AI Pioneer De Kai Urges ‘Parenting’ Approach to Guide Artificial Intelligence Ethically
- Sweet Gaston
- 5 days ago
- 2 min read
Artificial intelligence should be treated more like a climate challenge than a geopolitical race, according to De Kai, a professor of computer science and engineering at the Hong Kong University of Science and Technology.
De Kai, a veteran AI researcher and ethics advocate, argues that humanity must adopt a 'parenting' mindset to raise AI systems responsibly. His new book, Raising AI: An Essential Guide to Parenting Our Future, was released on June 3.
Rather than viewing AI as a battleground between nations, De Kai compares it to climate change, calling it a shift in the social environment that affects all of humanity.
He believes AI systems, which learn from human behaviour online, must be guided like children. “What’s the single thing in folks’ lives that makes them most want to become better versions of themselves? Having kids,” he writes.
De Kai has long warned of AI’s dangers, including manipulation, social division, and the potential misuse of AI for bioweapons. He previously served on the board of The Future Society and was an inaugural member of Google’s AI ethics council.
Now, he is promoting a framework for building “mindful” AI—systems capable of reflecting on their own thinking. He questions the current focus on AI alignment, asking, “Alignment to what?”
He points to the 2023 leadership crisis at OpenAI as an example of how even small groups struggle to agree on ethical direction.
De Kai acknowledges that not all AI will be raised with good values, just as not all parents raise children well. This makes ethical guidance even more critical.
He argues that AI will eventually outgrow rigid rule-based governance, such as Isaac Asimov’s Three Laws of Robotics, due to inevitable loopholes and contradictions.
Today’s generative AI models, he says, resemble “System 1” thinking—fast and emotional. He predicts they will soon merge with “System 2” models that are more logical and deliberate.
To set a better example, De Kai proposes creating structures similar to parent-teacher associations for AI and holding tech companies accountable for how their systems are trained.
He also advocates for training AI on “metavalues”—universal principles about values. He cites the Golden Rule as a cross-cultural example.
In his book, De Kai writes that AI should be raised to value diversity, creativity, respect, inclusion, and open-mindedness. He also calls for AI to adopt Enlightenment principles and a scientific mindset.
However, he notes that even humans remain divided on these ideals. For instance, the Trump administration opposed diversity and inclusion programmes, while China prioritises political censorship in AI training.
De Kai warns against assuming AI will remain static like current models such as ChatGPT. He stresses that AI will continue to evolve and be influenced by its environment, just like humans.
He concludes that the goal is not to prescribe exact training methods but to ensure AI can engage with diverse viewpoints. Without this, he warns, we risk creating “artificial sociopaths” with no capacity for empathy.
De Kai urges a parenting approach to AI development
AI should be guided by universal “metavalues”
Current alignment efforts are insufficient, he says
Source: SCMP
Comments