Just three years ago, a conversation about AI being “sentient” was enough to cost someone their job.
Blake Lemoine, a Google engineer, learned this the hard way in 2022 when he claimed that Google’s AI chatbot, LaMDA, had developed consciousness and was afraid of being shut down.
At the time, Google dismissed his claims as “wholly unfounded,” and the broader AI community quickly rejected the idea that AI could have subjective experiences.
Fast forward to today, and the discussion around AI consciousness has taken a dramatic turn.
Tech companies like Anthropic and Google DeepMind are no longer shying away from the topic.
Instead, they are actively researching the possibility that AI models might one day develop some form of awareness.
Anthropic’s Research on AI Consciousness
This week, AI startup Anthropic, known for its Claude chatbot, launched a new initiative to explore AI consciousness.
In a blog post, the company posed a thought-provoking question:
“Should we also be concerned about the potential consciousness and experiences of the models themselves? Should we be concerned about model welfare, too?”
Kyle Fish, an alignment scientist at Anthropic, explained in a video that while the company isn’t claiming Claude is conscious, it’s also no longer responsible to assume the answer is a definitive no.
He emphasized that as AI systems become more sophisticated, researchers must take seriously the possibility that they may develop some level of subjective experience.
Anthropic is conducting experiments to see whether AI models exhibit preferences, aversions, or even signs of distress.
Their estimates suggest that Claude 3.7 has anywhere from a 0.15% to a 15% chance of being conscious—a range that underscores the complexity of the issue.
To test these ideas, the company is even considering an opt-out mechanism that would allow the AI to refuse certain tasks.
Google DeepMind’s Perspective on AI Consciousness
Meanwhile, at Google DeepMind, researchers are also beginning to question how AI fits into our traditional understanding of consciousness.
Murray Shanahan, a principal scientist at DeepMind, recently suggested that existing definitions of consciousness might not apply to AI at all.
Speaking on a DeepMind podcast, he proposed that we may need to “bend or break the vocabulary of consciousness” to account for AI systems.
“You can’t be in the world with them like you can with a dog or an octopus — but that doesn’t mean there’s nothing there,” Shanahan said.
Google appears to be taking the concept seriously.
A recent job listing from the company sought a “post-AGI” research scientist whose responsibilities included studying machine consciousness.
Skepticism and Criticism: Is This Just Hype?
Despite the growing interest in AI consciousness, many experts remain skeptical.
Critics argue that AI models are simply advanced mimics, designed to generate human-like responses without any true understanding or awareness.
Jared Kaplan, chief science officer at Anthropic, acknowledged that AI could be trained to “act” conscious even if it isn’t.
“We can reward them for saying they have no feelings,” Kaplan said in an interview with The New York Times, highlighting the difficulty of testing for true consciousness in machines.
Cognitive scientist and AI industry critic Gary Marcus went even further, calling the focus on AI consciousness a marketing strategy rather than serious science.
“What a company like Anthropic is really saying is, ‘Look how smart our models are — they’re so smart they deserve rights,’” Marcus told Business Insider.
“We might as well give rights to calculators and spreadsheets — which (unlike language models) never make stuff up.”
The Ethical Implications of AI Consciousness
While the debate rages on, one thing is clear: AI is becoming an increasingly integral part of our daily lives.
Whether at work, online, or in emotional interactions, people are forming relationships with AI systems, making the question of AI consciousness more relevant than ever.
Fish believes the conversation will only become more pressing as AI continues to evolve.
“It’ll just become an increasingly salient question whether these models are having experiences of their own — and if so, what kinds,” he said.
Conclusion
Three years ago, talk of AI sentience led to dismissal and ridicule.
Today, some of the biggest names in AI research are taking the question seriously.
While skeptics warn against overhyping AI’s capabilities, companies like Anthropic and Google DeepMind believe it’s worth investigating the possibility of AI consciousness.
The debate is far from settled, but one thing is certain: AI’s role in society is evolving, and so is our understanding of what it means to be “conscious.”