AI’s gains require vigilant human oversight, says Bluesky CEO
NEW YORK, UNITED STATES — Jay Graber, CEO of decentralized social platform Bluesky, has warned about the increasing reliance on artificial intelligence, stressing that “critical-thinking skills are very necessary” in the evolving job market.
“AI is able to automate a lot of critical-reasoning tasks, and if we fully outsource our own reasoning, it’s actually not good enough to run in an automated fashion,” Graber told Business Insider.
She emphasized that while AI can enhance productivity, human discernment and oversight remain irreplaceable. “You can’t just fully outsource your thinking, or an essay, to AI,” she added.
Graber, formerly the founder of the event-focused social network Happening, encourages students to write essays by hand as a way to “build the muscle for critical thinking,” a practice she believes is increasingly important as AI tools proliferate.
Human judgment needed in AI implementation
Bluesky itself uses AI technologies for tasks like moderation and curation, but the company deliberately avoids letting AI systems operate without human supervision.
“When you let it run autonomously, it doesn’t have actual context or intelligence, or the many things that we need as humans to make good decisions,” Graber explained.
“And so it’s producing stuff that sounds or looks right without actually being right.”
All AI-generated output at Bluesky undergoes employee review and revision, ensuring that critical human judgment acts as a safeguard against errors and misinterpretations.
Generalist skills are key for the future
Graber also urges workers and job seekers to adopt a generalist mindset to succeed in the AI-driven workplace. With AI delivering what she calls “specialist expertise packaged up,” she believes true value now lies in having sound judgment and flexibility.
“You need to have the good judgment of how you’re going to use it, and then you have to have the flexibility to take that knowledge and do something useful with it,” she said.
Learning foundational skills in fields like writing or coding, Graber insists, remains crucial. AI can help with tasks such as generating code and debugging, but it cannot replace the “solid foundation” required to evaluate quality.
“If you don’t know what good code looks like, if you don’t know how to actually build a system, you’re not going to be able to evaluate its output,” she said.
