What is ChatGPT?
Lately, it’s been really hard to avoid #ChatGPT. The new #AI technology tickles us on social media, traditional press and TV.
AI technologies are nothing new but this time it feels different. Not another buzzword that will pass away as quickly as it arrived but here to stay.
In just 5 days of its launch, ChatGPT reached over a million users (!). In fact, it’s the fastest growing platform in history. In comparison, iPhone took 74 days to reach 1 million users, Instagram took two and a half months, and Netflix 3.5 years. It gives you a roportion…
Over the past few weeks I have been testing, trying, and challenging ChatGPT. I’ve already implemented some of the outputs I’ve produced and I feel just like a kid in the biggest toy store ever.
For the benefit of those who are interested in getting to know GPT and ChatGPT I decided to make some order and start from scratch:
ChatGPT technology was created by the artificial intelligence company #OpenAI. It is a non-profit company but its subsidiary, OpenAI LP, was established for profit. What is the profit potential? It’s impossible to estimate at this point, but about a week ago, its value was already estimated at $29 billion.
One of the most interesting investors in OpenAI is Microsoft, which invested $1 billion in the company. According to recent reports, Microsoft is considering another investment, this time totaling $ 10 billion and intentions to integrate OpenAI technologies into the Office suite. No less interesting than this, there will be the integration of OpenAI in bing, Microsoft’s search engine that has been trying to bite into Google, for many years but without success.
This is probably the first time that Google has faced a strategic threat to its search engine. How does Google respond and is it indeed a real threat? – About this in a separate post.
Let’s go back to the revolution… So what exactly is ChatGPT?
ChatGPT is a chatbot that you can “chat” with, similar to how we communicate with other bots in recent years, for example in customer service. Similar but only on the face of it.
The ChatGPT is a language model developed by OpenAI based on an engine called GPT-3.5 (this is the current version). It has an incredible ability to interact in a dialogue format and provide answers that seem surprisingly human. “Surprising” is an understatement…
How does this happen?
The language model is called
LLMs (Large Language Models) and through it you can predict the next word in a series of words in a particular sentence as well as the following sentences. This capability allows you to create full paragraphs and content pages.
But it turned out that LLM is limited. He does not always understand exactly what human beings are really asking for.
So what have they done?…
Because of this limitation, OpenAI has integrated an additional layer into ChatGPT that also includes human feedback.
The technology is called Reinforcement Learning with Human Feedback (RLHF). This layer helps ChatGPT learn how to follow the guidance of human users and create answers and content at a satisfactory level.
In fact, what distinguishes ChatGPT from a regular chatbot is the fact that it has been trained to understand the human intention hidden behind the question and to provide useful, real (under limitations…) and harmless answers.
And what can be done with all this beauty?
You are invited to continue following…