Just a few months ago ChatGPT was launched which changed many people’s perception of what AI can do. It was based on OpenAI’s GPT-3.5, which was also integrated into Microsoft’s Bing and Skype, Edge too. Now the company has confirmed that it has switched to the new and more powerful GPT-4 model.
In fact, it did a while ago: if you’re part of the Bing preview, you’ve been using GPT-4 for the past five weeks (you can sign up for the preview here). This is not plain GPT-4, by the way, but a version that has been customized by Microsoft for research.
So, what’s new in GPT-4? For starters, it’s a “multimode” model, which is a fancy way of saying you can attach images to your query, not just text. Here is an example of GPT-4 explaining a joke found on Reddit. Note that the output is text only (i.e. no images can be generated like with Stable Diffusion, MidJourney, etc).
GPT-4 explains what’s funny about a lightning cable shaped like a VGA cable
The new model is also smarter, the OpenAI team tested it with the 2022 and 2023 practice exam books. Note: The model doesn’t know anything after September 2021, so these exams (and their answers) weren’t part of the training data.
GPT-3.5 took the bar exam (which lawyers must pass) and scored in the bottom 10%. GPT-4 scored in the top 10%. The justice system isn’t ready for robo-lawyers yet, but they are on the horizon. GPT-4 also scored in the 88th percentile on the LSAT exam, v3.5 was in the 40th. For SAT Math, GPT-4 was in the 89th percentile, GPT-3.5 in the 70th. You can check the OpenAI announcement for more exam results.
The most important novelty of version 4 is the “steerability”. Previously, ChatGPT was forced to act as a digital assistant by prepending some rules. It was possible to trick the AI into revealing those rules, e.g Here you are what Microsoft told “Sydney” to do like Bing (including not revealing its Sydney codename):
The rules for an early version of Bing based on ChatGPT
Microsoft and OpenAI have worked to hide those rules (to prevent so-called “jailbreaking”), but now there’s a better way to do it: Enterprises can control AI style and activity with a system message. Here is an example:
AI personality change is now done with system prompts
It’s important to note that GPT-4 still has limitations, especially when it comes to facts. Like its predecessor, the model can invent things, these are called “hallucinations”. The new version is significantly better (scoring 40% higher in internal tests) than GPT-3.5 at sticking to the facts and not making any logical errors, but it’s still not perfect. However, GPT-3 was released in mid 2020, GPT-3.5 arrived in early 2022 (a later improvement was used for ChatGPT), so the pace of improvement is nothing short of incredible.
Now all we want to know is this: can we have a GPT-4 powered Cortana?
Start a new Thread