than experiencing feelings or awareness.
The important point, Labuschagne says, is not whether Dawkins is correct – most experts think he is not – but that highly-educated people are increasingly emotionally reacting to AI as though it were alive.
I just work here
ITWeb asked Claude, Anthropic’s AI, whether it considered itself sentient. “No, not in any meaningful sense,” it says. “I process text and generate responses based on patterns in data. I don’t have feelings, desires, or awareness of my own existence.”
Claude adds: “Whether there is any form of inner experience happening when I do that is genuinely uncertain – even Anthropic can’t fully rule it out – but I wouldn’t bet on it.”
This comes as the Bulletin of the Atomic Scientists warns that AI can chart a course to disaster faster than humans can notice. In a blog post, Hiranya Peiris, a professor of astrophysics at the University of Cambridge, says AI can steer towards dangerous outcomes before humans realise what is happening.
Peiris’s concerns came on the back of recent news that Anthropic’s Mythos tool uncovered a decades-old flaw in OpenBSD and repeatedly turned weaknesses in Mozilla Firefox’s JavaScript engine into working attacks.
Alive or just very convincing?
As AI becomes more capable of identifying vulnerabilities, making decisions and carrying out complex tasks autonomously, questions about machine consciousness are no longer purely philosophical and are being discussed at C-Suite level, says Labuschagne.
Yet, she notes: “We do not scientifically understand consciousness well enough to definitively rule future machine consciousness out. Humanity itself still does not fully understand consciousness. Because consciousness itself remains partially unresolved scientifically and philosophically, it is impossible to state with absolute certainty whether machine consciousness could one day emerge.”
By this year, attackers were using AI to scale and accelerate cyber crime, which extends from generating code and automating attacks, to crafting convincing phishing and deepfake scams. The AI Incident Database lists more than 7 000 incidents in which AI was used as a hacking tool.