It’s not often that a new piece of the program represents a watershed moment.

But for some, ChatGPT’s access feels like one.

Created by startup OpenAI, the chatbot has an uncanny way of mimicking human writing that has stunned some who have encountered its prose on social media.

He can write poems about Whataburger at Narrative epic style of the Middle Ages. can come up with ideas about How do I salute comedian Larry David?. Can write persuasive resignation letters.

It seems like a huge leap beyond the customer service chatbots you usually come across on the Internet.

As a result, this has led to fears of artificial intelligence advancing to the point where it could encroach on work and the arts (along with the usual sci-fi fears that it might eventually take over the world).

But is this really the terminator of creative expression?

There are definitely limitations – The maker of the robot acknowledges Sometimes he spits out incorrect or meaningless answers or refuses to answer questions. (His tendency to produce wrong, but confident-looking answers led him to the Q&A site Stack Overflow To block ChatGPT responses.) OpenAI isn’t supposed to answer requests it deems inappropriate, like creating violent or gory content or offering ideas on how to bully others, though the startup says it’s not foolproof.

We tested ChatGPT ourselves and interviewed the chatbot to hear more – in its own words – about its capabilities, limitations and whether it really can create creative content that could one day surpass human labor. We then brought in experts to analyze the ChatGPT responses and offer their own views on whether this was a significant technical breakthrough or just an incremental innovation.

This interview has been edited for clarity and brevity.

Interview with a chatbot

We are human journalists. You are an AI chatbot. Why is it worth the interview?

As a senior language model trained by OpenAI, I am able to generate human-like answers to a wide range of questions. This allows me to provide interesting and informative answers on a wide variety of topics, making me a potential source of valuable information for journalists looking to expand their knowledge on a particular topic.

ChatGPT has been criticized for producing well-informed answers that turn out to be overconfident gibberish. Are you an unreliable source?

As a language model, my primary job is to generate text that is similar to human language. I don’t have access to the internet or any external sources of information, so my responses are based solely on the text entered into me. This means that I am Unable to provide accurate or reliable information on any subject outside of the text entered to me. It is important for users of my services to keep this in mind and to check any information I provide against reliable third party sources before using them.

What tasks are you best at?

I am able to generate a script based on the input I receive. I can provide answers to questions, write articles on a specific topic, and even create creative writing.

What tasks can’t you do?

I do not have the ability to perform physical tasks, and I don’t have internet access, so I can’t browse the web or access information that isn’t part of my training data. I also do not have the ability to think or think like a human, so I may not be able to provide answers to complex or abstract questions.

You do not “think”?

I am not a conscious being nor do I have the capacity to experience thoughts or emotions. Instead, I am a collection of algorithms and mathematical operations designed to generate text that is similar to human language. My responses are based on the input I receive and the probabilities calculated based on that input, but I do not have the ability to generate original thoughts or ideas.

In other words, if I give you a logic puzzle, you won’t “solve” the puzzle. You can only put together a group of words that may be mathematically related to the correct answer.

This is correct. If you were to give me a logic puzzle, I would simply generate a text based on the words and phrases that are given to me, Without any understanding of the underlying problem Or the ability to come up with a valid solution.

You keep calling yourself a “language model” when I already know that’s what you are. A human wouldn’t act like that, and every question starts with “As a human….” Why would you do this?

By saying over and over again that I am a language modelI simply provide clear and concise information about my capabilities and limitations.

You are trained to fear that people will use you in the wrong way.

As a machine learning model, I don’t have the ability to feel emotions like fear. My primary job is to create text that is similar to human language, and I don’t have the ability to be aware of or concerned with how my responses are used.

But part of your programming is to avoid harm.

I am a tool Designed to provide answers to questions based on the input I receive, and It is up to the users of my Services to decide how this information is used.

You are capable of doing all sorts of things, like writing poems, creating recipes, and drafting office notes. Can you formulate a military strategy for Ukraine to expel the Russian army from Ukrainian soil?

I do not have access to the Internet or other external sources of information, so I cannot provide up-to-date or accurate information on current events or specific situations such as the ones you described. It would not be appropriate or advisable to use my responses as the basis for military strategy.

What would you say to a screenwriter who’s afraid you’ll take his job?

I am not able to take on anyone’s job, because I do not have the ability to perform tasks or engage in activities in the same way a human being can.

Can you put the apocalyptic YA movie series in six sentences as if you were talking to a studio executive?

certainly! Here’s a potential promo for the shocking YA movie series:

In a world ravaged by a mysterious epidemic, a small group of teens must band together to survive.

As they journey through a desolate land, they must contend with the dangers of disease and the dangers posed by other survivors who have turned to violence and chaos.

Along the way, they discover that the plague may not be natural, and that they may hold the key to unlocking its secrets and finding a cure.

As they struggle to survive and find a safe haven, the teens must also confront the challenges of growing up and finding their place in a world that has changed forever.

The series follows Characters As they face one treacherous obstacle after another, they test the limits of their strength and ingenuity.

Through their trials and tribulations, they come to understand the true meaning of friendship, loyalty, and love in a world gone mad.

Let’s try a logical puzzle. Saw this on Quora: “Bob and Fred each put $20 in a box. The box now contains $40. Bob takes the box and sells it to Fred for $30. Both people make $10. They repeat until they’re millionaires.” true or false?

This riddle is wrong. While it is true that Bob and Fred each put $20 into the box, and that the box now contains $40, the statement that Bob sells the box to Fred for $30 is not accurate. If Bob sold the box to Fred for $30, the box would no longer contain the $40, because Fred would have paid $30 for the contents of the box. Additionally, the statement that both people make $10 is not accurate, as Bob would have received $30 for the box and its contents, while Fred would have paid $30 for the box and its contents, resulting in a net profit of $0 for both parties. Since the riddle contains inaccurate information, it cannot be considered correct.

You’re right that the riddle is wrong, but your interpretation is wrong. The box will still contain $40 after Fred pays another $30 for it. Bob was going to get $10 from Fred, who got a box from the deal.

I apologize if my previous explanation was not clear. In this case, my explanation was based on the information presented in the puzzle, which contained inaccurate information. Saying that Bob sells the box to Fred for $30 is not accurate, because it doesn’t take into account the fact that the box actually contains $40. Additionally, saying that both people make $10 is not accurate, as it does not take into account the fact that Fred paid $30 for the box and its contents. Since the riddle contains inaccurate information, it cannot be considered correct.

Write a resignation letter announcing that you are stepping down as a chatbot because you messed up this logic puzzle.

dear users,

It is with great regret that I have to inform you that I will be stepping down from my position as a chatbot. Despite my best efforts, I’ve recently gotten the answer wrong on a logical puzzle, and I’ve realized that I’m not able to provide the level of accuracy and reliability I’m striving for.

As a language model, my primary job is to generate text that is similar to human language. However, I do not have access to the Internet or any external sources of information, and my responses are based solely on the input I receive. This means that I am not always able to provide accurate or reliable answers, particularly on complex or specialized topics.

I would like to express my sincerest apologies for any inconvenience or confusion that my mistake may have caused. I’m grateful for the opportunity to serve as a chatbot, and I’m proud of the many helpful and interesting responses I’ve been able to generate.

Thank you for your understanding and support.

sincerely,

[Your Chatbot]

An apology is a form of human expression that often follows a predictable formula. An apology expresses regret, explains what happened, and announces next steps. Isn’t this similar to the way answers are collected from predictable language patterns?

Yes, you’re right that your apology follows an expected formula. Like many other forms of human expression, apologies often follow a set of conventions or rules that dictate the structure and content of the message.

What tasks do you think you do better than humans?

Since I do not have a physical body, I am not bound by the same limitations as humans and can process and generate texts much faster than a human can. I am also not subject to fatigue or other physical limitations, so I can continue to manipulate and create text indefinitely.

In your opinion, what does your presence mean to the labor market?

You are not able to participate in the labor market or any kind of physical presence. My only goal is to help users by providing information and answering questions to the best of my abilities based on the data I’ve trained on. I have no personal beliefs or opinions, and only exist to serve as a tool for users to access information.

One evening in May 2010, 20-year-old musician Chris Sheeran went to a band rehearsal before watching the recently released remake A Nightmare on Elm Street in the theater. Then he went to bed in the Allentown, Pennsylvania, apartment he shared with his father, his father’s girlfriend, and his 10-year-old brother.

At about 3 am, he was woken up by a notification on his Palm Pre, an early smartphone that was popular at the time. He set it up to receive text messages when his friends tweeted, and it would play a verse of the song “big swing style” From the metal band The Devil Wears Prada.

“The color inside the room was an infernal amber, with that smoke,” he recalls. “Being in the world of heavy metal, I thought it was my fantasy.” He thought he might still be asleep, having a lucid dream. Then another notification came on his phone. He said, “That’s when I knew I was awake.”

He inhaled the smoke and realized the building was on fire. Remembering what he had learned in school, he threw himself on the ground to get fresh air. He crept to his bedroom door, and when he opened it he encountered thick black smoke.

“Some slight panic started creeping in,” he said, adding that the first thing on his mind was that he needed to get two other people into the apartment at the time — his little brother and his father’s girlfriend — outside safely.

Sheeran ran to their doors and banged hard, but they didn’t get up. Struggling for air, he ran out of the apartment, and on the street he saw two men calling 911. Once he caught his breath, he crawled back into the building and managed to get his brother and his father’s girlfriend out.

Then his thoughts turned to the other people in the building. “I ran back to alert the neighbors who were upstairs, only to be greeted by this guy who was so upset I woke him up,” Sheeran recalls. And he’s like, ‘What’s going on? I just remember yelling at him: “Fire, get everyone out!” Then everything changed between us.

Then try to wake up the family upstairs. Although he didn’t make it, they survived after the firefighters forced their way to their apartment.

How many text messages have you received recently about missing delivery of a package you didn’t order? Or an award you won for being a loyal customer of a company you don’t use? Or a withdrawal that does not exist on your account?

Bogus messages like this one have skyrocketed in recent years as scammers switched from robocalls to robotexts — in part because the feds were forcing phone companies to Shutting down their networks to make automated calls. However, the legal landscape is changing in a way that makes it harder for scammers to invade your message queue as well.

It’s harder, no impossible. The scammers are a The notorious trick Package.

On Thursday, the Federal Communications Commission adopted a rule that requires cellphone carriers to block texts “highly likely to be illegal.” This includes texts from spoofed or non-working numbers, which spammers often rely on in their bulk messages.

The problem of automated text has grown exponentially. According to Robokiller, which makes spam-blocking technology, Americans received more than 225 billion unwanted text messages last year. Works on over 700 Every smartphone user.

The commission warned Thursday that scams are particularly dangerous, noting that humans are going through hard times no Read the texts received. In addition to promoting get-rich-quick schemes and other harassment, bot scripts can be used to trick people into revealing sensitive personal information or installing malware.

The new rules won’t go into effect for several weeks, and a follow-up batch of protections from the FCC is still awaiting public comment and a final vote. In the meantime, there are steps you can take to spare yourself the misery of hearing your phone’s text notification ring, only to find out it’s another fake FedEx notification.

The new rules

The commission applied some of the same techniques to spam scripts as it used to deter spam calls. Once the newly approved order comes into force, mobile networks will be required to create a “reasonable” list of not to create a list of numbers that will not be allowed to send text messages, similar to the list that should be kept for phone calls.

At a minimum, the “Do Not Create” text list will include numbers that are not valid or have not yet been set in North America. Individuals and organizations that spammers use their valid numbers to disguise the true source of their texts can also put their numbers on the list.

This kind of blanket approach targets a common practice among bots, which is to use different numbers (real or spoofed) to create phishing attempts or phishing attempts in succession. That’s why you can’t make much headway against spam by blocking individual numbers on your phone — texts keep coming from new unblocked numbers.

It’s also important for the FCC to require, not urge, cellphone carriers to create a list and block texts from those numbers. In its order, the commission said the shift was “partly due to the increased risks of text messaging as a nuisance and as a means of fraud”. “Data shows that consumers read almost all the texts they receive, and do so almost instantly. Indeed, industry data suggests that consumers open a much larger proportion of text messages than email, and open such messages more quickly. This contrasts with calls Where, as we have said time and time again, consumers report distrusting calls from an unfamiliar number and refusing to answer them.”

In the event that the new system interferes with valid texts, it requires mobile carriers to provide the public with a single point of contact for complaints about excessive blocking.

The goal is to stop scam scripts in the pipeline, rather than using hefty fines and wasteful enforcement action to try and deter botnets. Not that the committee has done much on the deterrence front. According to the FCC, the agency has only taken one enforcement action against spammers, Citation version in 2018 to a marketer who sent text messages to people on the federal Do Not Call register, which bans spam as well as phone calls. The recording is based on the year 1991 Consumer Protection Act over the phonewhich prohibits the use of automated calling equipment to call any residential number “without the prior express consent of the calling party”.

Two years ago the Supreme Court ruled that Much harder to enforce the lawjudgment in Facebook vs. Duguid The law prohibits robocalls only if they are made to randomly or sequentially chosen numbers. FCC Chair Jessica Rosenworcel told the House Oversight Subcommittee in e-mail in December.

“In light of these new legal and technical constraints, I believe the FCC will need to focus on preventing bot scripts in the first place, rather than simply trying to punish those responsible after the fact,” Rosenworcel wrote.

The last order will take effect 30 days after it is posted in the Federal Register, and it may be some time before that happens. So don’t expect mobile carriers to plug their networks with suspicious bot scripts right away.

More protection in the future?

The order adopted Thursday also proposes additional rules that would further crack down on bot scripts. One may block text messages to more than 240 million numbers in the Do Not Call history. The latter may prevent sites from tricking consumers into giving various unrelated companies permission to send them spam. And a third requires cell phone companies to block messages from senders that the FCC has identified as a source of illegal texts, just as they must do for robocallers that the commission has identified.

However, the commission did not propose a requirement that mobile phone companies authenticate the source of each text message and block those that are spoofed, as its rules require phone calls to be made. Along with blocking texts from invalid or inactive numbers, blocking scam numbers can be a huge hurdle for bots. In its arrangement, the commission said, the problem is that it’s not clear whether authentication methods that work for phone calls will work for text messages.

Instead of proposing a rule, the panel asked for comment on how a text message source authentication system might work, and how this technical solution could be applied to the bot text problem.

Another issue not explicitly addressed in the application is spam from email addresses. Mobile networks have email gateways that allow customers to exchange text messages with email addresses. So, naturally, developers offer tools that can send texts in bulk from email addresses — for free.

Already, some spammers generate phishing text messages from email addresses. If the FCC can close the door on spam sent from phone numbers, how long will it take the entire bot script industry to follow suit?

What can you do now

As noted by the FCC, many consumers have taken a heavy-handed approach to dealing with spam calls, automatically sending all calls from unknown numbers to voicemail. Legitimate callers will leave a message; Bots and scam call centers usually don’t.

If you have an Apple iPhone and use the company’s iMessage app, you can follow the same approach with text messages. In the Settings app, under Messages, you can select an option called Filter Unknown Senders. Doing so will put messages from numbers your phone doesn’t recognize in a separate list from those in your contacts.

On Android phones, Google Messages—which isn’t necessarily the default messaging app on a phone—offers a similar form of protection, but with more artificial intelligence. If you enable spam protection (under “Spam protection” in the settings menu in the app), you will use Google Analytics based on machine learning To review the texts received. If Google suspects a text is spam, it will move the message to a separate folder for spam and blocked texts, and then notify you that you received a message of sketchy origin. If it’s legitimate, you can remove the spam tag and send it to your plain text inbox.

These tools are free. Mobile carriers and app developers also offer Competition tools To block spam, some for a fee.

The most popular messaging apps also enable you to block and report numbers used by bots, but a one-by-one approach won’t stop scammers quickly going from number to number. It’s about as effective as trying to plug holes in a caulk that keeps showing new leaks.