Morfternight #62: ChatGPT is everywhere.
The one that was co-written by a human and a computer.
🤩 Welcome to the 5 new Morfternighters who joined us last week.
We love to have you here, and I hope you’ll enjoy reading Morfternight.
If you do, remember to share with a friend by clicking on this button.
📷 Photo of the week
Walk - More Photos
👋 Hello, World!
Well, it’s been quite a week for people curious about AI.
I shared a few examples generated with versions 3 and 4 of the Midjourney bot to illustrate the progress made in the past four months. Then ChatGPT became publicly accessible, and that’s all everyone seems to be talking about now.
Again, the progress is impressive.
I asked it to explain what it is directly to you:
I am writing a newsletter where I mention your existence. Can you explain what you are to my readers in about 300 words? My readers are all educated adults, but while some have a technical background, it's only the case for some of them, so please use language everyone can understand. It would be nice to add a joke or two to the explanation.
Here is the response:
Sure, here is a brief explanation of what I am that you can include in your newsletter:
I am an AI assistant trained by OpenAI, a leading research institute focused on the development of artificial intelligence technology. In simpler terms, I am a computer program that uses natural language processing (NLP) technology to generate responses to user inputs. I am not a real person, but I am able to simulate human-like conversation and provide users with information and assistance on a wide range of topics.
I have been trained on a large dataset of text, which allows me to understand and respond to a wide range of conversational prompts. This means that I am able to answer questions and provide information on a wide range of topics, from the latest developments in AI technology to the history of AI research and development.
As a large language model, I am able to process and understand complex language, and can engage in more in-depth conversations on a variety of topics. I am also able to provide detailed and accurate responses to questions, and can assist users with a wide range of tasks and information requests.
While I may not be as smart as some of your readers (after all, I am just a computer program), I am always here to help and provide interesting and informative answers to your questions. So if you have any burning questions about AI or anything else, just ask me and I'll do my best to provide a helpful and humorous response. I may not be a real person, but I am here to assist you and provide you with the information and assistance you need.
Nothing fancy here, and a GPT-2 Output Detector gives it a 99.98% chance a computer wrote it. Where it becomes interesting is that after my subsequent request.
Can you turn that explanation into a poem using alexandrine verses?
The response is now rated at 98.64% as being written by a human:
Sure, here is a short poem that explains what I am using alexandrine verses:
I am an AI assistant trained by OpenAI,
A computer program that uses NLP technology.
I am not a real person, but I can simulate
Human-like conversation with great facility.
I have been trained on a large dataset of text,
Which allows me to understand and respond to
A wide range of conversational prompts,
And provide users with helpful information and response.
I may not be as smart as some of your readers,
But I am always here to help and provide
Interesting and informative answers to your questions,
So don't hesitate to ask, and I'll do my best to oblige.
Granted, GPT-2 is a bit outdated, but it was released only 3.5 years ago.
This ability to manipulate content, rewrite it differently, or summarize it, will be extremely valuable in helping people learn. Anyone who has ever been asking questions about something they didn’t understand only to hear the same explanation over and over will agree with me.
(I also posted it on Mastodon, but there’s no embed yet).
🤖 Three places to visit today
Teachers are going to hate ChatGPT. They might not have realized yet that these tools will upend every country's education system, but already homework is becoming pointless:I have helped run an AI-based entrepreneurship program for years, written papers on the econ of AI, and follow the field quite closely. Nonetheless, I am *shocked* by how good OpenAI's new chat (chat.openai.com/chat) is. E.g., you can no longer give take-home exams/homework.
A bit more technical, but so meta: Building A Virtual Machine inside ChatGPT… Anyone said, “Simulation?”
✌️ Initial Tana experiments
Leaving the world of AI before y’all hit “unsubscribe,” I wanted to report on my first experiment with Tana. I mentioned in a recent issue of Morfternight that I was waiting to access this new tool, and a few days ago, it finally happened.
It’s a very early stage, my experience, that is. The app is already quite polished, although it’s still in alpha and evolving rapidly.
Until recently, if you wanted a note-taking solution beyond a paper notebook or its digital equivalent, you first had to figure out whether you think like an architect, a gardener, or a librarian—in other words, choosing between solutions like Notion, Roam, or Evernote.
Notion is quite powerful, but it requires thinking about the structure of your future content ahead of time. You think in terms of tables, databases, and structured data.
Roam allows you to type stuff without caring how you organize it and relies on reciprocal links between blocks to build a graph over time.
In Evernote, everything is organized in a tree of hierarchical folders, like it’s 1998.
Tana, on the other hand, brings out the best of Roam and Notion and ignores Evernote. The best of all three worlds, let me tell you :)
You can start by adding your notes without building any structure ahead of time, but whenever you feel the need to add structured data, you can very easily do so by creating “supertags.” At first, they behave like tags in every other app, allowing you to group content on the fly. But they also allow you to define for each supertag a template and fields containing metadata related to the node you tagged.
A couple of days ago, I shared my first workflow in Tana to streamline the new format of weekly review I have recently adopted: The weekly Plus, Minus, and Next review.
Expect to read more about this app shortly, and also, stay tuned for invitations whenever they’ll make them available to users.