OpenAI announced the launch of its newest large language model, the GPT-4, which it claims exhibits “human-level performance” on many professional tests.
What Is Multimodal GPT-4?
Multimodal GPT-4 is the latest version of OpenAI’s GPT large language model, which powers many of the artificial intelligence demos in the technology industry. The new model is “larger” than previous versions, which means it has been trained on more data. making it more expensive as well. OpenAI used Microsoft Azure to train the model; Microsoft has invested billions in the startup.
What Are the Advantages of Multimodal GPT-4?
OpenAI claims that the new model will produce fewer factually incorrect answers and perform better than humans on many standardized tests. GPT-4 performed at the 90th percentile on a simulated bar exam, the 93rd percentile on an SAT reading exam, and the 89th percentile on the SAT Math exam, OpenAI claimed.
The new model is more reliable, creative, and able to handle much more nuanced instructions than GPT-3.5. However, OpenAI warns that the new software is not perfect yet and is less capable than humans in many cases. It still is not reliable to provide facts, the company said.
What Are the Limitations of Multimodal GPT-4?
Multimodal GPT-4 still has many known limitations that OpenAI is working to address, such as social biases and adversarial prompts. In a casual conversation, the distinction between GPT-3.5 and GPT-4 can be subtle. The difference is more tangible when the tasks are more complex. According to the company, GPT-4 is more reliable, creative, and able to handle much more nuanced instructions than GPT-3.5.
How Can You Access Multimodal GPT-4?
The new model will be available to paid ChatGPT subscribers and will also be available as part of an API which allows programmers to integrate the AI into their apps.
Microsoft’s AI-powered Bing chatbot is already running on GPT-4, customized for search. The new Bing will benefit from OpenAI’s future improvements to “GPT-4 and beyond.” The company is also easing up on the limits it placed on the Bing AI chatbot to prevent some of its stranger behaviors; users will now be able to have 15 turns per session and a maximum of 150 per day, according to Bing CVP Jordi Ribas.
AI is continuously triggering debates about the ethics of AI especially when it comes to its role in generating misinformation. Evidently, the more advanced AI gets, the more factually reliable it should become. GPT-4 would face a scathing criticism if it fails to provide more accurate information.