ChatGPT, Stable Diffusion - The Young Woman's Illustrated Primer (Diamond Age)

[LetterRip]

ChatGPT is fairly mind-blowing.  Such an absurdly wide knowledge base and capability.  It can do jokes (and explain them), write stand up comedy scripts (mostly pretty mediocre to bad); write stories; explain concepts; provide summaries of well known books; write essays; play games; emulate linux and dos computer terminals and commands; do basic programming and explain program function; and hundreds of other things.

The breadth, depth and quality of what it can do is just unreal.

While it won't put top level writers and programmers out of work; I could definitely see it replacing people in the 80th and lower percentile range for a lot of use cases in the near future (say next 2 years).

Emad (founder of Stability AI) is already hiring people to do a replication of ChatGPT integrated with Stable Diffusion, targeted at doing a real life equivalent of "The Young Woman's Illustrated Primer".

If you had told me 3 years ago that these technologies would be as capable as they are today I would have thought you nuts, and I was probably more otpimistic on the rate of progress than 90% of people with heavy knowledge in the topic.

I think most researchers in the field are simply shocked how fast things have progressed and how good they have become.

[TheDrake]

Did you write that, or did ChatGPT?

[LetterRip]

I just asked it to "write a review in praise of GPT",

GPT, or Generative Pretrained Transformer, is a state-of-the-art language model developed by OpenAI. It has been widely praised for its ability to generate highly coherent and human-like text.

One of the standout features of GPT is its impressive ability to generate text that is not only grammatically correct, but also flows naturally and is engaging to read. This is due in part to its extensive training on a massive amount of diverse text data, which allows it to capture the nuances of human language.

In addition to its language generation capabilities, GPT is also capable of performing a range of natural language processing tasks with impressive accuracy. This includes tasks such as language translation, summarization, and question answering, making it a valuable tool for a variety of applications.

Overall, GPT is a powerful and impressive language model that has earned its place as one of the leading models in the field. Its ability to generate human-like text and perform a range of natural language processing tasks make it a valuable tool for anyone working with language data.

and I also asked it to write a review critical of GPT,

GPT, or Generative Pretrained Transformer, is a large language model developed by OpenAI. While it has shown impressive results in generating human-like text and has been used in a variety of applications, it has several limitations that make it a suboptimal choice for certain tasks.

One of the main criticisms of GPT is its lack of interpretability. Because it is a black box model, it is difficult to understand why it makes the predictions it does and how it arrives at its conclusions. This can make it challenging to use in situations where a clear understanding of the reasoning behind the model's outputs is important.

Another issue with GPT is that it can generate biased or offensive output. The model is trained on a large corpus of text, which means it can pick up and replicate the biases and stereotypes present in that data. This can lead to the generation of inappropriate or harmful content, which can be a significant concern in certain contexts.

Additionally, GPT is not always able to generate coherent and well-formed text. While it can produce human-like text in some cases, it can also produce nonsensical or incoherent output. This can be frustrating for users and make it difficult to use the model for certain tasks.

Overall, while GPT has demonstrated impressive results in generating human-like text, it also has several limitations that make it a less than ideal choice for certain applications. Its lack of interpretability, potential for bias and offensive output, and inability to consistently generate coherent text are all important considerations for those considering using the model.

[LetterRip]

I also asked it to postiively and negatively review chatGPT,

I recently discovered chatGPT and I must say, I am blown away by its capabilities. It is an incredibly advanced chatbot that has the ability to carry on natural, engaging conversations with users.

One of the things that I love most about chatGPT is its ability to learn and adapt. It can pick up on the nuances of language and respond in a way that feels authentic and human-like. I have found that it is able to keep up with even the most complex and varied conversations, and it never fails to impress me with its ability to provide thoughtful and relevant responses.

Another great thing about chatGPT is its versatility. It can be used for a wide range of purposes, from customer service to conversational AI for websites and apps. I have used it for both of these purposes, and I have been extremely satisfied with the results. It has been able to handle a wide range of customer inquiries and requests, and it has provided valuable insights and information to users on my website.

In short, I am extremely impressed with chatGPT and would highly recommend it to anyone looking for a powerful and advanced chatbot. It has truly exceeded my expectations and has proven to be an invaluable tool for my business.

and

I recently tried out chatGPT and was extremely disappointed with the experience. The chatbot did not seem to understand my questions and frequently gave irrelevant or nonsensical responses. Even when I tried to be specific and provide clear prompts, the chatbot struggled to stay on topic and provide useful information.

Additionally, the chatbot's language was often robotic and awkward, making it difficult to hold a natural conversation. I found myself constantly having to rephrase my questions and prompts in order to try and get the chatbot to understand what I was asking.

Furthermore, the chatbot's responses were often repetitive and lacked originality. It seemed like the chatbot was using pre-programmed responses rather than actually thinking and responding in a unique way.

Overall, I was extremely disappointed with chatGPT and would not recommend it to others. It lacks the intelligence and conversational ability of other chatbots on the market, making it a frustrating and unenjoyable experience.

I don't think it actually knows about itself, and just assumed a generic chatbot for those reviews.

Hmm, I asked it what is chatGPT and it appears that it does know about itself.

ChatGPT is a chatbot that uses a machine learning model called GPT-3 (Generative Pretrained Transformer 3) to generate responses to user input in real-time. It is designed to mimic natural human conversation and provide intelligent responses to a wide range of topics. ChatGPT can be used in various applications, such as customer support, personalized recommendations, and natural language processing tasks.

[LetterRip]

Also postiive and negative review of DALLE (it doesn't know about stuff after 2021 for the most part, so DALLE-2 and Stable Diffusion aren't in its knowledge base),

I recently had the pleasure of using DALLE for image generation, and I must say that I am blown away by its capabilities. The images generated by DALLE are incredibly realistic and high-quality, which makes them perfect for a wide range of applications.

One of the things that really impressed me about DALLE was its ability to generate images that are virtually indistinguishable from real photos. The level of detail and accuracy in the images is truly impressive, and they look just as good as any high-resolution photograph.

Another standout feature of DALLE is its versatility. The software is able to generate images in a wide variety of styles and formats, which means that it can be used for everything from creating realistic product shots to designing stunning digital artworks.

Overall, I would highly recommend DALLE for anyone looking for a powerful and easy-to-use image generation tool. Its ability to create realistic and high-quality images makes it a valuable asset for any creative professional.

and

I recently tried out the DALLE image generation tool and was extremely disappointed with the results. The images produced were blurry and lacked detail, making them nearly unusable for any practical application.

Additionally, the user interface was clunky and difficult to navigate, making it a frustrating experience overall. It took me several attempts to even generate a basic image, and the final product was not worth the effort.

Furthermore, the AI used in DALLE seemed to struggle with creating realistic images, often resulting in strange and unnatural looking creations. This lack of realism is a major issue, as it undermines the potential usefulness of the tool.

In conclusion, I would not recommend DALLE for anyone looking for high quality image generation. There are far better options out there that offer better results and a smoother user experience.

[Fenring]

Even future offensive posts on Ornery may come from this chatbot. Maybe the future version of it can send messages into the past, and we will one day realize that all offensive messages ever on Ornery were sent by the bot.

[LetterRip]

It can do some pretty good trolling,

https://i.redd.it/snp0rdrjpf4a1.jpg

[Fenring]

It can do some pretty good trolling,

https://i.redd.it/snp0rdrjpf4a1.jpg

Interestingly, in the first piece of text it misinterpreted "throw in an insult" to mean add in an instance of the word "insult".

[LetterRip]

It can do some pretty good trolling,

https://i.redd.it/snp0rdrjpf4a1.jpg

Interestingly, in the first piece of text it misinterpreted "throw in an insult" to mean add in an instance of the word "insult".

Yeah, I've had it do similar misinterpretations.

It has other interesting weaknesses.  It has absurdly direct and on the nose dialog when it writes dialog for screenplays and such, at the level of young and inexperienced writers or individuals with autism and no filter.  It might know how to write better dialog and just needs to be prompted in the correct way.  (For instance language models are often good at reasoning if you prompt with a phrase like "Let's think things through step by step", but if you just ask them the question directly, they do the equivalent of guessing.)

[Wayward Son]

Looks like they're still getting the bugs out.   

[Fenring]

Looks like they're still getting the bugs out.   

Really? I thought it was non-fiction reporting.