Simplified Tech News Reporting

Understanding ChatGPT: What’s Cool & What’s Not

0 35

In layperson’s terms, ChatGPT is an interactive AI designed by OpenAI to generate human-like responses to a wide range of prompts and inquiries. Its latest model ChatGPT-4, was recently released to the public, superseding its immediate predecessor ChatGPT-3.5 in several ways, including language comprehension, the ability to adopt a desired style, and tone and stay within bounds set by its developers.

The use of ChatGPT does not require any unique skill set. If you can use Google search, you can use ChatGPT, but unlike Google search, which is a form of narrow AI, ChatGPT is a more advanced AI. An amazing AI is so good that it is like a human in that it makes mistakes and changes its mind.

On a more serious note, ChatGPT isn’t the equivalent of the human mind. It is all just programming and maths at arguably the highest level. ChatGPT is not a conscious being. While it can draw information on facts and principles, it doesn’t teach them to engage in common sense reasoning or make new insights about how life works (at least not yet).

When generating a response to your prompt, ChatGPT doesn’t make a factual assessment or conduct an ethical evaluation of the text it provides its user. ChatGPT’s response is based on algorithmically synthesised information in a novel way because it simultaneously has access to vast amounts of information.

To manage your expectations and help you be a productive use of this AI, here are the ‘what’s cool’ and ‘what’s not cool’ things to know about ChatGPT.

 

What’s Cool About ChatGPT

  1. ChatGPT understands human-to-human language communication. After inputting your query, it generates coherent responses relevant to your inquiry. It can also execute tasks such as language translation, summarisation, proofreading and question answering. In January this year, ChatGPT-3.5 reportedly passed exams in four law school courses at the University of Minnesota, and ChatGPT-4 has proven to be brighter than its predecessor. Imagine that!

 

  1. ChatGPT is context-sensitive. It recognises context and generates contextually relevant responses more often than not. Hence, you can frequently have more in-depth and factual conversations with this AI chatbot.

 

  1. Considering the complex prompts and queries it is subjected to by millions of users and the great frequency of factual accuracy in its response, ChatGPT’s response speed of about 15-20 seconds is quite commendable.

 

What’s Not Cool About ChatGPT

  1. ChatGPT was trained on enormous text data from publicly available internet sources. To put it clearly, we humans (imperfect beings) are the primary source of the data on which this AI was trained, making it vulnerable to less frequent factual inaccuracy. ChatGPT sometimes generates fairly spectacular nonsense, which was highly evident in Microsoft’s Bing search engine responses according to negative reviews from its users. This new Bing search engine had a variation of ChatGPT built into it.

 

  1. ChatGPT will inevitably give you a wrong response if the accurate answer to your query or prompts depends on up-to-date information. This is because the bank of publicly available internet data ChatGPT was trained on is limited to the year 2021. Starting from 2022, ChatGPT doesn’t know ‘what’s up’; hence it can’t help you with the latest information after 2021. This can be considered one of the few advantages Google search still has over ChatGPT.

 

  1. Most AI, like ChatGPT, is often subject to limited training data. Since humans are sometimes biassed, the human-generated data it is trained on is also biassed, resulting in a biassed AI. As a result, it’s essential to be aware of ChatGPT’s partial responses and work to curb them. It would help if you did not take every reaction from the ChatGPT hook, line and sinker.

 

ChatGPT-3.5 vs ChatGPT-4.0: What Improved?

 

GPT-3.5 only understands text, while GPT-4 can also understand photos, diagrams, and screenshots.

GPT-4 can handle 25,000 words at a time, compared to the 3,000 words limit of GPT-3.5.

GPT-4 can handle more nuanced instructions, showing more advanced reasoning capabilities than GPT-3.5.

GPT-4 is safer in regulating access to information about disallowed or sensitive topics such as self-harm or how to obtain dangerous substances.

GPT-4 comes with a new type of feature seen as the “playground” on GPT -4’s interface, which allows users to instruct the model on the style, tone, and wordiness to use in subsequent interactions. For instance, a user can use the playground feature technically called “system message” to condition the model to respond as Socrates would to an inquiry.

 

Understand How ChatGPT Works By Using It.

The best way to learn anything is via hands-on experience. The more you use ChatGPT, the more you know what it can do and how to use it. You are most likely to find Apps on Google Play Store that claim to be ChatGPT, but they are from scammers. ChatGPT is currently not available as an App on any mobile app store. You can only access it on OpenAI’s website after opening an account. While ChatGPT-3.5 is free, its latest model ChatGPT-4.0 is only available to ChatGPT plus subscribers. This subscription costs $20 per month and comes with advantages those using the free plan don’t enjoy.

ChatGPT-3.5 is only available when demand is low. While ChatGPT/ChatGPT-4.0 is always available, it comes faster and gives priority access to new features.

 

Still curious about ChatGPT’s capabilities? Watch the live stream of ChatGPT in action.

Written by Moses Atoyebi

 

 

 

Leave A Reply

Your email address will not be published.