Is chatGPT the new Google Search?

What is ChatGPT?

ChatGPT is a variant of the GPT (Generative Pre-trained Transformer) language model specifically designed by OpenAI to have conversations with users in a natural and lifelike manner, and it can generate responses that are appropriate and coherent in a given context.

Large language models like GPT-3.5 are trained to predict the next word in a series of words.

ChatGPT has also been trained using Reinforcement Learning with Human Feedback (RLHF), which helps it learn to follow directions and produce responses that are satisfactory to humans. This additional layer of training helps ChatGPT to better understand and respond to user input.

Large Language Models

ChatGPT is a large language model (LLM).

A large language model is a type of artificial intelligence model that is trained on a very large dataset of human-generated text and is able to generate natural language text.

These models are called "large" because they are trained on a much larger dataset than smaller language models, which allows them to learn more about the structure and patterns of human language.

According to?Stanford University :

“GPT-3 has 175 billion parameters and was trained on 570 gigabytes of text. For comparison, its predecessor, GPT-2, was over 100 times smaller at 1.5 billion parameters.

This increase in scale drastically changes the behavior of the model — GPT-3 is able to perform tasks it was not explicitly trained on, like translating sentences from English to French, with few to no training examples.

This behavior was mostly absent in GPT-2. Furthermore, for some tasks, GPT-3 outperforms models that were explicitly trained to solve those tasks, although in other tasks it falls short.”

Large language models (LLMs) are able to predict the next word in a series of words in a sentence and even generate entire paragraphs or pages of content. They work somewhat like autocomplete, but on a much larger scale.

While LLMs are very good at generating coherent and natural-sounding text, they can sometimes struggle to fully understand the specific intentions of a human user. That is where Reinforcement Learning with Human Feedback (RLHF) plays an important role in the ChatGPT model.

How Was ChatGPT Trained?

GPT-3.5 was trained on a very large dataset of human-generated text, including books, articles, and information from the internet, including sources like Reddit discussions, to help ChatGPT learn dialogue and attain a human style of responding.

In addition to being trained on this dataset, ChatGPT was also trained using a technique called Reinforcement Learning with Human Feedback (RLHF). This technique involves using human feedback to help the chatbot learn to follow directions and generate responses that are satisfactory to humans.

A March 2022 research paper titled?Training Language Models to Follow Instructions with Human Feedback ?explains why this is a breakthrough approach:

This approach involved working with a team of human contractors (called labelers)?who provided feedback on the chatbot's responses. The labelers would present a prompt to the chatbot, and the chatbot would generate a response. The labelers would then rate the response based on how closely it adhered to the prompt and how human-like it sounded. This feedback was used to train the chatbot to better understand and respond to user input."

What sets ChatGPT apart from a simple chatbot is that it was specifically trained to understand the human intent in a question and provide helpful, truthful, and harmless answers.

What are the Limitations of ChatGPT?

1. It is a statistical model and relies on patterns in the data it was trained on, so it may not always produce accurate or appropriate responses to unusual or unforeseen input.
2. It is not able to provide definitive answers to factual questions or perform tasks beyond generating text.
3. It does not have access to the internet or external knowledge sources, so it cannot provide up-to-date information or engage in discussions about current events.
4. It is not able to understand or interpret the emotions or intentions of the user, so it may not always generate responses that are sensitive to the user's feelings or needs.
5. It is not capable of learning from new experiences or adapting its behavior based on past interactions, so it may not be able to improve its performance over time

Is ChatGPT Free To Use?

Currently, ChatGPT is available for users to test and provide feedback on during the "research preview" period at no cost. This allows users to try out the chatbot and helps the AI improve its performance by learning from any mistakes it makes in its responses to questions.

There is currently a contest with a prize of $500 in ChatGPT credits to encourage the public to rate the responses. The currently ongoing contest ends at 11:59 p.m. PST on December 31, 2022. So you have two days to win yourself 500$ worth of ChatGPT API credits.

Will Language Models Replace Google Search?

Google has developed an AI chatbot called LaMDA , which has performed so convincingly in simulating human conversation that a Google engineer claimed it was sentient.

Given the capabilities of large language models to answer a wide range of questions, it is not unreasonable to consider the possibility that companies such as OpenAI, Google, or Microsoft may one day replace traditional search with an AI chatbot.

But ChatGPT given its current limitations, is far from being the Google killer (at the moment). Google provides information from multiple credible sources, constantly updating its algorithm to filter wrong information. ChatGPT is revolutionary, and it will be a blast to see its development in the future.