Random facts about ChatGPT and when it will replace us.

AI technology has been on the rise for years, but in 2023, it seems to be snowballing even further.

Big news every day: 谷歌 recently opened limited access to its chatbot Bard for US and UK users, while OpenAI opened ChatGPT-4. With so much AI hype, it's worth taking a moment to learn some random facts about ChatGPT:

• Created by the company OpenAI . A random guy Elon Musk used to be there in the beginning but have nothing to do with it anymore.
• ChatGPT3.5 created on the model GPT (generative pre-trained transformer) 3.5. For learning OpenAI used web scrappers and a lot of other tools (for example previous models used just web texts),so, it scrapped internet data till 2021 year. It takes text, splits it in tokens (~4 symbols) then it learns how these tokens are seen in text in which order and connections based on this info it can build its own sentences consisting of these tokens and their probabilities and a lot of calculations, bla-bla-bla. So in other words, it repeats what was said on the internet and does not do any calculations. Considering that free ChatGpt 3.5 is so good in writing some texts, emulating something, helping with basic code (imagine learning on StackOverflow Q/A), but there are a lot of mistakes and false statements, because...well, sometimes someone is wrong in the internet.
• Notion is one of the first partners of OpenAI, so NotionAI uses GPT model from OpenAI. The same.
• 微软 is the main investor of OpenAI . They funded them long ago and according to contract OpenAI shoud use Azure resources for their computing and development, and they have a remarkably close partnership. So, MS has a HUUUGE advantage in AI race because of this experience and using Azure for this area. And that's why they are implementing AI and GPT models everywhere (Bing, Teams, Azure AI, etc)
• - ChatGPT4 is a huge deal. It has a dataset and variables in the model in >100 times more and the dataset is more up to date
• There is already a new title "prompt developer" - a person who knows how to properly talk and do the job using ChatGPT.
• ChatGPT remembers the context in bounds of open chat tab. So you can and should ask it for more details and manipulate context and ask it to change/extend answers, etc.
• For a good answer you should describe the context of your question and describe HOW you want it to act as. For example:

"I want you to act as a cyber security specialist. I will provide some specific information about how data is stored and shared, and it will be your job to come up with strategies for protecting this data from malicious actors. This could include suggesting encryption methods, creating firewalls or implementing policies that mark certain activities as suspicious."

• There is a good list af already created prompts for preconfiguration: https://github.com/f/awesome-chatgpt-prompts
• For calculations use WolframAlpha. It's a true computing system. maybe I'll write about it further.
• GitHub just announced GitHub Copilot X:

- Copilot Chat - "ChatGPT-like experience in your editor" powered by GPT-4

- Copilot for Pull Requests - AI-generated descriptions for pull requests on GitHub

- GitHub Copilot for Docs - chat for *any* company's repos and docs, starting with React/Azure/MDN

- GitHub Copilot CLI - compose commands and loops, and throw around obscure find flags to satisfy your query.

My opinion:

• Have we reached technological singularity (https://en.wikipedia.org/wiki/Technological_singularity), where machine can build a machine and do it better than human? No.
• But it will be reached through Breaking Points. Previously such Breaking Points were a creation of internet, social networks, mobile devices, and current state of AI is definitely such a thing
• Will it replace men of ours professions? I don't think so. BUT and BUT, people who know how to use it in work/day life will be/are in a much superior position that people who think that it's a toy or do not use it, and I'm talking not only about tech professions.
• Currently from dev/ops perspective it's in the state of someone of junior grade (non-offence) to whom you describe what you want to do and then it does a basic thing, then you ask to fix/expand/describe here and there. But it cannot write a complex system because it's just a system that operates tokens and cannot "think" and compute.?

#chatgpt #ai #ml #artificialintelligence #gpt #future #openai #microsoft #machinelearning