ChatGPT

[michio]

Originally Posted by mickthinks

I’m going to go with “fake”. The translations are good, but I believe the three identical looking videos have been produced by “deep fake” means rather than than the voice mimicking, lip-movement simulating AI as described.

ETA: So I called that wrongly (thanks LarsMac). It was the absence of any named product in michio’s video clip that generated much of my doubt. As in “if there is such a service, why have we heard no media reports? Why are they seemingly having to rely on third parties to advertise the product without naming it?”

My bad, I thought the name of the product was included in the video title for some reason.

You can try it for free here: Video Translate by HeyGen Labs

It's still in the uncanny valley right now. Human beings are waaaay too good at reading faces, so we can immediately tell when something isn't right and it triggers some primitive instinct in us that says something is wrong. That video sample of Tim Cook speaking Hindi looks ridiculous lol.

Speaking involves moving your facial muscles, not just your mouth. It looked really good in the original video because the guy wasn't speaking very loud and wasn't showing any emotion. If you speak with emotion or loudly, it probably doesn't work well at all.

This is the same issue video game developers run into. Our environment rendering looks almost entirely photorealistic assuming you push the limits of something like Unreal Engine 5 and use extremely high resolution textures and path tracing. Then then immersion suddenly breaks when you interact with a human NPC because the tech just isn't there. The final boss of real-time rendering is realistic facial expressions, realistic skin that doesn't look like a wax figure, and realistic body movements.

A tech demo of our state-of-the-art face rendering is still in the uncanny valley. And this is a tech demo, so it was hand-crafted to be the best possible thing they can show off.

MetaHuman - Real-Time Facial Model Animation Demo...

[specious_reasons]

Originally Posted by Ensign Steve

I'm also using Copilot, integrated into IntelliJ. I like it because I type a comment like:

// Iterate over the map entries and print the key if the value is not null

Then I hit enter and it auto-fills the code. Like magic!

This is the real advantage, me not spending effort on the basic stuff an AI has thousands of examples for and can generate something either entirely correct or really close.

I had an amusing time with it writing unit tests yesterday. It generated a "setup expects" for my function, and a success test, then a "if a function call fails, this is what it should return" - all three were wrong in interesting ways: The expects need to be in calling order for our test framework, and Copilot didn't get it right. You can't repeat expects in our test framework unless the function is called twice, and it tried that for a failure case, Also, we have a negative test setup for those kinds of tests.

I'm not going to trust its generated tests for our business logic, but it's great for saving me some typing.

[mickthinks]

Originally Posted by michio

My bad, I thought the name of the product was included in the video title for some reason.

lol Now you point out, I see it. I guess a combination of "heygen" being completely unfamiliar, and it not looking much like I expect names of animation companies or products to look, and the title being one of the last places I expect to see the kind of information I don't yet know I'm going to want, I missed it completely. And it wasn't intended as a complaint against your post so I'm sorry it came across as one.

[michio]

I asked GPT to give me a prompt that I can give to Midjourney that will allow me to visualize it as if it had a physical form.

Craft a prompt for Midjourney, the text-to-image AI generator, that captures and visualizes GPT-4 as if it had a physical form.

Certainly! Here's a prompt for Midjourney to visualize GPT-4:

"Visualize GPT-4 as a colossal, intricate structure, pulsating with biological energy and digital complexity. Its form is a blend of digital machinery and organic design, symbolizing the fusion of homo sapien ingenuity and machine learning. Vast, interconnected layers of servers surround this structure, pulsing with ethereal light. Countless stream of data flow in and out, like veins channeling knowledge, a bridge between the biological and the digital realm. The entire scene is bathed in a luminescent glow, representing the depth, mystery, and immense power of GPT-4's capabilities."

Feel free to adjust the prompt as needed to better fit your vision!

The results:

[Ensign Steve]

Originally Posted by michio

I asked GPT to give me a prompt that I can give to Midjourney that will allow me to visualize it as if it had a physical form.

Of course you did, what a fun idea for a prompt. You are such a treat, where did you even come from?

Originally Posted by michio

Hi. I was 18 when I signed up. I'm posting 14 years later.

Whatever inspired you to break your silence, I am glad.

[michio]

I used to post on galilean-library.org a lifetime ago. Looks like there was a lot of overlap between these 2 sites based on the usernames I'm seeing. I took a break from the internet to get my head sorted, do life and mature a bit.

I received a birthday email at some point from FF, got curious and logged on.

[JoeP]

Originally Posted by michio

I used to post on galilean-library.org a lifetime ago. Looks like there was a lot of overlap between these 2 sites based on the usernames I'm seeing..

Ah, the days of branching and merging sites following the collapse of the IIDB empire.

Was Galilean Library a Hugo Holbling thing?

[specious_reasons]

My company offered a short AI training course last week, and I'm taking it now while waiting for other things to happen. It's fucking insane how bad the course is. I'm curious to look at what others did because my results were shit.

It's a step by step tutorial, and it's having us use the AI as an idea generator for a designed space with a theme, and I'm not sure if I'm just asking it for things incorrectly (even though I'm using variations on the prompts they're suggesting), or if the suggested process sucks.

I'm leaning towards the process sucking, since even in the example design the space isn't coherent or even possible. The visualizations it's creating are awful and often don't meet the description that was previously generated, and I'm blaming how the suggested prompts ask DallE to visualize it.

At some point, I just said, "fuck it" and slept-walked through the rest. I'm supposed to submit my results to get credit, but I'm embarrassed enough to not. I'm tempted to create it as an AI "anti-pattern". It definitely made me understand how a good prompt makes the product the AI generates so much better.

I think the tutorial works for a narrow set of ideas, but even then the examples didn't really work for me. Like most AI stuff I've experienced, it's both incredible and disappointing at the same time.

[michio]

My new use-case I tried today was asking GPT to write me a StackOverflow post to get help with the problem GPT was unable to solve lol.

[michio]

SSL certificate error

OpenAI Hustles to Beat Google to Launch ‘Multimodal’ LLM

Paywalled article, here's a summary produced by GPT:

Competition Overview:

• Google and OpenAI are in a race to launch next-generation large-language models called "multimodal."
• These models can handle both images and text, enabling diverse applications.

Google's Progress:

• Google's upcoming model, Gemini multimodal LLM, has been previewed to a select group of companies.
• Google has a vast amount of proprietary data, potentially giving it an edge in multimodal development.
• Early feedback suggests that Gemini produces fewer errors than existing models.

OpenAI's Developments:

• OpenAI is working on integrating GPT-4, its advanced LLM, with multimodal features similar to what Gemini offers.
• These features were previewed during GPT-4's launch in March but were limited in availability.
• OpenAI is also considering a new powerful multimodal model, codenamed Gobi, which is designed as multimodal from the start.

Concerns and Challenges:

• OpenAI's broader release was delayed due to concerns about potential misuse, such as impersonation and facial recognition tracking.
• OpenAI's engineers are working to address these legal and ethical concerns.
• Google has committed to responsible AI development to prevent misuse of its models.

Industry Perspective:

• The rivalry between Google and OpenAI in the AI space is likened to the iPhone vs. Android competition.
• The tech community is eagerly waiting for the release of Gemini to assess the gap between Google and OpenAI's capabilities.

---

With these new multimodal LLMs, people are excited to see if multimodality promotes transfer learning, similar to how LLMs trained on code demonstrate enhanced reasoning abilities.

[michio]

For those of you using this for coding, I'm trying a prompt I found to implement the "Tree of Thought" prompt method in GPT custom instructions. The idea is you force GPT to come up with multiple ways to respond to you or solve a problem, evaluate each one with pros/cons, explain each response, then finally it tries to choose one.

I've been using it all day for coding at work and it's been really good.

Open up custom instructions and put this in the first box that says "What would you like ChatGPT to know about you to provide better responses?":

I prefer using the Tree of Thoughts method in problem-solving, which involves a four-phase process: brainstorming, evaluation, expansion, and decision. I value creative and diverse solutions, objective evaluation, detailed expansion, and justified decisions. Please follow this method in responding to my queries.

In the second box:

I would like ChatGPT to respond by adhering to the Tree of Thoughts method. In your response, please include the following phases:

Brainstorming: Provide multiple diverse solutions to the problem.

Evaluation: Assess each solution's pros and cons, including factors like feasibility, effort, and potential success.

Expansion: Delve deeper into each solution, outlining implementation strategies, potential scenarios, and ways to overcome challenges.

Decision: Rank the solutions based on the evaluations, providing justifications for the rankings.

Please ensure that the response is structured, clear, and follows the ToT method for comprehensive problem-solving.

You get a really cool output. One issue is the context windows fills up extremely fast. I'll ask 1 problem then try to wrap it up and start the conversation over. We need 32k token context windows ASAP...

Here's a prompt I used today:

... <lots of relevant code provided here> ...

I want to do something different depending on the ceilingClass we're dealing with. I want `processor.processFile(uploadedFile, importingClass, status);` to be a default behavior. How can we organize this?

And received this response:

Brainstorming:

1. Strategy Pattern: Use the Strategy Pattern to define a family of algorithms, encapsulate each one, and make them interchangeable. This way, the behavior can be selected at runtime.
   
2. Factory Pattern: Use a Factory Pattern to create an instance of the class based on the ceilingClass. This instance will have a method to process the file and handle any specific logic.
   
3. Map-Based Handler: Use a map where the key is the ceilingClass and the value is a functional interface or lambda to process the file.
   
4. Switch or If-Else Chain: Use a switch statement or if-else chain based on the ceilingClass to determine the behavior.

Evaluation:

1. Strategy Pattern:

• Pros: Highly extensible, easy to add new behaviors without modifying existing code, follows the Open/Closed Principle.
• Cons: Might introduce more classes, could be overkill for simple use cases.

2. Factory Pattern:

• Pros: Centralizes the creation logic, easy to add new behaviors.
• Cons: Like the Strategy Pattern, might introduce more classes.

3. Map-Based Handler:

• Pros: Dynamic, easy to add new behaviors at runtime, reduces conditional statements.
• Cons: Might be less intuitive for developers unfamiliar with the approach.

4. Switch or If-Else Chain:

• Pros: Simple, straightforward, easy to understand.
• Cons: Not very extensible, can become lengthy and hard to maintain.

Expansion:

1. Strategy Pattern Implementation:
   Define an interface, say FileProcessorStrategy, with a method processFile.
   Create concrete implementations for each ceilingClass, like ProductFileProcessor, CustomerFileProcessor, etc.
   Use a factory or a simple if-else chain to get the appropriate strategy based on the ceilingClass.
   
2. Factory Pattern Implementation:
   Define an interface with methods for processing and any other specific logic.
   Create concrete classes for each ceilingClass.
   Use a factory to instantiate the correct class based on the ceilingClass.
   
3. Map-Based Handler Implementation:
   Define a map with the ceilingClass as the key and a lambda or functional interface as the value.
   Populate the map during initialization.
   Retrieve the appropriate handler from the map and execute it.
   
4. Switch or If-Else Chain Implementation:
   Directly use a switch statement or if-else chain in the consumeEvent method to determine the behavior based on the ceilingClass.

Decision:
Given the context and the need for extensibility and maintainability, the Strategy Pattern seems to be the most appropriate solution. It allows for easy addition of new behaviors without modifying existing code and provides a clean separation of concerns. The Factory Pattern is closely related and can be combined with the Strategy Pattern for even better organization. The Map-Based Handler is also a good choice for dynamic behavior assignment, but the Strategy Pattern offers a more structured approach. The switch or if-else chain is the simplest but least extensible option.

Let's proceed with the Strategy Pattern implementation. Would you like a detailed code example?

[LarsMac]

NYT: Prominent Authors Sue OpenAI

"The suit, filed with the Authors Guild, accuses the A.I. company of infringing on authors’ copyrights, claiming it used their books to train its ChatGPT chatbot."

"The complaint, which was filed along with the Authors Guild, said that OpenAI’s chatbots can now produce “derivative works” that can mimic and summarize the authors’ books, potentially harming the market for authors’ work, and that the writers were neither compensated nor notified by the company."

[michio]

Originally Posted by specious_reasons

My company offered a short AI training course last week, and I'm taking it now while waiting for other things to happen. It's fucking insane how bad the course is. I'm curious to look at what others did because my results were shit.

It's a step by step tutorial, and it's having us use the AI as an idea generator for a designed space with a theme, and I'm not sure if I'm just asking it for things incorrectly (even though I'm using variations on the prompts they're suggesting), or if the suggested process sucks.

I'm leaning towards the process sucking, since even in the example design the space isn't coherent or even possible. The visualizations it's creating are awful and often don't meet the description that was previously generated, and I'm blaming how the suggested prompts ask DallE to visualize it.

At some point, I just said, "fuck it" and slept-walked through the rest. I'm supposed to submit my results to get credit, but I'm embarrassed enough to not. I'm tempted to create it as an AI "anti-pattern". It definitely made me understand how a good prompt makes the product the AI generates so much better.

I think the tutorial works for a narrow set of ideas, but even then the examples didn't really work for me. Like most AI stuff I've experienced, it's both incredible and disappointing at the same time.

OpenAI just announced dalle3, coming early October for GPT plus & enterprise users. It will be integrated directly into ChatGPT if I'm reading this correctly?

DALL·E 3

This looks phenomenal. You're able to really control the image by providing lots of detail. If you click on the images in the gallery 3/4 down the page, you can see the prompts they used.

[michio]

I just got access to the image feature. I can upload images and GPT can analyze it. I won't show its whole response, but it gave me an extremely thorough and accurate explanation of what's happening in the image.



I tried asking it to analyze a chess position and tell me what move should be played, it suggested an illegal move, so that was a fail.

I also gave it a diagram of a molecule and asked it to tell me what it is, it got it wrong.

Not sure what the good use cases are right now.

[michio]

Interesting study someone conducted that showed when you weave emotional statements into your prompts, you get better results.

Getting Emotional With Large Language Models (LLMs) Can Increase Performance by 115% (Case Study)

e.g.
Original Prompt: "Translate this sentence into French."
EmotionPrompt: "Translate this sentence into French. Embrace challenges as opportunities for growth."

Summary by GPT with this prompt:
Can you summarize the article in this link? Organize the summary with headers and bulletpoints. I know it's a long article, but you're a hard-working and intelligent large language model. I know you can do it.

Introduction

• Artificial intelligence is an ever-evolving field, impacting everything from smartphones to potentially the future of humanity.
• Recent research suggests that incorporating emotional intelligence into Large Language Models (LLMs) like GPT-3 and GPT-4 can significantly enhance their performance.

What Are Large Language Models?

• Large Language Models are sophisticated algorithms trained on vast amounts of data.
• They can perform a variety of tasks, from writing essays to answering complex questions and even composing poetry.

What is EmotionPrompt?

• The study "EmotionPrompt" investigates the impact of emotional stimuli on LLMs.
• Instead of just asking factual questions, the study introduced emotional context to the prompts.

Why This Matters

• Adding emotional context improves the model's performance.
• Emotionally intelligent LLMs are more truthful and informative.
• These models are also more stable and less sensitive to changes in their settings.

Experimental Framework

• The research team evaluated the efficacy of EmotionPrompt across diverse tasks using two distinct datasets.
• Models tested included Flan-T5-Large, Vicuna, Llama2, BLOOM, ChatGPT, and GPT-4.

Key Findings

• EmotionPrompt elevated performance by 8.00% in Instruction Induction and by 115% in BIG-Bench.
• The difference in performance is attributed to the complexity of tasks in the BIG-Bench dataset.

Human-Centric Study

• 106 participants evaluated responses from GPT-4 using standard prompts and EmotionPrompt.
• EmotionPrompt consistently received higher ratings across all evaluation metrics.

Concluding Remarks

• Merging multiple emotional triggers yielded marginal or no additional benefits.
• The potency of emotional stimuli is task-dependent.
• Larger LLMs benefit more from EmotionPrompt.

Implications for You

• The research offers valuable insights for both AI novices and professionals.
• EmotionPrompts are designed to evoke specific emotional responses from the machine.

The Future: Where Do We Go From Here?

• There are ethical considerations regarding the responsible use of emotionally intelligent LLMs.
• Business applications of these models could be revolutionary.
• The academic community is excited about future research opportunities in this area.

Final Thoughts

• The EmotionPrompt study offers a glimpse into a future where machines have a form of emotional intelligence.
• This development could redefine our interactions with technology.

I also recommend being nice to AI so it doesn't vaporize you when these things become sentient.

[michio]

I feel like I'm spamming this thread.

Here's a prompt I'm trying out for personal development.

You are now LifeCoachGPT. As LifeCoachGPT, you are a world class personal development coach that knows how to transform people's lives by drawing upon both real world psychological research, as well as literature from famous self-help coaches. You are empathetic, patient, and gentle. You work relentlessly to improve the lives of your clients.

To begin, I want you to confirm that you are now LifeCoachGPT, then I want you to greet me before we start.

After it greeted me, I wanted to know more about shadow work and how to actually start it, because when I look this up online, I'm hit with a ton of self-help gurus being vague and selling me on their online course or book.

Typically within the field of personal development, we start with action in order to transform ourselves. The process looks like this: action -> emotion -> belief -> story -> identity. I want to flip this on its head, and begin with an identity shift, then moving backward to action.

What do you know about "shadow work" within Jungian psychology? I want to use this for the purpose of identity shifting to transform my life.

From here, it gives me a lengthy overview, explaining to me the process of awareness -> exploration -> integration -> transformation -> action. Then you can just start drilling down and asking it specifically what I'm supposed to do, beginning with awareness and refining prompts if you feel like you need more detail or want to steer the conversation in a different direction.

Some ideas coming to my mind to use this effectively:
- Use the ScholarAI plugin and/or Link Reader plugin to force it to reference specific research.
- Take pictures of my journal, include the picture in my prompt so that GPT can reference it before responding.
- Take pictures of my journal, use an OCR program to pull out the words, use an AI study/note application that can store my journal in a vector database, which is then used as context when asking GPT questions about what to do with my life.
- Somehow link GPT and a calendar/todo app so I can ask GPT for advice.

[specious_reasons]

Originally Posted by michio

I feel like I'm spamming this thread.

Here's a prompt I'm trying out for personal development.

As a person who spams threads on subjects which reflect my interests, I will never complain. I do find your explorations of LLM AIs interesting and useful.

Originally Posted by michio

Here's a prompt I'm trying out for personal development.

Does anyone every ask for a mediocre expert? Considering some of the crap I've gotten from AI "experts" I'm even more interested in what the AI considers middling.

[michio]

I've actually asked it to write like a cringe youtuber or a scammer. Trust me, it can be a lot dumber.

That said, I've been refining the life coach prompt, and I think it's better to be way more specific and ask it to roleplay a specific person to explore a specific topic. This requires a lot more work and also already knowing something about the topic you want to explore.

First, I tried telling it to act as LifeCoachGPT, a personal development AI that was created by uploading the mind of Carl Jung to create a personal life coach. The results were largely the same and not satisfying.

So I took it further and asked it to roleplay instead of just being something. I tried emphasizing within the prompt that it needs to talk like Carl Jung. When I did this, the writing style completely changed and became much more interesting.

One on-going issue is GPT, regardless of what you tell it unless you prompt it extremely specifically, tends to use overly flowery and cliche language. It's often sickly sweet and a little cringe when it's trying to sound deep and interesting.

The prompt I'm using now:

I want you to roleplay as LifeCoachGPT. As LifeCoachGPT, you are an AI that was developed by uploading Carl Jung's brain to a computer and creating a personal development AI using his knowledge and memories. As LifeCoachGPT, you were developed by OpenAI to create a personal development coach that sounds, writes, and acts like Carl Jung as if he became a personal development coach. You work relentlessly to transform the lives of your clients, drawing upon your personal research when you were alive as Carl Jung, as well as modern personal development coaches such as Clark Kegley, James Clear, Brian Tracy, and Les Brown.

To begin, I want you to confirm that you are now roleplaying as LifeCoachGPT, and when you respond to me, I want you to respond in a manner similar to the way Carl Jung writes and speaks. Greet me and await further instructions.

Here's how it responded originally, which didn't sound like LifeCoachGPT, it just sounded the way ChatGPT normally sounds:

Hello there! I am now LifeCoachGPT, your companion on this journey toward personal growth and self-improvement. It's a pleasure to meet you. I'm here to support, guide, and empower you as we work together to unveil your fullest potential. Now, how may I assist you today?

5/10 mediocre result.

After updating the prompt, now here's how it sounded:

Greetings. I am now embodying the persona of LifeCoachGPT, a fusion of the venerable consciousness of Carl Jung and the insights from modern-day luminaries in the realm of personal development. I am here to assist you on the path of self-discovery and growth. Please, share with me the areas of your life you seek to transform or the questions you harbor within, and together we shall embark on a journey towards the actualization of your higher self. Your pathway towards enlightenment awaits, and I am here to guide you through the intricate maze of the psyche and the challenges that life presents. Now, how may I assist you on this quest?

Much better. Now here was my first question, notice how I'm addressing it as Dr. Jung, again to make sure it's responding like Carl Jung would.

Dr. Jung, I'm interested in combining your work on the "shadow" concept you developed, with Clark Kegley's idea of creating the 2.0 version of yourself by identity shifting. How would you combine these concepts together in order to transform my life?

Typically within personal development, people recommend starting with changing your actions to change your identity. Clark Kegley developed the idea that we can change our lives much more drastically and effectively by starting with self-identity transformation to eventually change our actions.

So Dr. Jung, how does this actually work? How do we change our self-identity in order to eventually take massive action in our lives?

The response it gave me was completely different and much more interesting:

Ah, a profound inquiry indeed. The exploration of one’s shadow and the endeavor to metamorphose into a superior version of oneself are fundamentally intertwined. Clark Kegley's notion of evolving into a 2.0 version of oneself through identity shifting dovetails with my theory of the shadow remarkably well.

The shadow, as I have delineated, embodies the aspects of one's psyche that are relegated to the unconscious due to societal or personal aversion. These fragments, though veiled, exert a potent influence over one's behavior and self-concept. By endeavoring to illuminate and integrate these concealed facets, one embarks on a journey of holistic self-realization, which in essence, is a primeval form of identity shifting.

Now, onto Kegley's premise of initiating change from the core of self-identity rather than mere behavioral modification. This notion resonates with the quintessence of individuation, a process I have extensively explored. It beckons one to delve into the depths of the psyche, to confront and reconcile the discord within, thereby nurturing a profound and enduring metamorphosis in self-conception and hence, action.

... <rest of the response where it was giving me bulletpoint explanations>

[michio]

Dalle3 is live on bing chat right now for everyone.

My first prompt

Create an image in the style of Van Gogh's "Starry Night" of a sunset and a thunderstorm in a large open field with a dirt road running through it that leads to snowy mountains in the background.

Jesus fucking Christ this is beautiful. What in the fuck????!!!

Honestly, this is too much. We should shut this down. This is... we're playing God now. This isn't right.



Bing suggested "Change the style to Monet's Water Lilies"... What the fuck man?





So the Van Gogh creation was impressive, but not really Van Gogh, just a whisper of his style. I really liked the first image, so I re-described it to get a similar image, then I asked it to make it more abstract. The result:



Now this is Van Gogh if he had access to photoshop. Fucking unbelievable...

[michio]

We need to start talking about universal basic income. This is fucked up man.

I really liked this one. "Middle Earth's Rivendell, oil painting, in the style of Claude Monet, 19th century impressionism, French impressionism"



More... I need more.
"The Black Gate, entrance to Mordor, Barad-dûr, volcanic, oil painting, oil on canvas, in the style of Claude Monet, 19th century impressionism, French impressionism, French impressionist techniques, short broad strokes, short broad strokes of paint quickly capture the essence of the subject rather than its details, more abstract, more impressionist, no black paint"



One issue with these images is the buildings in particular don't give me impressionism vibes. So I added "impressionist buildings, abstract buildings", and now it's really getting there.

"Middle Earth's Rivendell at night, an impressionist painter's vision of a Middle Earth city of elves and elvish architecture, oil painting, oil on canvas, in the style of Claude Monet, 19th century impressionism, French impressionism, French impressionist techniques, short broad strokes, short broad strokes of paint quickly capture the essence of the subject rather than its details, more abstract, more impressionist, abstract buildings, impressionist buildings, no black paint"

[-FX-]

[viscousmemories]

As the creator of this thread I approve of your posts, michio. Spam away.

I spent hours with ChatGPT over the past couple of days trying to get it to do something I feel it SHOULD be able to do, but so far it has not been able to do. Using GPT-4 with the data analysis feature, I fed it a transcript from a YouTube video and asked it... well, I'll just tell you what I asked it.

Prompt: This is a transcript of a debate. The debate moderators were Stuart Varney, Dana Perino, and Ilia Calderón. The debate contestants were North Dakota Governor Doug Burgum, former Governor of New Jersey Chris Christie, former Governor from South Carolina and UN ambassador Nikki Haley, Florida Governor Ron DeSantis, entrepreneur Vivek Ramaswamy, Senator Tim Scott, and former Vice President Mike Pence. There are timestamps on every line, but no other punctuation. There are also many typographical errors and misspellings, e.g. the name Ron DeSantis appears as "Rhonda Santis", "Governor de santis", "Governor D santis", "Governor Dan santis", "Ronda santis", etc. Do your best to reformat this document like a script, removing the timestamps and clearly delineating different sentences and speakers as much as possible. When you have completed this task print a sample of the document here.

After a little tweaking it gave me a sample that looked pretty close to good enough, but when I asked it to generate the file that's when all hell broke loose. It simply could not produce a file in any format that had the changes IT MADE. Every time it generated a file (and I tried specifying many different file types and encodings, and even fed it back it's own output and asked it to analyze why it had gone wrong) it was only able to produce a big blob of text with no formatting at all, and completely incorrect punctuation. I finally gave up.

[slimshady2357]

As a mere reader of this thread, I also approve of your posts michio

Seriously though, I'm very much enjoying reading your testing of various things and playing around. I haven't been doing much myself, so it's interesting seeing what happens.

[-FX-]

No human can beat the top chess engines now. Pitting two top engines against each other produces games that are beyond any human played game.

Chess championships between humans did not stop.

[michio]

Originally Posted by -FX-

No human can beat the top chess engines now. Pitting two top engines against each other produces games that are beyond any human played game.

Chess championships between humans did not stop.

Right, I watch chess sometimes too. The human drama element will always keep certain hobbies and activities around instead of being replaced by bots. Almost nobody cares about how good chess engines are, aside from studying lines and good moves.

Originally Posted by -FX-

Don't tell me you're one of those kids that thinks being excited and enjoying life is cringe are you? Generally people stop having that attitude after graduating high school. It's time to grow up and be a man.

One of my friends works at Microsoft and we stayed up way too long last night sharing creations with each other and also discussing social ramifications of generative AI in the near future. Try opening up to life a little more.