OpenAI's GPT-4 raises the bar for AI language models

Nick's ESL News Lessons

A printable worksheet for news-based English conversation classes.

Date: 2023/03/16

Difficulty: Upper Intermediate B2

PDF: OpenAI's GPT-4 raises the bar for AI language models
YouTube: OpenAI's GPT-4 raises the bar for AI language models

Vocabulary Match

groundbreaking •• teacher that asks questions
predecessor •• true
Socratic tutor •• previous version
factual •• innovative
skeptical •• doubtful

Vocabulary Answers

groundbreakinginnovative
predecessorprevious version
Socratic tutorteacher that asks questions
factualtrue
skepticaldoubtful

Article

GPT-4 is the latest version of OpenAI's groundbreaking AI language model.

According to the company, GPT-4 scores higher on intelligence and knowledge tests than its predecessor, GPT-3, and can now respond to images as well as text. Among its new features, GPT-4 can act as a Socratic tutor, and discuss the contents of photographs.

The new release is 40 percent more likely to provide factual responses and 82 percent less likely to respond to requests that should be disallowed.

However, GPT-4 still suffers from problems seen in GPT-3, such as showing preferences, providing false information, or accepting requests to role-play upsetting personalities.

Some AI experts are skeptical of its usefulness. Oren Etzioni, a professor at the University of Washington said, "While they've made a lot of progress, it's clearly not trustworthy."

Quiz

  1. How does GPT-4 perform on intelligence and knowledge tests compared to ChatGPT?
  2. What are some problems that GPT-4 still suffers from?

Quiz Answers (examples)

  1. It scores higher on intelligence and knowledge tests than GPT-3.
  2. It shows preferences, provides false information, and accepts requests to role-play upsetting personalities.

Conversation

A:Hey, have you heard about the new version of ChatGPT?
B:Yeah, it's even smarter and can understand pictures now, right?
A:Yep, it's awesome, we could totally use it to do our homework for us.
B:But wouldn't that be cheating? What if we get caught?
A:No chance! We're going to be top of the class this year!
B:But what if it gets the answers wrong? I heard it's not always factual.
A:Hmm, you have a point. Maybe we should just use it as a personal tutor.

Discussion

  1. Have you ever used an AI chatbot or virtual assistant, like Siri or Alexa? What was it like?
  2. If you could have an AI assistant help you, what would you want it to do?
  3. Do you think AI language models will eventually replace human writers and journalists?

Source Material

https://www.wired.com/story/gpt-4-openai-will-make-chatgpt-smarter-but-wont-fix-its-flaws/

MORE: