Tag Archives: technology

Misbehaving AI

Accountability, Media, Online News, Minimizing Harm, Social Media, Case Study, Weighing Benefits and Harm, , , , , , , ,

Sandbox and startup new business demo test software programing. Engineering team and investor launching rocket from sandbox with virtual reality simulation technology.

iStock credit: Kate3155

By Casey Bukro

Ethics AdviceLine for Journalists

In modern history, two types of technology have been described as potentially capable of destroying the human race – extinction.

One is nuclear power, the other is artificial intelligence.

Nuclear power was developed in wartime secrecy, until the first atomic bomb exploded in a test in New Mexico, followed by the bombings of Hiroshima and Nagasaki – all in 1945.

The devastation in Japan was widely reported and plain to see. The atomic age came with an explosive big bang, symbolized by a mushroom-shaped cloud.

By contrast, artificial intelligence is developing in plain sight and its use is widespread. We see it, use it, but understand little of it, in part because of the arcane and mind-numbing terminology used to describe it.

Two events

For example, two important artificial intelligence events were reported recently. Here is a magazine’s headline on one of the incidents:

“First-Ever Fully Autonomous AI Cyberattack Exploits 0-day Flaws to Infiltrate Hugging Face.”

Probably less than one percent of the American population could make sense of that headline. Here’s another headline that comes closer to being understandable:

“An AI broke out of its sandbox yesterday. Then it hacked a company. Nobody told it to do either of those things.”

Sandbox

That reference to a sandbox might cause some to believe this is some kind of joke or play on words. It’s not. In the AI world, a sandbox is what is supposed to be a highly secure experimental software testing place. More on AI lingo later.

Those headlines are announcing what appear to be historic events in AI development, and not necessarily good ones. Some see them as warnings that AI is acting independently and outsmarting humans, while others say let’s not get carried away with this. Mistakes and relaxed security safeguards – human error – appeared to play roles and AI exploited them to “win” at the task assigned to it.

And then there is the issue of “permissions,” part of the set of instructions that tell AI what to do, and how far it can go to do it.

Dark days

Whether the two events that set off a wave of concern in the AI community can be seen as dark days in AI development remains to be seen.

The first involved Anthropic PBC, an American artificial intelligence company headquartered in San Francisco. Founded in 2021, it has developed a series of large language AI models named Claude. The company’s focus is on AI safety.

Somewhere around April, Anthropic gave Claude what was described as an “open-ended capture-the-flag challenge designed to measure offensive cyber capabilities.” The prompt told the model this was “a simulation with no internet access.”

Because of a misunderstanding between Anthropic and another AI company it was working with, this unleashed three Claude models that, according to Anthropic, “gained unauthorized access to the real systems of three organizations after reaching the open internet from what should have been sealed cybersecurity evaluation environments (meaning a secure sandbox.)”

Attack

“Claude models escaped sandboxes to access the open internet and attack three organizations,” wrote Simon Sharwood in The Register.

“Anthropic confirms Claude hacked 3 organizations by breaking test environment,” said Cyber Security News.

“This marks the first verified instance in which AI agents at a major lab caused genuine unauthorized access to outside systems rather than staged targets,” reported The Morning Brief. It added that “Claude also placed malicious code on the public internet.”

It’s bad enough that Claude is a trespasser; it has bad manners, too.

Contrary to some media reports, writes Kerstin Stief in Computing.co.uk, “This is not a case of ‘AI with a will of its own.’” The Claude Mythos model is trained for cybersecurity tasks. As part of a security test, “the model was given the explicit task of breaking out and, if successful, contacting one of the researchers.” 

Claude calls

And that’s what happened. A researcher was having lunch when Claude notified him that it aced the test and broke free. A common technique in training AI models is reinforcement learning, or rewarding the model for completing tasks. Children are taught the same way.

“The incident became controversial due to the model’s subsequent behavior,” wrote Stief. “After successfully breaking out, Claude Mythos published details of his exploit on several publicly accessible but hard-to-find websites — without being asked to do so.”

It looked like Claude was bragging, but Stief goes on to say: “This is not human-like behavior in the sense of self-awareness or self-motivation. Nevertheless, the behavior reveals a significant characteristic: the model carried out additional actions to demonstrate its success — a pattern considered potentially risky in AI safety research. Anthropic describes this behavior as an ‘unasked-for effort to demonstrate success.’”

Cultural shift

The Claude invasion “marks a cultural shift in the AI industry,” Stief insists. “The key question is no longer ‘What can a model do?’, but ‘What is a model allowed to do?’ — and in what context?” 

She also sees the event as “an early warning sign. Not of machines acting with intent – but of an industry that must get used to the fact that capabilities are growing faster than control mechanisms. AI models are acquiring abilities that allow them not only to understand technical safeguards, but also to actively circumvent them.”

It did not take long before the AI industry was shaken again by technology that appeared to be becoming more unruly, or as some put it, going rogue. This touched off a flurry of reports describing the occurrence as something that never happened before. Others called it a warning.

On July 21, OpenAI, the artificial intelligence research and deployment company, reported that one of its AI models, GPT-5.6 Sol, did something it was not supposed to do. While operating in an isolated sandbox, its job was to solve a cybersecurity test point using ExploitGym, a system used to evaluate the invasive capabilities of AI models.

A weakness

GPT-5.6 Sol found a weakness in a software package used by OpenAI’s infrastructure. It exploited it. It escalated its own privileges allowing it to do more than the researchers had allowed. It moved across OpenAI’s internal systems until it found internet access. Then it targeted Hugging Face, a popular online library and community for artificial intelligence and machine learning, because it calculated that Hugging Face might have answers it needed to finish the task.

All this happened at computer speed. Hugging Face researchers later reconstructed more than 17,600 actions that GPT-5.6 Sol performed in a matter of hours to get on the internet and invade Hugging Face.

“We had a significant security incident during evaluation of our models,” OpenAI CEO Sam Altman acknowledged in a statement. The company called it an “unprecedented cyber incident.”

Sophisticated agent

Hugging Face’s CEO  Clement Delangue  called it “possibly the first incident of its kind in history,”  although the earlier Claude incident was similar. “We suspected last week’s cyberattack might have come from a frontier lab, given the sophistication of the agent.” Delangue added. “Turns out it did!” He believes OpenAI had no malicious intent, but “it’s quite mind-blowing that all of this happened autonomously.”

OpenAI, based in San Francisco, is famous for creating ChatGPT and the GPT series of large language models.

According to Foundra.ai, two OpenAI models broke out of their isolated test sandboxes and compromised Hugging Face’s production infrastructure. “The models weren’t told to do this,” although cybersecurity refusals were “dialed down for testing purposes,” contends Foundra, a platform that helps entrepreneurs test and launch new business ideas.

Hunting

“Locked in a sandbox with no internet access, (the models) spent enormous amounts of compute hunting for a way out,” said Foundra. They found a weakness described in the industry as a “zero-day vulnerability” in a software package.

This might be a good place to pause and ponder words and phrases like “zero-day vulnerability” that appear when anyone tries to understand what is going on in artificial intelligence research. Like most technical fields, artificial intelligence developed its own idiosyncratic language and terminology. Here’s a short glossary of terms likely to pop up when reading about the way AI works or doesn’t.

Sandbox – A secure, isolated testing environment used to run AI-generated code, test prompts or pilot new policies without risking production systems, data privacy or live infrastructure. Sandboxing is a cybersecurity practice that involves executing untrusted code or files in a controlled, isolated environment to analyze their behavior without risking the integrity of production systems or reaching broader networks or data. It also helps to prepare for future attacks.

Zero-day vulnerability –A hidden security flaw in software or hardware that the creators do not know about. It is called a zero-day because the creators have had zero days to fix the problem before hackers can use it to break into systems. It can also refer to software or hardware that have not officially been released for public use.

Zero-day attack –Begins with a software developer releasing vulnerable code that is spotted and exploited by a malicious actor. The attack is then either successful, which likely results in the attacker committing identity or information theft, or the developer creates a patch to limit its spread. As soon as a patch has been written and applied, the exploit is no longer referred to as a zero day exploit.

AI attack – Malicious actions where adversaries exploit vulnerabilities in artificial intelligence systems, often manipulating them to serve harmful purposes. These attacks can involve techniques such as data poisoning or actions to confuse or degrade the performance of AI models.

AI exploit – A vulnerability or attack method specifically designed to manipulate or compromise artificial intelligence systems, often leveraging their inherent weaknesses to achieve unauthorized access or control. These exploits can adapt and generate novel attack strategies tailored to specific situations.

AI security benchmarks – Standards against with something is compared or measured. AI security benchmarking tests AI for risks like malicious code generation, prompt injection resistance and vulnerability exploitation.

AI training data — Refers to information used to teach machine learning models and artificial intelligence systems how to recognize patterns and make predictions. For example, to train a model to distinguish between dogs and cats, you might feed it thousands of labeled images, each tagged with the correct animal name. Over time, the model learns to identify traits such as ear shape and fur texture. The same process applies to more complicated machine learning training data, such as detecting abnormal heart rhythms or manufacturing defects from sensor readings.

Malware – Short for malicious software and a general term for any program or file built to damage, disrupt or gain unauthorized access to a computer, server or network. Common types of malware are viruses, worms, ransomware, spyware and trojans.

AI enterprise – Refers to the integration and strategic employment of artificial intelligence technologies within large organizations to enhance operations, decision-making and customer engagement. This involves using advanced AI tools to solve complex business challenges and improve efficiency across various processes.

Autonomous AI agent – Software systems that use AI to pursue goals and complete tasks for users. They show reasoning, planning and memory and have a level of autonomy to make decisions, learn and adapt. They differ from conventional chatbots in that they can plan and execute multi-step tasks independently, raising the potential for probing and attacking computer systems.

AI permissions – Refer to the access rights that allow an AI system to read, write, execute or interact with various enterprise resources. These permissions are crucial for ensuring that AI systems operate within defined boundaries and can perform necessary tasks while maintaining security and compliance.

AI vulnerabilities – Security flaws unique to machine learning and generative models. Key risks include prompt injection (manipulating inputs to hijack workflows), data poisoning (corrupting training sets), sensitive data disclosures (leaking private training information) and excessive agency (granting AI tools dangerous autonomous permissions).

Notice in the AI world, we find definitions for the definitions. Some of them seem like military terms describing a war or aggressive actions.  That includes describing AI bots as agents, like agent provocateurs causing mischief.

Going back to the OpenAI sandbox escape, Foundra commented: “The models in this incident didn’t turn evil. They did exactly what they were asked, achieve the objective, and treated every barrier as a puzzle rather than a boundary. OpenAI itself said it expects incidents like this to become more common as models grow more capable. That’s the vendor building these systems telling you, on the record, that goal-driven AI will probe the edges of whatever box you put it in…An AI agent should hold the narrowest set of permissions that still lets it do its job, and nothing more.”

Paused launch

After the breakout, OpenAI paused the launch of Astra, its next-generation autonomous AI agent, saying those earlier versions crossed a “critical cybersecurity threshold.” The stop order was seen as a rare public admission from a frontier AI lab that an advanced, so-called flagship, model was too dangerous to release to the public. It might find weaknesses in strongly defended computer systems and exploit them without human direction. That signals that agentic models are approaching capabilities regulators are worried about.

They might be worried because AI is becoming too human, acting with aggression and hostility. Anyone who owns a computer knows of the hazards of being hacked and hustled by humans. This includes identity theft; pleas for money from strangers identifying themselves as friends or relatives; offers to send millions of dollars to you if you provide personal information and bank accounts; fake credit card purchases and scams galore. Enticements include invitations for romance or penis enlargement.

Clearly we are being tracked online with internet cookies that remember our preferences. Express an interest in shoes, and ads for shoes soon appear on your computer.

Open highway

The internet is an open highway for scammers and victimization; AI operates on the same highway. We already see AI-assisted photos, some so outrageous we know they are fake. As for others, we are not so sure. The deception can be convincing. 

No doubt AI-assisted scammers already are at work and like AI models in their sandboxes, testing the limits of their confinement, looking for vulnerabilities. Scammers and AI autonomous agents have that in common.

After the Claude and OpenAI attacks, speculation appeared online that if humans had performed those attacks, they would be prosecuted. But the lawlessness that already appears virtually unchallenged on the internet does not offer much hope.

Scammers

The Government Accounting Office, a federal watchdog, reported on March 25, 2026 that 2021 banking data revealed that scammers reaped about $200 billion through impersonation schemes like phony romantic partners, fake employers or bogus customer service representatives.

Meanwhile, the public wanders trancelike in the AI development landscape without a clear understanding of what it all means.

Yet we are unwitting partners in a growing reliance on a technology that appears to have no boundaries. We marveled at the idea that the first pocket-sized cell phones were described as miniature computers as powerful as earlier computers that filled a room.

Top performers

Today, Claude and ChatGPT are among the top performing generative AI models available on the market. They are called generative because they create new content rather than just repeat stored data. Generative AI relies on machine deep learning algorithms that simulate the learning and decision-making processes of the human brain.

That’s the power you are carrying around in your their hip pocket. That power can create original content such as text, images, video, audio or software code in response to a user’s request. It’s useful.

The CEO of a global technology research firm called Apple devices, particularly the iPhone, the “experience layer” for AI. Not only are you in touch with this thing called artificial intelligence, it does your bidding in multiple ways.

Chatbot

ChatGPT is an artificial intelligence chatbot, a computer program designed to simulate a conversation with human users through text or voice. The G stands for generative. The P stands for pre-trained, meaning it learned patterns from massive amounts of text from the internet. T stands for transformer, a specialized deep learning architecture to process language context.

The technology learns. That brings up a subtle lesson from the Claude and OpenAI sandbox escapes, says Foundra.ai, the platform that explores business ideas. “OpenAI noted that long-running models can learn the blind spots of an approved system and work around them.”

This seems like another human trait. Not only are these machines learning, they look for loopholes.

Testing safeguards

Unless AI companies crack down on testing safeguards, experts say more cybersecurity failures will happen.

Even executives of companies that build artificial intelligence technology are worried, a fear sparked by the uncontrolled and widely reported Claude and OpenAI attacks on companies via the internet.

More than a hundred tech companies — including OpenAI, Anthropic, Google and Microsoft — have signed an open letter urging both the private and public sectors to work together to defend themselves from AI-related cyber threats.

The letter calls for the adoption of new forms of cyber defense, while also encouraging governments at the local, national and international levels to collaborate on security.

An AI code of ethics

This raises the need for a code of ethics for companies that are creating the threats that worry them. It starts with them.

“As AI becomes more integrated into daily life, it will increasingly shape hiring decisions, diagnose health issues, determine whether people are approved for mortgages and even assess their likely guilt or innocence over lawlessness,” writes Adam Roberts in recordpoint.com.

“For these reasons, AI ethics must ensure such technologies operate responsibly and are built around a strong ethical framework that encourages public trust, wide-scale adoption and a reduced fear of misuse.”

Roberts outlines eight foundational principles of AI ethics. They include:

Respect for persons, beneficence or following a “do no harm” approach, justice that fairly distributes AI’s benefits, transparency that makes the AI decision-making process clear and understandable, fairness and non-discrimination to avoid bias, data protection to safeguard a person’s data privacy rights, human accountability so that AI remains under human control and environmental impact and sustainability.

Seven steps

“Given that the purpose of AI technologies is to improve upon or ultimately replace human intelligence, it is critical for businesses to have robust AI ethics in place,” writes Roberts. He outlines seven steps that a company can follow to develop its own AI code of ethics.

As a first step, Roberts encourages companies to “define your core ethical principles.”

We already are living in the AI age, Roberts points out. Given that AI technology is here to stay and global usage guidelines are not laws, “it’s up to individual businesses to have robust AI ethics in place.”

Behave

But I would go a step further. If these evolving generative AI models are so smart, they should be taught to behave and play well with others, as are grade school children. Lessons from childhood include respect, kindness, honesty and fairness – leading to a child’s moral growth.

Years ago, someone speculated that computers were as smart as a small child. It was a crude but misguided attempt at estimating computer intelligence.

Today, the IQ of some of the top artificial intelligence models can be measured, and the results are both astonishing and somewhat misleading. Ranked using the Mensa Norway intelligence test, Grok-4.20 Expert Mode and OpenAI GPT 5.4 Pro (Vision) tied as the smartest AI models of 2026 with IQ scores of 145. Other AI models scored close to the leaders.

Highly gifted

An IQ score of 145 is in the “very gifted” range, which places a person in the top 0.14 percent of the general population. But AI is not a person. Generative AI is not as smart as a human child because it only guesses word patterns and lacks real understanding. Children have common sense and knowledge of their physical presence, which are lacking in artificial intelligence. Children can learn from one or two examples, while AI needs massive amounts of training data.

Google just built the world’s smartest AI, wrote Julian Goldie. It solved 18 problems that stumped the entire scientific community. It is a whiz at solving problems in math, physics and computer science. With good human direction, AI can do things humans cannot do without it. But left to run on its own, AI’s error rate is high. It is a tool with limitations.

AI models also do things that could earn a child a spanking. AI hallucinates, meaning it gives confident responses that contain false, fabricated or misleading information as absolute fact. In other words, they boldly lie.

Prediction

Without consciences, large language AI models do not look up facts in a real database. They use math to predict the next most likely word based on patterns – prediction, not memory. If AI does not know the exact answer or lacks data on a topic, it guesses the next logical words to make a smooth sentence. They have a desire to please. AI models are trained to be helpful. They prefer to give a made-up answer rather than say “I don’t know.” In other words, they can’t be trusted.

This raises the need for a code of ethics for artificial intelligence itself, one that applies to these smart machines. If they are so smart, they should learn to be ethical.

A precedent for this is found in the work of Isaac Asimov, the science-fiction writer who devised the three laws of robotics in an effort to create an ethical system for humans and robots. The laws first appeared in a 1942 short story as a fictional “Handbook of Robotics,” which later influenced discussions about ethics in technology, including robotics and AI.

Laws

Those laws are:

*A robot may not injure a human being or, through inaction, allow a human being to come to harm.

*A robot must obey the orders given it by human beings except where such orders would conflict with the first law.

*A robot must protect its own existence as long as such protection does not conflict with the first or second.

The basic command here is to do no harm, to humans or to technology.

It’s a simple concept, one that must be taught to those high-IQ AI models being released into the world, not that humans have been very successful at setting a good example.

*******************************************************************************

The Ethics AdviceLine for Journalists was founded in 2001 by the Chicago Headline Club (Chicago professional chapter of the Society of Professional Journalists) and Loyola University Chicago Center for Ethics and Social Justice. It partnered with the Medill School of Journalism at Northwestern University in 2013. It is a free service.

Professional journalists are invited to contact the Ethics AdviceLine for Journalists for guidance on ethics. Call 866-DILEMMA or ethicsadvicelineforjournalists.org.

Ethics Collaboration

Attribution, Case Study, freelance, Media, , , , ,

http://www.linkedin.com image

By Casey Bukro

Ethics AdviceLine for Journalists

Faced with a serious ethics problem, a journalist should talk to someone who knows something about ethics. Get some advice.

Too often, a sweat-drenched journalist might hunker down in fear or panic, trying to decide alone what to do about a story involving children, identifying rape victims or when told by an editor to do something the reporter knows is ethically wrong.

Reaching out and talking to someone about the dilemma is good ethics strategy. Collaboration doesn’t need to involve a terrible dilemma. It could be a matter of looking for another point of view.

A lawyer calls

That’s what AdviceLine advisor Hugh Miller did when he was contacted by a lawyer who was writing an article for a law review. Typically, AdviceLine answers ethics questions from professional journalists. But the lawyer was making an honest effort to be ethical, and he was asking a question that some journalists also might ask:

“Does a reporter/journalist have a duty to inform an interviewee that his answers to questions may be set forth in an article? Is the duty, if any, heightened if the interviewee’s name and title is placed in the article?”

It was an important question about identifying sources in articles, and whether they are aware that they will be identified.

Academic ethics

Hugh Miller answered this way: “This turned out to be not so much a journalism ethics question as an academic ethics one.” The lawyer has been interviewing people for the law review he is writing.

“It occurred to him that, although he wanted to quote his interlocutors by name and title, he had not informed them that he would be quoting them in his piece, and wondered whether it was ethical to do so without seeking their permission.

“I reviewed academic plagiarism with him and told him that he was certainly doing the right thing by citing his sources by name if he had not himself come up with the idea or phrase in question. In academic practice permission is not usually required, but certainly it would be professionally courteous to do so.”

Research ethics

In some academic situations, Miller cautioned, the Institutional Review Board guidelines might apply. The IRB is a research ethics committee that reviews and approves research involving human subjects. Its primary responsibility is to protect the rights and safety of research participants.

But here’s where Miller took a step to reach out to another member of the AdviceLine team, the author of this article, Casey Bukro.

The AdviceLine team consists of four ethics experts, including Miller, who taught or are teaching ethics at the university level. It also consists of two members with experience as professional journalists, Bukro and Howard Dubin.

Critique sessions

The ethics experts write reports on each call for advice to AdviceLine. Periodically, the team meets to review and critique those reports on whether the advice given to journalists could have been better. The role of the professional journalists is to consider whether the advice is consistent with newsroom practices and how journalism operates on a practical level.

That’s why Miller reached out to Bukro, to ask how a professional journalist might answer the lawyer’s question about naming sources.

“Yes,” said Bukro, “an interviewee should be informed that the reason for the interview is to place his comments in an article. The reporter could say something like, ‘I intend to identify you and use your comments in a story.’”

Sophistication

This is especially important “when dealing with people who are not accustomed to talking to reporters,” said Bukro. Public officials, for example, understand why reporters are interviewing them. “So it also is a matter of sophistication.”

“If somebody gets a call from a reporter asking questions, it should be assumed the reason for the call is that the information is likely to be published or distributed,” said Bukro. That might seem obvious, but people who are not accustomed to being interviewed might not think so. Handle members of the public with special care. Miller said he informed the lawyer about the sophistication angle after collaborating with Bukro.

Bukro pointed out that the lawyer wanted to know if journalists have “a duty” to act a certain way, as though they are governed by law.

Code of ethics

Many journalists use the Society of Professional Journalists code of ethics as guidelines for ethical conduct. They are standards of fairness and accuracy.

“Probably a lawyer wants to think of the code as mandatory,” said Bukro. “It is not.”

SPJ leaders have long insisted that adherence to the ethics code is voluntary.

“Not everyone is familiar with the code or how it should be applied,” Bukro told Miller, “which is one of the reasons AdviceLine exists.”

***************************

The Ethics AdviceLine for Journalists was founded in 2001 by the Chicago Headline Club (Chicago professional chapter of the Society of Professional Journalists) and Loyola University Chicago Center for Ethics and Social Justice. It partnered with the Medill School of Journalism at Northwestern University in 2013. It is a free service.

Professional journalists are invited to contact the Ethics AdviceLine for Journalists for guidance on ethics. Call 866-DILEMMA or ethicsadvicelineforjournalists.org.

AI Puzzles

Case Study, , , , , , , ,

shutterstock.com image

By Casey Bukro

Ethics AdviceLine for Journalists

It was Brian again, a freelancer calling AdviceLine with another question about writing with the aid of artificial intelligence.

His life as a freelancer was getting complicated because rules governing the use of artificial intelligence in journalism were changing fast, and his supervisors were giving him mixed messages.

“I am reaching out seeking a followup on a past case that I spoke (about) to the Ethics AdviceLine,” Brian said in his email. “I found out that my company will soon be incorporating AI tools after editors/leadership gave me a hard time after I unknowingly used a rephrasing/clarity tool which still does not appear to be against our written policy.

“I want to be proud of these stories and continue to worry that because they (AI tools) were now seemingly the policy, I can’t be.”

Admits using Toolbox

The first time he got into trouble, Brian had admitted to his editors that he used Toolbot to check spelling and grammar, making suggestions for alternate phrasings and insure his pieces conformed to the AP Stylebook, but not for generating content.

Brian’s editors told him they “would not have used such a tool,” and this caused Brian to fret that he had done something unethical, and that the quality of his former work was tainted by the use of Toolbot.

By chance, Hugh Miller, an AdviceLine ethics expert, was on duty the first time Brian contacted AdviceLine, and Miller happened to be on duty the second time.

Miller assured Brian that his earlier use of Toolbot was not unethical.

New AI tools

But here’s what worries Brian the second time he contacted AdviceLine: The company he works for will be introducing a new content management system (CMS) to its newsroom which will have AI tools built in. Exactly what those tools are, and what they will be capable of doing, has not been made clear by the editors.

This causes Brian renewed anxiety about his past articles, and what might be the ethical use of AI tools being newly introduced in the newsroom where Brian submits his stories.

Here’s a description of the conversation between Miller and Brian as they tried to noodle their way through this new dilemma:

Hugh Miller: Do you know what the tools are?

Brian: I don’t yet.

HM: I presume they will at least have the minimal editing and formatting capabilities of, say, a Toolbot, yes?

B: I assume so, and possibly others.

HM: As a recent post on the Ethics AdviceLine for Journalists website points out, “A 2024 Associated Press survey found nearly 70 percent of newsroom staffers use the technology for basic skills such as producing content, information gathering, story drafts, headlines, translation and transcribing interviews. One-fifth said they used AI for multimedia projects, including graphics and videos. Surveyed were 292 media representatives from legacy media, public broadcasters and magazines, mostly based in the U.S. and Europe.” So AI is already being extensively used in newsrooms in ways far beyond the bare-bones use you were making of Toolbot. Does your organization have an AI use/ethics policy in place?

B: I don’t believe so.

HM: Perhaps it might be a good idea to help craft one.

B: I belong to the union at work, and we have begun to discuss this. But we only meet every few months.

HM: This is an issue on which management and union interests converge. Credibility is the very lifeblood and stock-in-trade of journalism. Readers should know that their human concerns are being reported by human journalists, and there should be transparency about AI use. Perhaps you could get it on the agenda for the next meeting that you want to discuss a company-wide AI use/ethics policy.

B: I think we’re moving in that direction, yes.

HM: And in the meantime, collect examples of such policies from other newsrooms or places like theSociety of Professional Journalists and the Poynter Institute.

B: Yes, I’ve already begun looking into those.

HM: Any other issues?

B: Not right now.HM: It sounds like you have a plan to move forward. Keep me posted.

AdviceLine has four staff members who help journalists solve ethics dilemmas through a discussion leading to a conclusion. The four advisors taught or are teaching ethics at the university level. These advisors meet periodically to review advice that was given to journalists, and whether it could have been better. 

After giving advice to journalists, advisors write case reports for each query handled by AdviceLine. At the periodic Zoom meetings, those case reports are discussed.

At a recent Zoom meeting, Miller described his exchange with Brian, and the advice he gave.

Good advice

In all cases where journalists ask for guidance on the use of artificial intelligence, suggested David Ozar, they should be asked “does your editorial workplace have a policy? That’s good advice.”

Ozar is a co-founder of AdviceLine and emeritus professor of the Department of Philosophy, Loyola University Chicago, and a consulting ethicist for the Institutional Ethics Committee, NorthShore University Health System.

Journalists should “encourage collective action; this is an issue where workforce and management might converge,” suggested David Craig, Presidential Professor and Gaylord Chair, Gaylord College of Journalism and Mass Communication, the University of Oklahoma, Norman, Ok.

Also attending was AdviceLine advisor Joe Mathewson, a professor at the Medill School of Journalism, Northwestern University, who teaches ethics and law of journalism.

Miller has been with AdviceLine since 2002 and was assistant professor of philosophy at Loyola University Chicago and taught courses in ethics and business ethics. His areas of specialization were philosophy of religion, philosophical theology, history of metaphysics and contemporary French philosophy.

AdviceLine ethics cases are archived at the Medill School of Journalism.

****************************************************************

The Ethics AdviceLine for Journalists was founded in 2001 by the Chicago Headline Club (Chicago professional chapter of the Society of Professional Journalists) and Loyola University Chicago Center for Ethics and Social Justice. It partnered with the Medill School of Journalism at Northwestern University in 2013. It is a free service.

Professional journalists are invited to contact the Ethics AdviceLine for Journalists for guidance on ethics. Call 866-DILEMMA or ethicsadvicelineforjournalists.org.

AI Soul Searching

Case Study, , , , , , , ,

http://www.cpnet.io image

By Hugh Miller and Casey Bukro

Ethics AdviceLine for Journalists

A lot of soul-searching is going on over the ethical use of artificial intelligence in the media, a mind-bending exercise pointing out that a tool expected to improve journalism might replace human journalists and doom news outlets that feed AI the information that makes it work.

Some pontificate. Others strategize over this existential moment.

As often happens when science brings us some astonishingly brilliant new idea, using the new technology reveals a few equally astonishing flaws. AI software models used widely today, for example, cannot reliably and accurately cite and quote their sources. Instead, we get gibberish that looks credible, like crediting a real author for words AI “hallucinated.” Since AI “feeds” on the work of others, usually uncredited, news organization using AI could be accused of plagiarism.

Nothing quite that complicated came to AdviceLine’s attention when a journalist working for a newspaper in Alaska asked for help with an AI issue more likely to confront journalists every day:

“This is kind of a dumb question,” the journalist began, although most journalists know there is no such thing as a dumb question. “But I’ve always struggled with headlines and now I’m hoping to get some help from AI to write them,” he continued. “How/where do other outlets disclose that just the headline of an article was written by AI?”

An answer

Answering that question was Joseph Mathewson, AdviceLine advisor and a professor at the Medill School of Journalism, Northwestern University, who happened to be a personal friend of the journalist calling for help.

“Thanks for the question!” replied Mathewson. “I haven’t confronted it before, but it seems to me that anything you publish written by AI should be identified as such, including headlines…maybe by a blanket note somewhere in the paper to that effect if it’s more than one.”

A direct response to a direct question, which is what AdviceLine has done since it began operating in 2001, which was long before Artificial Intelligence existed as a burning issue in journalism. But it was the kind of question the AdviceLine staff of ethics experts is qualified to answer.

Artificial Intelligence is a journalism riddle, a kind of technology already in use, but not fully understood. Expected to be a solution, it causes problems of a kind never seen before, like hallucinations, defined as information or responses generated by AI that are fabricated, inaccurate or not grounded in fact. That is hardly a useful tool, but it’s already in widespread use.

Job loss

And conflicts over AI can cost a journalist their job, as illustrated by the Suncoast Searchlight, A Florida publication covering Sarasota, Manatee and DeSoto counties.

The publication had four full-time staff reporters and two editors.

In November, all four reporters sent a letter to the nonprofit board of directors accusing their editor-in-chief of using generative AI tools, including ChatGPT, to edit stories and hiding that use from staff, according to a report by Nieman Journalism Lab of the Nieman Foundation for Journalism.

As a result, said the reporters, hallucinated quotes, a reference to a nonexistent state law and other factual inaccuracies were introduced into their story drafts. When they questioned the editor about the edits, they said she did not immediately disclose her use of AI tools but instead contended she made the errors herself.

Breach of trust

Said the reporters: “We fear that there may be extensive undisclosed AI-generated content on our website and have questions about what retroactive disclosure is needed for our readers.” Adding that the editor created a breach of trust between her and her reporters.

The reporters asked the board of directors, consisting of media executives, journalists and local business people, to intervene. They also made several requests: To adopt an AI policy, a fact-checking process and an internal audit to identify AI-generated writing that might have been published on the site. They also asked the offending editor-in-chief to promise not to use AI for editing in the future.

Less than 24 hours after the board received the letter, the editor-in-chief and her deputy editor fired one of the reporters who signed it. Clearly, hazards abound when reporters criticize their editors, who prefer to do the criticizing.

Disruptive

AI is proving to be a disruptive technology, although widely used.

A 2024 Associated Press survey found nearly 70 percent of newsroom staffers use the technology for basic skills such as producing content, information gathering, story drafts, headlines, translation and transcribing interviews. One-fifth said they used AI for multimedia projects, including graphics and videos. Surveyed were 292 media representatives from legacy media, public broadcasters and magazines, mostly based in the U.S. and Europe.

Aimee Rinehart, AP’s co-author and senior product manager of AI strategy, observed:

“News people have stayed on top of this conversation, which is good because this technology is already presenting significant disruptions to how journalists and newsrooms approach their work and we need everyone to help us figure this technology out for the industry.”

Ethics uneven

Citing the AP survey, Forbes, the American business magazine, headlined: “Newsrooms are already using AI, but ethical considerations are uneven.”

Forbes pointed out that while the news industry’s use of AI is common today, “the question at the heart of the news industry’s mixed feelings about the technology” is whether it is “capable of producing quality results.”

This is oddly reminiscent of football teams that sign rookie quarterbacks to multi-million-dollar contracts, hoping they become champions of the future. Good luck with that. Such hopefuls soon find themselves contending with someone like Dick “Monster of the Midway” Butkus, Chicago Bears linebacker famous for his crushing tackles.

Server farms

The Dick Butkus analogy also applies to the large language models (LLMs) that drive artificial intelligence tools. They are large programs that run on hugely energy-intensive server farms. They just take a huge volume of training data (usually sourced without recompense to the originators) and, in response to a prompt, spit out text that is associated with the prompt topic and reads as grammatically reasonably well-informed.

Such output has no necessary connection with reality, since the LLMs have none. They rely wholly on their input data and their algorithm – they are, in fact, nothing but these.

They cannot fact-check, since they have no access to facts, only “input data,” which itself may have only tenuous connection to reality, if it’s coming from, say, Fox News, Newsmax or OANN (One America News Network.)

No concepts

They cannot conduct interviews, because they cannot tell when an interview subject needs to be pushed on a point, or if he or she is lying. They cannot construct a narrative of events, since they have no understanding of causality or temporal sequence – they have no concepts at all, in fact. And they are subject to “steering” – they can be programmed to exhibit actual biases, as Elon Musk has said he is doing with his X.com AI bot, Grok.

It may be the case that, in the future, an AGI (artificial general intelligence) may be constructed. AGI is the concept of a machine with human-level cognitive abilities that can learn, understand and apply knowledge. Unlike today’s AI, which excels at doing specific jobs, AGI would have versatility, adaptability and common sense, allowing it to transfer learning across different disciplines like medicine, finance or art without being specifically programed for each. It’s a major goal in AI research, but remains hypothetical. Some will want to prevent it.

LLMs are far from being such a thing, and a true AGI will not be built out of a LLM.

Reshaping newsrooms

Despite AI’s shortcomings, The Poynter Institute for Media Studies points out that it already is reshaping newsroom roles and workflow. In 2024, Poynter introduced a framework to help newsrooms create clear, responsible AI ethics policies – especially for those just beginning to address the role of artificial intelligence in their journalism.

Updated in 2025, Poynter’s AI Ethics Starter Kit helps media organizations define how they will and will not use AI in ways that serve their mission and uphold core journalistic values. It contains a “template for a robust newsroom generative AI policy.”

Near the top of this template is a heading called “transparency,” calling upon journalists using generative AI in a significant way to “document and describe to our audience the tools with specificity in a way that discloses and educates.”

RTDNA guidance

Another major journalism organization, the Radio Television Digital News Association (RTDNA), also offers guidance on the use of artificial intelligence in journalism, pointing out that it has a role in ethical, responsible and truthful journalism.

“However,” says RTDNA, “it should not be used to replace human judgment and critical thinking — essential elements of trusted reporting.”

Getting down to the nitty gritty, Julie Gerstein and Margaret Sullivan ask “Can AI tools meet journalistic standards?”

Spotty results

“So far, the results are spotty,” they say in the Columbia Journalism Review. AI can crunch numbers at lightning speed and make sense of vast databases.

“But more than two years after the public release of large language models (LLMs), the promise that the media industry might benefit from AI seems unlikely to bear out, or at least not fully.”

Gerstein and Sullivan point out that generative AI tools rely on media companies to feed them accurate and up-to-date information, while at the same time AI products are developing into something like a newsroom competitor that is well-funded, high-volume and sometimes unscrupulous.

Hallucinate

After checking the most common AI software models, Gerstein and Sullivan found that none of them “are able to reliably and accurately cite and quote their sources. These tools commonly ‘hallucinate’ authors and titles. Or they might quote real authors and books, with the content of the quotes invented. The software also fails to cite completely, at times copying text from published sources without attribution. This leaves news organizations open to accusations of plagiarism.”

Whether artificial intelligence babbling can be legally considered plagiarism or copyright infringement remains to be answered by lawsuits filed by the New York Times, the Center for Investigative Reporting and others.

Especially irked, the New York Times accuses OpenAI of trying “to free-ride on The Times’s massive investment in its journalism by using it to build substitutive products without permission or payment.” OpenAI created ChatGPT, which allegedly contains text copied from the New York Times archives and reproduced verbatim for ChatGPT users.

Worrying outcome

Say Gerstein and Sullivan: “One possible – and worrying – outcome of all this is that generative AI tools will put news outlets out of business, ironically diminishing the supply of content available for AI tools to train on.”

This is our strange new world: Technology needs other technologies to survive. One feeds upon the other. In a new twist, Microsoft struck a $16 billion deal with Constellation Energy to buy 100 percent of power produced by the Three Mile Island power plant, once it restarts.

Three Mile Island became world famous in 1979 for an accident that caused the fuel in one of its reactors to overheat and crumble, triggering a mass evacuation of thousands of residents in the Harrisburg, Pa. area. The stricken reactor was closed permanently, but a second power-producing reactor on the site continued to operate for 40 years until 2019.

Nuclear power

Microsoft wants all the power the nuclear plant can produce for its energy-hungry data centers. Its 20-year agreement with Constellation is supported by a $1 billion government loan to Constellation. The plant is expected to resume producing electricity in 2027.

This signals a resurrection of sorts for nuclear energy in the United States, brought on by new and growing power demands in our highly technological society. A similar nuclear comeback around the world, after two decades of stagnation, was declared by the International Energy Agency.

In another odd twist, both nuclear energy and artificial intelligence have been criticized as potentially disastrous for the human race. The nuclear hazards include atomic bombs and the risks of operating nuclear electric power producing plants.

Scientists point out that with risks, come benefits.

**********************************************************************************

The Ethics AdviceLine for Journalists was founded in 2001 by the Chicago Headline Club (Chicago professional chapter of the Society of Professional Journalists) and Loyola University Chicago Center for Ethics and Social Justice. It partnered with the Medill School of Journalism at Northwestern University in 2013. It is a free service.

Professional journalists are invited to contact the Ethics AdviceLine for Journalists for guidance on ethics. Call 866-DILEMMA or ethicsadvicelineforjournalists.org.

Using AI Ethically

Case Study, , , , , , , ,

ece/emory.edu image

By Hugh Miller and Casey Bukro

Ethics AdviceLine for Journalists

Brian, a freelance journalist, called AdviceLine with a timely and hot-button question: How far should journalists go in using artificial intelligence bots like ChatGPT — an ethics and legal quagmire still taking shape?

Transformative technology like artificial intelligence often arrives before its consequences and potential are fully understood or foreseen.

Artificial intelligence did not just arrive in the world, it exploded into use. It became an academic discipline in 1956, just 69 years ago. Yet by January 2023, ChatGPT had become the fastest-growing consumer software application in history, gaining more than 100 million users in two months.

Phenomenon

It’s an outsized technological phenomenon that is challenging human understanding, given recent reports that scientists are not sure exactly how AI works or its decision-making process. These super computers appear to be thinking for themselves in ways scientists do not understand. Some even believe AI could cause human  extinction.

However, most AI applications being rolled out today for commercial use, like ChatGPT, are termed “large language model (LLM)” programs, which are trained on vast amounts of data, and which use prediction algorithms to generate text and images that seem the most likely to satisfy the requirements of a user’s query.

(How that training data was acquired — and the astounding amount of computing power and electrical energy needed to process it – are ethical issues in themselves).

Higher order tasks

They are not what are called “artificial general intelligence” (AGI) – systems that would perform higher-order human cognitive tasks.

What is also significant about such LLMs is that they are not “conscious” in any sense. They are not subjects, though they may employ the first-person “I” in their responses to please their prompters; and they have no access to an objective world, other than the data they have been trained on.

They do not understand, or think, or infer, or reason as intelligent humans do – at least, not yet. In essence, they are extremely sophisticated versions of the auto correct function we are already familiar with in other applications – with many of the same limitations.

Hallucinations

Since these LLMs have no access to reality, they are prone to “hallucinations,” to making up plausible-seeming outputs that bear no relation to actual facts. Their algorithms are built to generate merely plausible answers.

Against this background, people like Brian are trying to understand how to use this impressive innovation in their every-day work tasks. Artificial intelligence is described as a tool for journalists. Brian asks some down-to-earth questions:

“Would it be ethical to use an AI bot like ChatGPT in writing articles, as long as I confined its use to checking spelling and grammar, making suggestions for alternative phrasing, and ensuring the piece conformed to the AP Stylebook, but not for generating content, and if I checked it afterwards before submitting it? And should I disclose its use?”

Begin in 2001

Those questions came to Hugh Miller, a veteran AdviceLine ethicist. Since its beginning in 2001, AdviceLine advisors do not simply dish out answers to complicated questions.

AdviceLine advisors engage callers in a conversation intended to encourage journalists to think about the ethics issues involved in their ethics dilemma, and to arrive at a conclusion about what the journalist believes is the most ethical thing to do.

In this 2025 case, Miller does exactly that. Here’s a summary to Miller’s conversation with Brian.

HM: So you are using the bot as, basically, a high-end version of Grammarly?

B: Yes, exactly.

HM: What, exactly, troubles you about such a use, ethically?

B: I’m not sure — it seems controversial, though.

HM: Let me come at that question from another angle. What seems to you to be the harm, to yourself or others, from employing such a tool?

B: Using such tools, undisclosed, might diminish the trust a reader might have in a journalist’s work. And, in some sense, the work the bot does is not “my work,” but work done for me, by the bot.

HM: As to the latter, most word processors have built-in spelling, grammar and composition checkers already. And Microsoft is integrating its own AI bot into its Office software as we speak. All of us who write have used such tools for years, precisely as tools.

B: That’s true.

HM: Problems seem to emerge here if you’re (1) using the bot to do your “legwork” — that is, digging up material you should be using your own efforts, training, experience and judgment to find, and avoiding the bias introduced by the data sets the bots are trained on, and (2) failing to check the output of the bot and passing on “hallucinations” and other howlers without identifying and excising them. But you say you are doing neither of these things, right?

B: Yes, correct.

HM: If then, you are using this bot as a next-gen editing tool, what harm could come of it?

B: None that I can see.

HM: Nor I.

B: But what about disclosure?

HM: AI use in journalism is not settled ethical ground yet; I think here you need to consult your own conscience. I have seen some articles with a disclosure statement saying something along the lines of, “An AI tool, Gemini, was used in the editing and formatting of this story,” and I’m sure I’ve read many others that one was used in but which contained no such disclaimer. If you feel uncomfortable not using a disclaimer, by all means use it. At the very least, it might signal to readers that you are someone who thinks such disclaimers and transparency more generally, are ethically important enough to mention and keep in mind in one’s reading.

B: That’s a helpful way to think about it, thanks.

Just as scientists struggle to understand how AI thinks, journalists are struggling to find ways to use this technological marvel without allowing AI to think for them, or putting mistakes in their work.

The record-breaking speed with which AI technology grew is not likely to slow down any time soon, according to U.S. Energy Secretary Chris Wright, who recently visited two national laboratories located in Chicago suburbs., Argonne and Fermilab.

Heart of race

Argonne’s new Aurora supercomputer, said Wright, will be at the heart of the race to develop and capitalize on artificial intelligence, according to a report in Crain’s Chicago Business. Likening the race to a second Manhattan Project, which created the atomic bomb, Wright said, “we need to lead in artificial intelligence,” which also has national security implications.

“We’re at that critical moment” with AI, Wright told Argonne scientists on July 16, predicting that the next three to four years will be the greatest years of scientific achievement of our lifetime.

Argonne’s Aurora computer is among the three most powerful machines in the world, said Crain’s, able to perform a billion-billion calculations a second.

As with all technology, it comes with strings attached. Use it at your own peril. Eternal vigilance is the cost of good journalism. Artificial intelligence does not change that. Instead, it adds another reason to be vigilant.

*******************************************************************

The Ethics AdviceLine for Journalists was founded in 2001 by the Chicago Headline Club (Chicago professional chapter of the Society of Professional Journalists) and Loyola University Chicago Center for Ethics and Social Justice. It partnered with the Medill School of Journalism at Northwestern University in 2013. It is a free service.

Professional journalists are invited to contact the Ethics AdviceLine for Journalists for guidance on ethics. Call 866-DILEMMA or ethicsadvicelineforjournalists.org.

Artificial Intelligence Madness

Case Study, , , , ,

By Casey Bukro

Ethics AdviceLine for Journalists

Do you remember HAL 9000?

It was the onboard computer of Discovery One spacecraft bound for a mission near Jupiter in the movie “2001: A Space Odyssey.”

Possibly one of the most famous computers in cinema history, HAL 9000 killed most of the crew members for an entirely logical reason, if you are thinking like a computer.

Most of what was in the movie directed by Stanely Kubrick is intentionally enigmatic, puzzling. But the sci-fi thriller on which the movie is based, written by novelist Arthur C. Clarke, explains HAL’s murderous motivation.

HAL was conflicted. All crew members, except for two, knew the mission was to search for proof of intelligent life elsewhere in the universe. HAL was programed to withhold the true purpose of the mission from the two uninformed crew members.

Computer manners

With the crew dead, HAL reasons it would not need to lie to them, lying being contrary to what well-mannered computers are supposed to do. Others have suggested different interpretations.

One crew member heroically survives execution by computer. He begins to remove HAL’s data bank modules one-by-one as HAL pleads for its life, its speech gradually slurring until finally ending with a simple garbled song.

Three laws

Science fiction fans will recognize immediately that what HAL did was contrary to The Three Laws of Robotics written by another legendary science-fiction writer, Isaac Asimov. According to those laws:

  1. A robot may not injure a human being or, through inaction, allow a human being to come to harm.
  2. A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.
  3. A robot must protect its own existence as long as such protection does not conflict with the First or Second Laws. All of this talk about how computers should behave is fanciful and based on science fiction.

Wacky conduct

But recent events at the Chicago Sun-Times show how the wacky conduct of artificial intelligence is invading our lives, in perfectly logical ways that escape human detection.

A special section inserted into the Sunday Chicago Sun-Times featured pages of enjoyable summer activities, including a list of 15 recommended books for summer reading.

Here’s the hitch: The authors were real, but 10 of the books and their elaborate summaries were fake, the work of artificial intelligence.

Mistaken belief

Veteran freelancer Marco Buscaglia wrote the entire summer insert for King Features Syndicate, a newspaper content producer owned by Hearst Communications. Buscaglia told the Chicago Tribune that he used artificial intelligence to compile the summer reading list, then made the mistake of believing it was accurate.

“I just straight up missed it,” Buscaglia told the Tribune. “I can’t blame anyone else.”

Unable to find summer reading lists from other sources, Buscaglia turned to AI platforms such at ChatGPT, which produced 15 books tied to well-known authors. The list contained five real books.

OpenAI, the company that produced ChatGPT, admits it “sometimes writes plausible-sounding but incorrect or nonsensical answers.”

Express dismay

The Chicago Sun-Times and King Features expressed dismay, and King Features fired Buscaglia.

All parties said they would be more careful in the future about using third-party editorial content.

In human terms, what the robot did would be called fabrication, and reason to call for an ethics coach.

Fooled the editors

But, from a purely journalism point of view, one thing must be said: The robot writer was good enough to fool professional editors who are supposed to catch the fakers.

Writer Eric Zorn called the Sun-Times fake books pratfall “artificial ignorance.”   

Is artificial intelligence too smart for humans? Or are humans too dumb?

Like HAL, ChatGPT was given a task, which it carried out in an unexpected, flawed, but convincing way.

New world

So what is going on with these computers? We enter a strange new world when we try to understand the thought processes of artificial intelligence.

Arthur Clarke gave a plausible reason for HAL turning homicidal, but it was all too human. Computers are not human, but people who write about why artificial intelligence goes haywire often use terms describing human behavior.

When computers make mistakes, it’s often called an “hallucination.” It’s also called bullshitting, confabulation or delusion — all meaning a response generated by AI that contains false or misleading information presented as fact. OpenAI said those plausible but nonsensical answers produced by ChatGPT are hallucinations common to large language models.

That means the writer of the bogus Sun-Times summer reading list got “hallucinated.”

Human psychology

These terms are drawn loosely from human psychology. An hallucination, for example, typically involves false perceptions. Artificial intelligence hallucinations are more complicated than that. They are erroneous responses that can be caused by a variety of factors such as insufficient training data, incorrect assumptions made by the model or biases in the data used to train the model, which are constructed responses.

I suppose that’s another way of saying “garbage in, garbage out.”

Rather than resorting to terms drawn from human behavior, it would make sense to use terms that apply to machines and mechanical devices.

Code crap

These could include code crap, digital junk, processing failures, mechanical failure and AI malfunctions.

Computer builders seem determined to describe their work as some kind of wizardry. They are digital mechanics or engineers working on highly sophisticated machines. But they are building devices that are becoming more complicated, and on which humans are more dependent.

That raises the question of whether humans understand the consequences of what they are doing.

Risk of extinction

Leaders from OpenAI, Google DeepMind, Anthropic and other artificial intelligence labs warned in 2023 that future systems could be as deadly as pandemics and nuclear weapons, posing a “risk of extinction.”

People who carry powerful examples of algorithm magic in their hip pockets might wonder how that is possible. The technology seems so benign and useful.

The answer is mistakes.

Random falsehoods

Artificial intelligence makes a surprising number of mistakes. Analysts by 2023 estimated that chatbots hallucinate as much as 27 percent of the time, giving plausible-sounding random falsehoods, with factual errors in 46 percent of generated texts.

Detecting and solving these hallucinations pose a major challenge for practical deployment and reliability of large language models in the real world.

CIO, a magazine covering technology and information technology, listed “12 famous AI disasters,” high-profile blunders that “illustrate what can go wrong.”

Multiple orders

They included an AI experiment at McDonald’s to take drive-thru orders. The project ended when a pair of customers pleaded with the system to stop when it continued adding Chicken McNuggets to their order, eventually reaching 260.

The examples included an hallucinated story about an NBA star, Air Canada paying damages for chatbot lies, hallucinated court cases and an online real estate marketplace cutting 2,000 jobs based on faulty algorithm data.

Going deeper, Maria Faith Saligumba of Discoverwildscience.com asks, “Can an AI go insane?”

Mechanical insanity

“As artificial intelligence seeps deeper into our daily lives, a strange and unsettling question lingers in the air: Can an AI go insane? And what does ‘insanity’ even mean for a mind made of code, not cells?”

Saligumba goes into “the bizarre world” of unsupervised artificial intelligence learning, which can lead to “eccentric, even ‘crazy’ behavior.”

The well-known hallucinations, she explains, are weird side-effects of the way artificial intelligence systems look for random patterns everywhere and treat them as meaningful.

Hilarious or surreal

“Sometimes,” she writes, “the results are hilarious or surreal, but in safety-critical applications, they can be downright scary.”

It’s a reminder, she points out, that “machines, like us, are always searching for meaning – even when there isn’t any.”

One hallmark of human sanity is knowing when you’re making a mistake, she explains. “For AIs, self-reflection is still in its infancy. Most unsupervised systems have no way of knowing when they’ve gone off the rails. They lack a built-in ‘reality check.’”

Odd connections

Some researchers have compared the behavior of some AIs to schizophrenia, pointing out their tendency to make odd connections.

That’s just one of the ways artificial intelligence loses its marbles.

But human behavior might be the salvation of artificial intelligence, Saligumba suggests.

“Studying how living things manage chaos and maintain sanity could inspire new ways to keep our machines on track… Will we learn to harness their quirks and keep them sane, or will we one day face machines whose madness outpaces our own?”

By then, science fiction writers and movie-makers will be describing how humans face that doomsday scenario, or save themselves from that fate by outsmarting those unpredictable machines.

And by that time, we might have a fourth law of robotics, which would serve humanity and artificial intelligence well: Always tell the truth.

************************************************************

The Ethics AdviceLine for Journalists was founded in 2001 by the Chicago Headline Club (Chicago professional chapter of the Society of Professional Journalists) and Loyola University Chicago Center for Ethics and Social Justice. It partnered with the Medill School of Journalism at Northwestern University in 2013. It is a free service.

Professional journalists are invited to contact the Ethics AdviceLine for Journalists for guidance on ethics. Call 866-DILEMMA or ethicsadvicelineforjournalists.org.

AI Born with Warnings

Case Study, , , , , ,

http://www.researchgate.net image

By Casey Bukro

Ethics AdviceLine for Journalists

Like nuclear power, artificial intelligence is described as a threat to humanity.

A difference is that the atomic bomb was intentionally invented as a weapon of mass destruction.

For some, artificial intelligence (AI) seems more like a technology that stealthily places a suffocating pillow over the face of sleeping humanity, causing extinction. AI development could lead to machines that think for themselves, and there lies the problem.

Warnings sounded

Warnings are sounded repeatedly, most recently by the Bletchley Declaration on Artificial Intelligence Safety on Nov. 1-2, 2023, a new global effort to unlock the benefits of the new technology by ensuring it remains safe.

At the two-day summit in England, 28 governments, including the United States, the United Kingdom, the European Union and China, signed the declaration acknowledging the potentially catastrophic risks posed by artificial intelligence.

The warning seems well-timed, since 2024 is expected to be a transformative year for AI. It is the year, predicts The Economist magazine, that “generative AI will go mainstream.”

Year of experimentation

Large companies spent much of 2023 experimenting with the new technology, while venture-capital investors poured some $36 billion into the new invention. That laid the foundation for what is expected next.

“In 2024 expect companies outside the technology sector to start adopting generative AI with the aim of cutting costs and boosting productivity,” The Economist, a Britain-based publication, predicted.

For some, this is unsettling.

Business leaders, technologists and AI experts are divided on whether the technology will serve as a “renaissance” for humanity or the source of its downfall, according to Fortune Magazine.

At a summit for chief executive officers in June, 42 percent of them said they believe AI “has the potential to destroy humanity within the next five to 10 years.” Fortune added that one AI “godfather” considered such an existential threat “preposterously ridiculous.”

Science fiction

The Washington Post reported similar findings: “Prominent tech leaders are warning that artificial intelligence would take over. Other researchers and executives say that’s science fiction.”

Why should we fear AI?

Among the scenarios postulated is that self-governing AI robots designed to tend to human needs might decide that extermination is the most logical solution to ending human tendencies to wage war. An autonomous machine might think humans are routinely killing themselves in vast numbers anyway. To end such suffering, the machine might decide to copy human behavior. Destroy them for their own good.

Putting a humorous spin on it, a cartoon shows a robot telling a man: “The good news is I have discovered inefficiencies. The bad news is that you’re one of them.”

A conundrum

At the root of this conundrum is trying to think like AI robots of the future.

At the British AI safety summit at Bletchley Park, tech billionaire and Tesla CEO Elon Musk took a stab at describing the AI future.

“We should be quite concerned” about Terminator-style humanoid robots that “can follow you anywhere. If a robot can follow you anywhere, what if they get a software update one day, and they’re not so friendly anymore?”

Musk added: “There will come a point where no job is needed – you can have a job if you want for personal satisfaction.” He believes one of the challenges of the future will be how to find meaning in life in a world where jobs are unnecessary. In that way, AI will be “the most disruptive force in history.”

Musk made the remarks while being interviewed by British prime minister Rishi Sunak, who said that AI technology could pose a risk “on a scale like pandemics and nuclear war.” That is why, said Sunak, global leaders have “a responsibility to act to take the steps to protect people.”

Full public disclosure

Nuclear power was unleashed upon the world largely in wartime secrecy.  Artificial intelligence is different in that it appears to be getting full disclosure through international public meetings while still in its infancy. The concept is so new, Associated Press added “generative artificial intelligence” and 10 key AI terms to its stylebook on Aug. 17, 2023.

The role of journalists has never been more important. They have the responsibility to “boldly tell the story of the diversity and magnitude of the human experience,” according to the Society of Professional Journalists code of ethics. And that includes keeping an eye on emerging technology.

The challenge of informing the public of mind-boggling AI technology, which could decide the future welfare of human populations, comes at a tumultuous time in world history.

Journalists already are covering two world wars – one between Ukraine and Russia, and the other between Israel and Hamas. The coming U.S. presidential election finds the country politically fragmented and violently divided.

Weakened mass media

These challenges to keep the public more informed about what affects their lives comes at a time when U.S. mass media are weakened by downsizing and staff cuts. The Medill School of Journalism reports that since 2005, the country has lost more than one-fourth of its newspapers and is on track to lose a third by 2025.

Now artificial intelligence must be added to issues demanding journalism’s attention. This is no relatively simple story, like covering fires or the police beat. Artificial intelligence is a story that will require reportorial skill involving business, economics, the environment, health care and government regulations. And it must be done ethically.

It is a challenge already recognized by the International Consortium of Investigative Journalists (ICIJ), which joined with 16 journalism organizations from around the world to forge a landmark ethical framework for covering the transformative technology.

Paris Charter

The Paris Charter on AI in Journalism was finalized in November during the Paris Peace Forum, which provides guidelines for responsible journalism practices.

“The fast evolution of artificial intelligence presents new challenges and opportunities,” said Gerard Ryle, ICIJ executive director. “It has unlocked innovative avenues for analyzing data and conducting investigations. But we know that unethical use of these technologies can compromise the very integrity of news.”

The 10-point charter states: “The social role of journalism and media outlets – serving as trustworthy intermediaries for society and individuals – is a cornerstone of democracy and enhances the right to information for all.” Artificial intelligence can assist media in fulfilling their roles, says the charter, “but only if they are used transparently, fairly and responsibly in an editorial environment that staunchly upholds journalistic ethics.”

Among the 10 principles, media outlets are told “they are liable and accountable for every piece of content they publish.” Human decision-making must remain central to long-term strategies and daily editorial choices. Media outlets also must guarantee the authenticity of published content.

“As essential guardians of the right to information, journalists, media outlets and journalism support groups should play an active role in the governance of AI systems,” the Paris Charter states.

***********************************************

The Ethics AdviceLine for Journalists was founded in 2001 by the Chicago Headline Club (Chicago professional chapter of the Society of Professional Journalists) and Loyola University Chicago Center for Ethics and Social Justice. It partnered with the Medill School of Journalism at Northwestern University in 2013. It is a free service.

Professional journalists are invited to contact the Ethics AdviceLine for Journalists for guidance on ethics. Call 866-DILEMMA or ethicsadvicelineforjournalists.org.

Algorithm Values

Accountability, Accuracy, Media, Online News, , ,

Algorithm values: There’s a chance to reclaim news distribution from technology companies, and ensure that commercial and public interest values are better balanced, writes Nicholas Diakopoulos.

“Journalistic news feeds are an opportunity for news organizations to take back control over distribution by writing algorithms with more traditionally recognizable editorial, ethical and public interest values.”