Chat GPT-4 is now passing the bar exam in the 90th percentile. It is getting as smart as a human. What does this mean for the future?
On Tuesday, OpenAI made a groundbreaking announcement about its latest model, GPT-4. The model is a large multimodal model that can accept both text and image inputs while returning text output that "exhibits human-level performance on various professional and academic benchmarks."
One of those benchmarks includes the Uniform Bar Exam, which the model was able to pass with a score in the top 10% of test takers.
What is GPT-4?
GPT-4 stands for "generative pre-trained transformer." It is part of a series of foundational language models that have been developed by OpenAI since 2018. GPT-4 is the latest and most advanced model in the series, with increased size and complexity resulting in generally better performance compared to previous models.
GPT-4's Capabilities
GPT-4 can follow complex instructions in natural language and generate technical or creative works. It supports generating and processing up to 32,768 tokens (around 25,000 words of text), allowing for longer content creation or document analysis than previous models.
In addition, GPT-4 has multimodal capabilities, which allow it to analyze the content of multiple images and make sense of them. It can understand a multi-image-sequence joke or extract information from a diagram.
GPT-4's Performance on Professional and Academic Benchmarks
OpenAI tested GPT-4's capabilities by having it take various professional and academic tests, including the Uniform Bar Exam, the Law School Admission Test (LSAT), the Graduate Record Examination (GRE) Quantitative, and various AP subject tests. The model performed at a human level on many of these tasks, scoring in the top 10% of test takers on the Uniform Bar Exam.
The Potential Impact of GPT-4
If GPT-4's performance is as claimed, it could represent the opening of a new era in artificial intelligence. GPT-4 has the potential to replace humans in intellectual tasks and perform general tasks at the level of a human.
It is not there yet, though. It still makes mistakes and has biases.
OpenAI's Plan for GPT-4
OpenAI plans to release GPT-4's text capability through ChatGPT and its commercial API, but with a waitlist at first. GPT-4 is currently available to subscribers of ChatGPT Plus.
The firm is also testing GPT-4's image input capability with a single partner, Be My Eyes, an upcoming smartphone app that can recognize a scene and describe it.
Limitations of GPT-4
Despite GPT-4's impressive performance, it is not without limitations. The model still reflects biases in its training dataset, hallucinates, and could potentially generate misinformation or harmful advice. OpenAI's goal is to develop artificial general intelligence (AGI) that can replace humans at any intellectual task, but GPT-4 is not there yet.
Conclusion
GPT-4 is a groundbreaking model with impressive capabilities and potential impact. Its ability to pass professional and academic benchmarks such as the Uniform Bar Exam at the 90th percentile is a remarkable achievement. However, it is important to recognize its limitations and continue to improve upon it to ensure that it is used ethically and responsibly.
Is there any concern with Chat GPT-4 passing the bar exam?
Online writer since 2008. I enjoy writing and have written nearly a thousand tech related articles about electric vehicles, software, tech companies, hardware, gaming, tech related products, and company developments in tech.
