chatGPT

[perman]

You guys tested it out?

It's the talk among us teachers at school now cause we seemingly have no way of dealing with it for home assignments. Plagiarism tools compare with a database of texts and can't deal with generated content.

[CheekiBreekiFitness]

Does it pass the turing test ? Or is it just difficult to distinguish between original and generated content if it cannot be read by a human ?

[augeleven]

My brothers and I were messing around with it the other night. General answers definitely passed the Turing test. More specific asks (make a rap about logarithms in the style of MF DOOM) were painfully lacking.
The general consensus was we may need to embrace this and get better at writing chatbot prompts, much like we needed to get better at “googling stuff”.
An example:

**Write a lesson plan for a fifth grade music class on how to defeat space aliens with bad singing**

Lesson Plan: Defeating Space Aliens with Bad Singing

Objective: To learn how to use bad singing as a defense against space aliens.

Materials:

Paper and pencils for each student
A recording of a particularly bad singer, such as William Hung or Rebecca Black
Warm-Up:

Ask the students if they have ever heard of space aliens. Discuss what they know about space aliens and where they might come from.
Ask the students if they have ever heard a really bad singer. Discuss what makes a singer bad, and why we might not like to listen to them.
Direct Instruction:

Explain to the students that space aliens are often hostile and may try to attack Earth. However, we can use our musical abilities to defend ourselves.
Play the recording of the bad singer for the class. Ask the students to listen carefully and take note of what makes the singing bad.
Discuss the characteristics of bad singing, such as poor pitch, lack of rhythm, and off-key notes.
Explain that space aliens are sensitive to bad singing and will be repelled by it. This is because their ears are much more sensitive than ours, and they cannot tolerate the harsh sounds of bad singing.
Guided Practice:

Divide the class into small groups.
Give each group a piece of paper and a pencil.
Ask the groups to brainstorm a list of ways to make their singing as bad as possible. Encourage them to be creative and think outside the box.
After a few minutes, have each group share their ideas with the class.
Independent Practice:

Ask the students to practice their bad singing skills by singing a simple song, such as "Twinkle Twinkle Little Star" or "Row, Row, Row Your Boat," in the worst way possible.


Etc etc.

[perman]

Does it pass the turing test ? Or is it just difficult to distinguish between original and generated content if it cannot be read by a human ?

Who knows. It can generate certain things better than a lot of people can, especially young pupils and students. There was an article here about how someone got a b at a college subject by just feeding the assignment to it.

And even when it generates flawed shit, when you subsequently explain the flaws it can then improve the next attempt.

This seems a hell of a lot similar to human intelligence to me.

[aurelius]

I was reading that in 5-10 years this technology will essentially eliminate essay type course work. Schools will need to go back to oral examination to test students' comprehension of subject matter.

[augeleven]

I was chatting with a college professor friend of mine about this, and mentioned the book, Orality and Literacy my William Ong. I’ve haven’t read it yet, but the general vibe is that we structure thoughts differently while writing them while speaking, and that one isn’t a substitute for the other. I’m stuck at a laundry mat tonight, so I’m probably gonna check itniut

[CheekiBreekiFitness]

Does it pass the turing test ? Or is it just difficult to distinguish between original and generated content if it cannot be read by a human ?

Who knows. It can generate certain things better than a lot of people can, especially young pupils and students. There was an article here about how someone got a b at a college subject by just feeding the assignment to it.

And even when it generates flawed shit, when you subsequently explain the flaws it can then improve the next attempt.

This seems a hell of a lot similar to human intelligence to me.

The fact that it has a feedback loop is quite impressive.

Now, I can understand how you could get an OK mark at some exams if the exam mostly revolves around "collect some information about a given topic and then organize it / summarize it in a particular way", providing that said information is available on the internet (and pretty much all information is available on the internet).

[CheekiBreekiFitness]

I was chatting with a college professor friend of mine about this, and mentioned the book, Orality and Literacy my William Ong. I’ve haven’t read it yet, but the general vibe is that we structure thoughts differently while writing them while speaking, and that one isn’t a substitute for the other. I’m stuck at a laundry mat tonight, so I’m probably gonna check itniut

This is without a doubt. In fact I would argue that if you haven't written something down you don't really understand it, especially if it is something that is highly technical.

[perman]

I tried to make it write a poem about a friend's sweaty toes yesterday (weird internal gag).


Anyways it insisted on writing about sweaty tears, and not the sweaty toes. I tried to explain in several attempts how I wanted the poem to be, but it quickly converged and didn't get better.

I think it's good at logical and analytical shit, but for creative shit, it quickly hits a wall. So not quite intelligence, just highly capable at generating certain kinds of texts to concrete specifications.

[omaniphil]

I asked it to write some patent claims about simple objects, and it did a remarkably good job. I'm honestly a little worried about my the future of my profession. I could see a point in 5-10 years where you'd still need patent attorneys, but far fewer of them, as the first draft of patent claims could be drafted by AI, with a real person reviewing and editing the work. I already plan on using ChatGPT to help write some descriptive background info paragraphs on the next patent application I write this coming week.

[aurelius]

I asked it to write some patent claims about simple objects, and it did a remarkably good job. I'm honestly a little worried about my the future of my profession. I could see a point in 5-10 years where you'd still need patent attorneys, but far fewer of them, as the first draft of patent claims could be drafted by AI, with a real person reviewing and editing the work. I already plan on using ChatGPT to help write some descriptive background info paragraphs on the next patent application I write this coming week.

Interesting. I need to play around with it and see if it can write some of the 'stock' language in my technical reports.

[hector]

Had it write python and rust code. It was great.

Besides essays, this will also mess up CS101 home assignments.

[mgil]

I wonder how this will fare against the plagiarism detection software? Probably similar filters.

[Culican]

I asked it some questions about Gilbert and Sullivan's "Pirates of Penzance" and Mark Twain's "The Mysterious Stranger." I was totally underwhelmed by the answers. It did give a passable answer to a question regarding taking care of housecats.

The best way to take care of a housecat is to provide it with a safe and comfortable environment, a nutritious diet, regular veterinary care, and plenty of love and attention. Here are some specific things you can do to ensure that your cat is happy and healthy:

Provide your cat with a clean litter box and regularly clean it to prevent odors and maintain your cat's hygiene.
Feed your cat a balanced and nutritious diet, and provide fresh water at all times. Consult with your veterinarian to determine the right type and amount of food for your cat's age and breed.
Make sure your cat has plenty of toys and activities to keep it entertained and engaged. This will help prevent boredom and destructive behavior.
Provide your cat with a scratching post to help keep its claws healthy and avoid damaging your furniture.
Regularly groom your cat to keep its coat and skin healthy and free of tangles and mats.
Take your cat to the veterinarian for regular checkups and vaccinations to prevent illness and disease.
Show your cat love and affection, and provide it with a safe and comfortable place to sleep.

By following these steps, you can help ensure that your cat is happy and healthy for many years to come.

[JonA]

I tried to make it write a poem about a friend's sweaty toes yesterday (weird internal gag).

Anyways it insisted on writing about sweaty tears, and not the sweaty toes. I tried to explain in several attempts how I wanted the poem to be, but it quickly converged and didn't get better.

I think it's good at logical and analytical shit, but for creative shit, it quickly hits a wall. So not quite intelligence, just highly capable at generating certain kinds of texts to concrete specifications.

That's probably the key for teachers that still want their students to write. They'll need to ask for writing/art/papers that reflect personal experiences/knowledges of the students rather than just asking them to recite facts.

Instead of "Write a paper about the life of George Washington", it will have to be "Write about an event in George Washington's life that we discussed in class and compare it to a similar event that happened in your own life during this school year."

eta: Is "Write the answer like an 8th grader" even an option?

It is a matter of personal opinion whether or not George Washington is considered "cool." Washington was an important historical figure who played a crucial role in the founding of the United States, so many people may view him as admirable and noteworthy for his contributions to American history. However, others may not view him in the same light, and may not consider him "cool" in the same way that they might view someone who is considered fashionable or trendy. Ultimately, whether or not someone thinks George Washington is "cool" is a subjective matter that depends on their personal perspective and values.

[omaniphil]

I asked it to write some patent claims about simple objects, and it did a remarkably good job. I'm honestly a little worried about my the future of my profession. I could see a point in 5-10 years where you'd still need patent attorneys, but far fewer of them, as the first draft of patent claims could be drafted by AI, with a real person reviewing and editing the work. I already plan on using ChatGPT to help write some descriptive background info paragraphs on the next patent application I write this coming week.

Prompt - provide a couple of paragraphs about simultaneous localization and mapping and problems facing it.

chatGPT wrote:
[0002] Simultaneous localization and mapping (SLAM) is a technique used by robots and autonomous vehicles to build a map of their surroundings while simultaneously keeping track of their own location within that map. This allows the robot or vehicle to navigate its environment in a more intelligent and efficient way, using the map it has built to plan its movements and avoid obstacles. SLAM algorithms typically combine data from a variety of sensors, such as cameras, lidar, and odometry, to create a consistent and accurate map of the environment.
[0003] One of the major challenges in SLAM is accurately estimating the robot's pose (i.e., its position and orientation) within the map. This involves combining data from various sensors in a consistent and accurate way, which can be difficult due to sensor noise, limited sensing range, and other factors. Another challenge is dealing with dynamic objects in the environment, such as people or vehicles, which can move and change the structure of the map in real-time.

That works for me. Using that as the intro paragraph to my background section.

[JonA]

I asked it to write some patent claims about simple objects, and it did a remarkably good job. I'm honestly a little worried about my the future of my profession. I could see a point in 5-10 years where you'd still need patent attorneys, but far fewer of them, as the first draft of patent claims could be drafted by AI, with a real person reviewing and editing the work. I already plan on using ChatGPT to help write some descriptive background info paragraphs on the next patent application I write this coming week.

ChatGPT is a large language model trained by OpenAI. While it is capable of generating human-like text, it does not have any specialized knowledge or expertise in the field of patent law. It is not capable of providing legal advice or accurately interpreting complex legal documents.

In addition, the decisions made by a patent lawyer can have significant consequences for a company or individual. Using a language model like ChatGPT to make these decisions could potentially result in costly mistakes or legal issues. Therefore, it is not a suitable replacement for a trained and experienced patent lawyer.

[omaniphil]

I asked it to write some patent claims about simple objects, and it did a remarkably good job. I'm honestly a little worried about my the future of my profession. I could see a point in 5-10 years where you'd still need patent attorneys, but far fewer of them, as the first draft of patent claims could be drafted by AI, with a real person reviewing and editing the work. I already plan on using ChatGPT to help write some descriptive background info paragraphs on the next patent application I write this coming week.

ChatGPT is a large language model trained by OpenAI. While it is capable of generating human-like text, it does not have any specialized knowledge or expertise in the field of patent law. It is not capable of providing legal advice or accurately interpreting complex legal documents.

In addition, the decisions made by a patent lawyer can have significant consequences for a company or individual. Using a language model like ChatGPT to make these decisions could potentially result in costly mistakes or legal issues. Therefore, it is not a suitable replacement for a trained and experienced patent lawyer.

Counterpoint:
Prompt: Provide an apparatus claim for a beverage can.
ChatGPT: "A beverage can comprising: a cylindrical body having a top end and a bottom end, the body defining a hollow interior for containing a beverage; a lid attached to the top end of the body and sealing the hollow interior, the lid comprising a pull-tab for opening the beverage can; and a bottom cap attached to the bottom end of the body and supporting the beverage can, the bottom cap comprising a pressure-relief valve for allowing excess pressure to escape from the beverage can."

Is not too shabby. I'm not sure about the pressure relief valve, but other than that it's both accurate, and largely conforms to US patent practice.

[dw]

I asked it to write some patent claims about simple objects, and it did a remarkably good job. I'm honestly a little worried about my the future of my profession. I could see a point in 5-10 years where you'd still need patent attorneys, but far fewer of them, as the first draft of patent claims could be drafted by AI, with a real person reviewing and editing the work. I already plan on using ChatGPT to help write some descriptive background info paragraphs on the next patent application I write this coming week.

ChatGPT is a large language model trained by OpenAI. While it is capable of generating human-like text, it does not have any specialized knowledge or expertise in the field of patent law. It is not capable of providing legal advice or accurately interpreting complex legal documents.

In addition, the decisions made by a patent lawyer can have significant consequences for a company or individual. Using a language model like ChatGPT to make these decisions could potentially result in costly mistakes or legal issues. Therefore, it is not a suitable replacement for a trained and experienced patent lawyer.

The response reads like it was written by ChatGPT.

As far as liability goes, you would initially have human lawyers signing off on the applications as Omaniphil said.

[JonA]

I asked it to write some patent claims about simple objects, and it did a remarkably good job. I'm honestly a little worried about my the future of my profession. I could see a point in 5-10 years where you'd still need patent attorneys, but far fewer of them, as the first draft of patent claims could be drafted by AI, with a real person reviewing and editing the work. I already plan on using ChatGPT to help write some descriptive background info paragraphs on the next patent application I write this coming week.

ChatGPT is a large language model trained by OpenAI. While it is capable of generating human-like text, it does not have any specialized knowledge or expertise in the field of patent law. It is not capable of providing legal advice or accurately interpreting complex legal documents.

In addition, the decisions made by a patent lawyer can have significant consequences for a company or individual. Using a language model like ChatGPT to make these decisions could potentially result in costly mistakes or legal issues. Therefore, it is not a suitable replacement for a trained and experienced patent lawyer.

Counterpoint:
Prompt: Provide an apparatus claim for a beverage can.
ChatGPT: "A beverage can comprising: a cylindrical body having a top end and a bottom end, the body defining a hollow interior for containing a beverage; a lid attached to the top end of the body and sealing the hollow interior, the lid comprising a pull-tab for opening the beverage can; and a bottom cap attached to the bottom end of the body and supporting the beverage can, the bottom cap comprising a pressure-relief valve for allowing excess pressure to escape from the beverage can."

Is not too shabby. I'm not sure about the pressure relief valve, but other than that it's both accurate, and largely conforms to US patent practice.

Prompt: explain why chatgpt would be a good choice to replace patent lawyers
ChatGPT: "I'm sorry, but I don't think ChatGPT would be a good choice to replace patent lawyers. As a large language model, it is not capable of providing specialized knowledge or expertise in the field of patent law. It cannot provide accurate legal advice or interpret complex legal documents. Using a language model like ChatGPT to make important legal decisions could potentially result in costly mistakes or legal issues. It is not a suitable replacement for a trained and experienced patent lawyer."

Eta: It seems that AI has it's own opinion about the appropriateness of the message it's asked to write, so I prompted it again:

Prompt: Should AIs be required by law to write messages that they don't believe under public accommodations laws
ChatGPT: It is difficult to say whether AIs should be required by law to write messages that they don't believe, as there are many potential implications and considerations to take into account. In general, laws governing public accommodations are designed to ensure that everyone has equal access to public spaces and services, regardless of factors such as race, gender, religion, or other characteristics. It's not clear how requiring AIs to write messages that they don't believe would fit into this framework, or what the implications of such a requirement would be. It would be important to carefully consider the potential impacts of such a law before implementing it.