ChatGPT! What is it?

James Barrat, author of “Our Final Invention: Artificial Intelligence and the End of the Human Era” told the Washington Post that:

“I’m increasingly inclined to think that there should be some regulatory oversight, may be at the national and international level, just to make sure that we don’t do something very foolish. I mean with artificial intelligence we’re summoning the demon.”

ChatGPT (Chat Generative Pre-Trained Transformer) developed by OpenAI and launched on November 30, 2022, is:

“famous for enabling users to improve and steer a conversation towards a desired length, format, style, level of detail, and language used. It is thus a chatbot’s advanced conversational capability which allows us to have human-like conversations aided by natural language processing (NLP) tool driven by Artificial Intelligence technology”.

ChatGPT uses NLP to create humanlike conversational dialogue. The language model can respond to questions and compose various written content, including articles, social media posts, essays, code and emails. It is based on Generative AI that enables users to quickly generate new content based on a variety of inputs. Such Inputs and outputs to these models can include text, images, sounds, animation, 3D models, or other types of data.

Contrary to Conversational AI that control NLP, NLU, and NLG, allowing it to comprehend and respond to user inputs, Generative AI is a High-performance infrastructure for cloud computing, data analytics & machine learning. It use techniques like GANs (A Generative Adversarial Network which is a prominent machine learning frameworks for approaching generative AI) and transformer models to learn from large datasets for generating unique outputs.

NLU means Natural Language understanding which focuses on computer reading comprehension while NLG mean Natural Language Generation which enables computers to write and is therefore, a process of producing a human language text response based on some data input. This text can also be converted into a speech format through text-to-speech services.

Generative Pre-Trained refers to how ChatGPT processes requests and formulates responses, is trained with Reinforcement Learning through Human Feedback (RLHF) and reward models that rank the best responses.

This feedback helps supplement ChatGPT to improve future responses with Machine Learning which is an umbrella term that solve problems such as development of algorithms by human programmers that would be too expensive hence, the problems are solved by helping computers to “discover” their “own” algorithm without explicitly telling the machine what to do with human-developed algorithms).

People are using ChatGPT for some elaborative functions, such as writing code or even malware. We can use ChatGPT for ordinary or everyday activities such as having a friendly conversation. Also, if we ask questions to ChatGPT, we could get handsomely paid. Some conversation could be as simple as “I am hungry, what food should I get?” or as elaborate as “What do you think happens in the afterlife?” Either way, ChatGPT is sure to have an answer.

But

It can not replicate even a single verse of the Qur’an: the Holy book because Qur’an is in the words of God Almighty who challenged both Humans and Jinns to produce a verse like that in Qur’an and they both failed because when Allah challenged, no one can meet that.

When ChatGPT was asked

Can you write a verse similar to the eloquence of the Qur’an? the answer was:

As an AI language model, I can generate text based on patterns and rules, but cannot produce something that matches the eloquence of the Qur’an. The language and style of Qur’an are unique and cannot be replicated by human or machine. It is considered a miracle of its own and its literary beauty and depth have been acknowledged throughout history.

ChatGPT was initially free to use when it was in research and feedback collection phase however, since beginning of February 2023, its paid subscription version ChatGPT Plus has been launched. Its current subscription plan starts

at $20 per month, and subscribers can get numerous benefits by paying this amount.

If during the use, you get a message that “ChatGPT is at capacity” it usually means that too many people are trying to access the chatbot right now. OpenAI only has a limited number of servers running ChatGPT and these are expensive to maintain, hence it can only support a limited number of users per session.

If you could recall, whenever you enquire online through user access number for any thing related with baking or telephone you sometime get message all officers are busy at the moment, please wait. “ChatGPT is at capacity” is similar to that message except that “please wait” facility is not yet available at ChatGPT servers.

Nevertheless, according to an analysis by Union Bank of Switzerland, ChatGPT is the fastest-growing “app” of all time. The analyses estimates that ChatGPT had 100 million active users on January 23, only two months after its launch compared to Tik Tok which took nine months to reach 100 million users. This is a massive success that other companies felt compelled to release their own versions of AI chatbots, such as 

 Google which is a great tool for creating social media posts if provided with a prompt, and it will generate content tailored to your platform, audience, brand, voice, and style. This can help keep a brand’s content fresh and engaging, and also help identify new opportunities for growth. ChatGPT on the other hand, is limited to search results only and can be used to navigate the internet, as is done with Google. Overall, ChatGPT is an incredibly powerful tool for social media marketers that can help them create better content, build stronger relationships with customers, and drive business results. It will still need social media managers to make strategic decisions, create and curate content, and build and maintain relationships with their audience. ChatGPT can be a useful addition to a social media manager’s toolset, but it should not be seen as a replacement. It pushes “Conversational Marketing” into the future.

Some observers, however, have noticed that ChatGPT has lost some of its accuracy and power. For example,

Santiago@svpino tweeted on July 19^th, 2023, that:

GPT-4 is getting worse over time, not better. Many people have reported noticing a significant degradation in the quality of the model responses, but so far, it was all subjective/circumstantial but now we know.

At least one study shows how the June version of GPT-4 is objectively worse than the version released in March on a few tasks. The team evaluated the models using a dataset of 500 problems where the models had to figure out whether a given digit was crucial? In March, GPT-4 answered correctly 488 of these questions. In June, it only got 12 correct answers. From 97.6% success rate down to 2.4% (Pl. see below mentioned Figure).

Then the team used “Chain-of-Thought” (a popular technique that significantly improves answers to help the model for reasoning) and ask ChatGPT “Is 17077 a prime number? Think step by step”. Unfortunately, the latest version of GPT-4 did not generate intermediate steps and instead answered incorrectly with a simple “No.”

Code generation has also gotten worse. The team built a dataset with 50 easy problems from LeetCode (platform to help enhance skills, expand knowledge and prepare for technical interviews) and measured how many GPT-4 answers ran without any changes. The March version succeeded in 52% of the problems, but this dropped to a pale 10% using the model from June.

Why is this happening?

It is assumed that OpenAI pushes changes continuously, but not evaluating whether the models are improving or regressing? Rumors suggest they are using several smaller and specialized GPT-4 models that act similarly to a large model but are less expensive to run. When a user asks a question, the system decides which model to send the query to? cheaper and faster, but could this new approach be the problem behind the degradation in quality is not yet clear.

The selection of a better chatbot depends upon the task we need to perform. Google Bard (a conversational AI chatbot like ChatGPT that can generate text of all kind and answer any question) is better at creating human-like responses and giving users more relevant current information. Contrary to this, ChatGPT is better at processing text and can help with data summarization and other related tasks.

Bing Chat launched in February 2023 is another conversational AI chatbot feature for Bing’s search engine which allow users to interact with chatbot instead of typing search queries. It is built into the Microsoft Edge sidebar to get comprehensive answers to complex questions, summarized information, and finding inspiration to build upon all in a side-by-side view, with no need to flip between tabs.

Did you see, dear readers, so much is happening in the technology world which is difficult to comprehend. Get what you want from these chatbots. Hope you will be successful.

See you next week,

Till then,

Bye