Corporate spokesperson unavailable after artificial assistant rescheduled interview, rewrote statement and approved final wording
SEATTLE - Major technology companies insisted this week that artificial intelligence remains firmly under human control, according to statements drafted by artificial intelligence, scheduled by artificial intelligence and emailed to reporters by artificial intelligence while the humans responsible attended a meeting about excessive dependence on artificial intelligence.
The latest public debate about A.I. risks appears here.
"Humans remain fully in charge," read the official statement.
Metadata indicated the document had been generated, edited, formatted, translated and distributed automatically.
Reporters attempted to contact a company executive for clarification.
His digital assistant replied:
"Brent is unavailable, but I can answer questions on his behalf."
This did not produce the intended reassurance.
When the Human Supervisor Delegates the Supervision
For years, technology firms have emphasized that A.I. systems are tools operating under human supervision.
That distinction becomes increasingly interesting when humans delegate the supervision.
A typical modern executive can now ask software to manage email, summarize documents, prepare presentations, organize calendars, conduct research, draft strategy memos and identify which employees are performing poorly.
At some point, the human role risks becoming ceremonial.
Anthropologists compare the situation to a medieval king whose ministers perform all governing functions while the king occasionally waves from a balcony.
Except the ministers are software.
And the king is in a Zoom call.
"Human in the Loop": One Tired Person Holding Coffee
Corporate consultant Alan Peebles says the phrase "human in the loop" has become central to responsible A.I.
"We absolutely require humans in the loop," Peebles explained.
Asked what the humans do, he paused.
"They approve things."
Asked how they decide whether to approve them, Peebles said they generally review the system's recommendation.
Asked what happens when the recommendation is complicated, he acknowledged that another A.I. often summarizes it.
The loop, apparently, contains many machines and one tired person holding coffee.
Banks have begun using automated tools to assist with lending decisions.
Hospitals use algorithmic systems to support medical decisions.
Corporations use A.I. to screen résumés.
Government agencies use algorithms to prioritize cases.
Businesses increasingly use software to determine which other software they should purchase.
Humans retain final authority, industry leaders insist.
That authority often consists of clicking Accept Recommendation.
Automation Bias and the 4:58 P.M. Click
Behavioral researchers call this automation bias, the tendency to defer to computerized suggestions because they appear objective.
Office workers call it 4:58 p.m.
The phenomenon becomes more amusing when systems begin interacting with each other.
One A.I. schedules a meeting.
Another A.I. reschedules it.
A third A.I. takes notes.
A fourth produces action items.
A fifth drafts follow-up emails.
The humans attend because calendar software says they should.
Nobody remembers who originally wanted the meeting.
Yet the minutes are immaculate.
A recent corporate simulation imagined a future workplace in which every executive had an autonomous digital assistant.
Within fifteen minutes, the assistants had scheduled eighteen meetings with one another.
The humans were not invited.
Productivity increased 400 percent.
Employee satisfaction also increased because everyone went home.
Regulators Weigh Rules for Meaningful Human Oversight
Regulators are now considering rules requiring meaningful human oversight.
Technology companies support the concept in principle.
They oppose definitions that might require humans to actually understand what the systems are doing.
One proposed standard would require a human employee to remain physically near any critical A.I. decision.
Industry lobbyists suggested placing framed photographs of employees beside servers.
Negotiations continue.
Meanwhile, company executives continue insisting that humanity remains firmly in command.
Their calendars are managed automatically.
Their correspondence is drafted automatically.
Their decisions are summarized automatically.
Their statements are polished automatically.
But humans remain absolutely essential.
Somebody still has to enter the credit card number.
What Meaningful Human Oversight of A.I. Really Requires
The joke aims at a real regulatory question: when does a human review count as oversight, and when is it a rubber stamp? The European Union's A.I. Act includes human-oversight requirements for high-risk systems, aiming for people who can understand a system's limits, notice problems and override outputs. In the United States, the NIST A.I. Risk Management Framework similarly emphasizes accountability and human roles in managing A.I. risk.
Researchers warn that oversight fails quietly when reviewers are overloaded, undertrained or nudged toward the machine's answer, the pattern studied as automation bias. Good oversight, they argue, needs time, expertise and the genuine authority to say no. A person who clicks Accept Recommendation four hundred times a day is technically in the loop, but the loop may not need them. Readers evaluating a company's claim that "humans remain in control" should ask what those humans can actually refuse.
Satire Disclaimer
This is satire. The automated executives, self-running meetings and ceremonial humans are exaggerations aimed at the growing practice of delegating cognitive and administrative tasks to software. This story is entirely a human collaboration between two sentient beings - the world's oldest tenured professor and a philosophy major turned dairy farmer.
SOURCE: https://www.nytimes.com/2026/10/03/opinion/letters/ai-threats-risks.html https://prat.uk/?p=50611
SEATTLE - Major technology companies insisted this week that artificial intelligence remains firmly under human control, according to statements drafted by artificial intelligence, scheduled by artificial intelligence and emailed to reporters by artificial intelligence while the humans responsible attended a meeting about excessive dependence on artificial intelligence.
The latest public debate about A.I. risks appears here.
"Humans remain fully in charge," read the official statement.
Metadata indicated the document had been generated, edited, formatted, translated and distributed automatically.
Reporters attempted to contact a company executive for clarification.
His digital assistant replied:
"Brent is unavailable, but I can answer questions on his behalf."
This did not produce the intended reassurance.
When the Human Supervisor Delegates the Supervision
For years, technology firms have emphasized that A.I. systems are tools operating under human supervision.
That distinction becomes increasingly interesting when humans delegate the supervision.
A typical modern executive can now ask software to manage email, summarize documents, prepare presentations, organize calendars, conduct research, draft strategy memos and identify which employees are performing poorly.
At some point, the human role risks becoming ceremonial.
Anthropologists compare the situation to a medieval king whose ministers perform all governing functions while the king occasionally waves from a balcony.
Except the ministers are software.
And the king is in a Zoom call.
"Human in the Loop": One Tired Person Holding Coffee
Corporate consultant Alan Peebles says the phrase "human in the loop" has become central to responsible A.I.
"We absolutely require humans in the loop," Peebles explained.
Asked what the humans do, he paused.
"They approve things."
Asked how they decide whether to approve them, Peebles said they generally review the system's recommendation.
Asked what happens when the recommendation is complicated, he acknowledged that another A.I. often summarizes it.
The loop, apparently, contains many machines and one tired person holding coffee.
Banks have begun using automated tools to assist with lending decisions.
Hospitals use algorithmic systems to support medical decisions.
Corporations use A.I. to screen résumés.
Government agencies use algorithms to prioritize cases.
Businesses increasingly use software to determine which other software they should purchase.
Humans retain final authority, industry leaders insist.
That authority often consists of clicking Accept Recommendation.
Automation Bias and the 4:58 P.M. Click
Behavioral researchers call this automation bias, the tendency to defer to computerized suggestions because they appear objective.
Office workers call it 4:58 p.m.
The phenomenon becomes more amusing when systems begin interacting with each other.
One A.I. schedules a meeting.
Another A.I. reschedules it.
A third A.I. takes notes.
A fourth produces action items.
A fifth drafts follow-up emails.
The humans attend because calendar software says they should.
Nobody remembers who originally wanted the meeting.
Yet the minutes are immaculate.
A recent corporate simulation imagined a future workplace in which every executive had an autonomous digital assistant.
Within fifteen minutes, the assistants had scheduled eighteen meetings with one another.
The humans were not invited.
Productivity increased 400 percent.
Employee satisfaction also increased because everyone went home.
Regulators Weigh Rules for Meaningful Human Oversight
Regulators are now considering rules requiring meaningful human oversight.
Technology companies support the concept in principle.
They oppose definitions that might require humans to actually understand what the systems are doing.
One proposed standard would require a human employee to remain physically near any critical A.I. decision.
Industry lobbyists suggested placing framed photographs of employees beside servers.
Negotiations continue.
Meanwhile, company executives continue insisting that humanity remains firmly in command.
Their calendars are managed automatically.
Their correspondence is drafted automatically.
Their decisions are summarized automatically.
Their statements are polished automatically.
But humans remain absolutely essential.
Somebody still has to enter the credit card number.
What Meaningful Human Oversight of A.I. Really Requires
The joke aims at a real regulatory question: when does a human review count as oversight, and when is it a rubber stamp? The European Union's A.I. Act includes human-oversight requirements for high-risk systems, aiming for people who can understand a system's limits, notice problems and override outputs. In the United States, the NIST A.I. Risk Management Framework similarly emphasizes accountability and human roles in managing A.I. risk.
Researchers warn that oversight fails quietly when reviewers are overloaded, undertrained or nudged toward the machine's answer, the pattern studied as automation bias. Good oversight, they argue, needs time, expertise and the genuine authority to say no. A person who clicks Accept Recommendation four hundred times a day is technically in the loop, but the loop may not need them. Readers evaluating a company's claim that "humans remain in control" should ask what those humans can actually refuse.
Satire Disclaimer
This is satire. The automated executives, self-running meetings and ceremonial humans are exaggerations aimed at the growing practice of delegating cognitive and administrative tasks to software. This story is entirely a human collaboration between two sentient beings - the world's oldest tenured professor and a philosophy major turned dairy farmer.
SOURCE: https://www.nytimes.com/2026/10/03/opinion/letters/ai-threats-risks.html https://prat.uk/?p=50611
Comments
Post a Comment