The AI chatbot that has the whole world in a flutter

What is ChatGPT and what does it do?

The third iteration of Generative Pre-trained Transformer (GPT-3) impressed many with its potential to write poems, articles, books, tweets, and even code like humans. It is trained to predict the next word on a large dataset of text, but GPT-3 can also generate untruthful and toxic comments, spread misinformation, spam, and write fraudulent academic essays. OpenAI is attempting to address these limitations to make it “more truthful and less toxic” with human supervision. OpenAI is a non-profit AI research company founded by Elon Musk (who resigned from the board but remained a co-chair) and others.

How did the internet receive ChatGPT?

Twitter is abuzz with adjectives like “amazing”, “mind-blowing”, and “astonishing” for ChatGPT, with some even describing it a faster Google search. Computer scientist Paul Graham tweeted: “My feed is half ChatGPT screenshots. It’s the new Wordle.” Chris Anderson, curator of TED, called ChatGPT “…the biggest tech innovation of the year…”. Elon Musk tweeted: “ChatGPT is scary good. We are not far from dangerously-strong AI”, to which Sam Altman (CEO of OpenAI) responded: “I agree on being close to dangerously strong AI in the sense of an AI that poses e.g. a huge cybersecurity risk…”

How does one access ChatGPT’s functionalities?

You can test this model at chat.openai.com for free. If you use ChatGPT’s dialogue box to ask:“Is AI becoming sentient?”, the response is: “No, AI is not becoming sentient. AI is a set of algorithms and programs designed to imitate human behavior and intelligence, but it does not possess the same level of consciousness or self-awareness as a human being.”

What limitations does ChatGPT have?

The models may lack knowledge of current events since they were trained on data till end 2021. OpenAI also acknowledges that ChatGPT “sometimes writes plausible-sounding but incorrect or nonsensical answers”. The model overuses certain phrases, which arise from biases in the training data and over-optimization issues. Current models typically guess what the user intended, rather than clarifying what the user means. ChatGPT can also be manipulated to respond to harmful questions.

What does AI prowess mean for humans?

OpenAI’s DALL-E is trained to generate images from mere text. OpenAI’s neural net called Whisper enables transcription in multiple languages, and translation from those languages into English. Large language models are trained on billions of words, which makes them more knowledgeable than humans. GPT-3 and ChatGPT still do not understand what they read or write; that said, when released, GPT-4 is expected to outperform GPT-3 and ChatGPT, and the excitement will only continue.