Shownotes Transcript
Technological Capabilities of ChatGPT 4
—
Today's Amazon Deals - https://amzn.to/3FeoGyg)
—-
Technological Capabilities of ChatGPT 4
GPT-4, the latest milestone in OpenAI’s effort to scale up deep learning, has been created by us. A large multimodal model that accepts image and text inputs and emits text outputs, GPT-4 exhibits human-level performance on various professional and academic benchmarks, although it is less capable than humans in many real-world scenarios. For example, a simulated bar exam was passed with a score around the top 10% of test takers, whereas GPT-3.5 scored around the bottom 10%. GPT-4 has been iteratively aligned for 6 months using lessons from our adversarial testing program as well as ChatGPT, resulting in our best-ever results (though far from perfect) on factuality, steerability, and refusing to go outside of guardrails.
GPT-4 was tested with several different exams around the world and with recent publications of exam editions and olympiads test cases, along with many other important testing benchmarks. It performed exceedingly well in many exams, especially better than GPT - 3.5.
Figure 32.1: Performance comparison GPT- 4 over academic and professional exams
[Source: GPT- 4 technical paper]
GPT-4 was again verified against the some concrete state-of-the-art(SOTA) ML models training available and which may include benchmark-specific crafting or additional training protocols, against some potential benchmarks. GPT-4 almost outperformed the other models significantly.
Figure 32.2: Performance of GPT-4 with some LM models against some benchmarks
[source: GPT- 4 technical paper]
Some Use Cases
GPT-4 already seemed to be exceeding the expectations of existing chatGPT with 3.5 versions. It seems to be exceeding the ChatGPT in advanced reasoning as their blog already showed an example of it, displaying more correctness than ever:
Figure 32.3: GPT’s advanced reasoning ability over ChatGPT
[Source: OpenAI blog]
With the new multi-modal ability, it can also have visual readability abilities and can go through visual reasoning and logic extracted from a picture. It can even perform visual question answering (VQA) tasks with a good perfection, with similar capabilities like it does for textual data.
Figure 32.4: GPT-4 performing VQA logical task
[Source: GPT- 4 technical paper]