Is GPT-4 better than ChatGPT? What are the differences?

Those interested can now get their hands on the highly anticipated Chat-GPT 4, the fourth dedication of the AI chatbot developed by Open AI.

“We’ve created GPT-4, the latest milestone in OpenAI’s effort in scaling up deep learning,” announced the tech company earlier this week when releasing the product. The team at Open AI touted the chatbot’s high scores on a “simulated bar exam,” where they say the responses ranked in the top ten percent; the previous edition fell in the bottom ten percent. Aside from the bar, exam scores for the fourth edition are higher across the board compared to the 3.5 version many uses became familiar over the last few months.

“Though far from perfect,” Open AI sees Chat-GPT as the most advanced AI of this kind. The message is very clear. Open AI is pleased with the developments made but understands there is a lot of work to do.

Some users have pointed out errors they have taken note of online.

Interacting with Chat-GPT

Since the company was so persistent about the chatbox’s score on the bar, I thought I would ask it a question related to the law.

I asked, "What was decided in the Supreme Court case Brown vs. Board of Education?" The chatbot responded with a concise four-paragraph summary of the case and its influence on the Civil Rights movement.

The answers are a bit vague

The only part that may feel off to users is how vague and passive the writing is. The last paragraph begins, “The decision in Brown v. Board of Education was a turning point in the Civil Rights Movement,” but has not previously mentioned the context under which the decision was made.

The bot continued, saying that the court’s decision “led to the desegregation of public schools throughout the United States.” The bot doesn’t mention that schools are increasingly segregated (a point that will become important in a moment). Instead, it describes the landmark case as “a significant victory in the struggle for racial equality.”

A follow-up question

When I asked, “How segregated are schools today?” The chatbox said that they were “somewhat segregated,” noting regional and district-level variations in the “degree of segregation.” Then, it went on to cite a report from the National Center for Education Statistics which found that during the 2017/2018 school year, fifty-two percent “of all public schools in the United States had student populations that were 50% or more economically disadvantaged” and that “these schools” were disproportionally located in urban areas and were “more likely to serve students of color.”

Then it went on to cite another study, this one from the Civil Rights Project at UCLA which “found that segregation in public schools has been on the rise in recent years, with more than one in five black students attending a school that is highly segregated by race and poverty.” The chatbot ended with a call to action for “policymakers, educators, and communities to work together to address the factors that contribute to school segregation and ensure that all students have access to high-quality education regardless of their race or socioeconomic status.” Again, there is vagueness to this final message. What are the “factors that contribute to social segregation”? Racism? Classism? We become aware that there is a problem, but we are no better equipped to describe it.

Overall, it is clear that improvements have been made, and the finish feels cleaner. However, the content remains full of jargon that fills space but does not actually tell a reader very much. The vagueness allows a cover for accuracy, and it remains to be seen how long that cover will be needed.