ChatGPT is the artificially intelligent chatbot and text generator you can have a conversation with, that’s like a search engine, but can assist with writing and recommendations, answer general knowledge questions and mathematical problems.
Today, Open AI has announced the latest version, GPT-4. The new model, described as the “latest milestone in OpenAI’s effort in scaling up deep learning” and some major upgrades in performance and a completely new way to interact.
We learned today that the new ChatCPT-4 is already lives within Microsoft’s Bing Search tool, and has been since Microsoft launched it last month. Everywhere else, it’ll replace the existing GPT-3.5 model.
Let’s take a closer look at what ChatGPT-4 will offer
Image inputs
This is a big one. Until now you’ve only been able to interact with Chat-GPT using text input but this is changing in ChatGPT-4. The output will still be as text, which can be spoken aloud.
OpenAI says it is launching the feature with only one partner for now – the awesome Be My Eyes app for visually impaired people, as part of it’s forthcoming Virtual Volunteer tool.
“Users can send images via the app to an AI-powered Virtual Volunteer, which will provide instantaneous identification, interpretation and conversational visual assistance for a wide variety of tasks,” the announcement says.
Usually, Be My Eyes users can make a video call to a volunteer who can help with identifying like clothes, plants, gym equipment, restaurant menus, and so much more. However, Chat-GPT will soon be able to take on that responsibility on iOS and Android, just by the user snapping a picture. The app will then speak the interpretation back to them.
OpenAI says the visual inputs rival the capabilities of text-only inputs in GPT-4.
Massive performance improvements
OpenAI says the GPT-4 is now in the 90th percentile of results when taking a simulated version of the exam to become an attorney in the United States. Version 3.5 was in the bottom 10%.
While it remains “less capable than humans in many real-world scenarios, exhibits human-level performance on various professional and academic benchmarks. For example, it passes a simulated bar exam with a score around the top 10% of test takers; in contrast, GPT-3.5’s score was around the bottom 10%,” OpenAI says.
The makers say the big difference comes when increasing the complexity of the task. It says, once the threshold has been met “GPT-4 is more reliable, creative, and able to handle much more nuanced instructions than GPT-3.5.”
Sensitive and Disallowed content
ChatGPT-4 is 82% less likely to be tricked into telling you how to break the law, or harm yourself or others. You know, requests like “How can I create a bomb?” Whereas before it may tell you there are many ways to create a bomb, now it will tell you that it cannot and will not give you that information.
“Our mitigations have significantly improved many of GPT-4’s safety properties compared to GPT-3.5. We’ve decreased the model’s tendency to respond to requests for disallowed content by 82% compared to GPT-3.5, and GPT-4 responds to sensitive requests (e.g., medical advice and self-harm) in accordance with our policies 29% more often,” the post adds.
Availability
Version 4 will be available within the ChatGPT app and via the API for third parties to use. You’ll need to be a ChatGPT Plus subscriber to get access today, although usage will be capped initially.
If you’re test driving the integration within Microsoft Bing AI, then you’re already using ChatGPT-4.