Chat GPT-4 surpasses its predecessor by comprehending images, among other capabilities
On Tuesday, OpenAI announced the release of GPT-4, an upgraded version of their popular ChatGPT tool. This advanced AI technology boasts improved capabilities, demonstrated by its successful performance in various tests and its ability to generate responses based on image prompts. The new GPT-4 represents a significant advancement in the field of generative AI.
Can Chat GPT-4 really accepts visual inputs?
OpenAI has recently introduced a new version of its AI tool, GPT-4, which is capable of accepting visual inputs in addition to text. This means that users can now ask questions based on queries that include images. For instance, in an example provided by OpenAI, a user asked GPT-4 “What is unusual about this image?” while including a picture of a man ironing on an ironing board attached to the roof of a moving taxi. GPT-4 responded by identifying the unusual aspect of the image.
Moreover, GPT-4 is now equipped to comprehend and explain the humour in memes. For instance, in another example provided by OpenAI, a user shared a meme with a picture of chicken nuggets laid out to show a map of the world along with text that reads, “Sometimes I just look at pictures of the earth from space and I marvel at how beautiful it all is.” The user then asked GPT-4 to explain the meme. In response, GPT-4 analyzed the image and the text to point out the unexpected juxtaposition that made the meme humorous.
Chat GPT-4 is performing exceptionally well on exams.
GPT-4 outperforms its predecessor when it comes to exams, such as the SAT and Advanced Placement-level tests. In a simulated bar exam, GPT-4 passed with a score in the top 10 per cent, compared to the bottom 10 per cent that GPT-3.5 scored. GPT-4 scored 710 out of 800 on the SAT reading and writing section, which is 40 points higher than GPT-3.5, while on the SAT math section, GPT-4 scored 700, which is 110 points more than GPT-3.5. Moreover, GPT-4 scored a 5 on AP Biology, which is an improvement from the 4 that GPT-3.5 achieved. However, GPT-4 and GPT-3.5 both received a score of 2 on the AP English Language and the AP English Literature exams.
GPT-4’s personality is customizable
According to OpenAI, GPT-4 now has an updated “steerability” function, which enables it to modify its behaviour based on commands. This means that instead of having a fixed personality, verbosity, tone, and styles like the previous version, developers and users of ChatGPT can now customize the AI’s style and task by providing specific instructions in the “system” message.
The newly updated GPT-4 boasts an improved “steerability” function, which allows developers and users to dictate the AI’s behaviour and style by giving specific commands in the “system” message. This marks a departure from the previous version of ChatGPT, which had a fixed personality, tone, and style.
To illustrate this function, one example shows GPT-4 set up as a “Socratic tutor” who engages with the user in solving a linear equation by analyzing it together, rather than providing a straightforward answer. In another instance, the AI was set up as a “Shakespearean pirate” and responded to a user’s request for help in locating their nonqualified plans on their W-2 with a humorous and poetic response.
It is worth noting that OpenAI, the makers of GPT-4 and ChatGPT, has received significant investment from Microsoft in recent years. Microsoft has announced a multiyear, multibillion-dollar investment in OpenAI, recognizing the potential of AI technology like ChatGPT to generate readable text, images, and computer code.
Limitations of GPT-4 still exist
Despite being an expansion of earlier models, GPT-4 still has limitations, according to OpenAI. The company admitted that the AI is not fully reliable, still “hallucinates,” and has limited knowledge of events that occurred after September 2021. Additionally, the model can have biases in its output, which has been a concern for critics. Access to GPT-4 is currently available for ChatGPT Plus subscribers for a $20 monthly fee. Microsoft’s new revamped Bing search engine is powered by GPT-4 and is available to a select group of users with advanced access to the platform
In conclusion, OpenAI has partnered with several organizations, including Duolingo, Morgan Stanley, and Khan Academy, to innovate products with GPT-4.
To learn more about Chat Gpt and Digital Marketing you can join the Best Digital Marketing institute in Delhi