Microsoft has officially confirmed that Bing is using OpenAI’s GPT-4 AI language model, which was announced earlier this week. The AI-powered chatbot is currently powering Bing Chat, and Microsoft has reportedly been using it for five weeks.
It’s also a key component of Duolingo’s new AI-backed features, and online learning platform Khan Academy has started evaluating it as well. Other early adopters include the Iceland government, Morgan Stanley and Stripe.
GPT-4 has a 25,000-word capacity for input and output
Microsoft Confirms It Is Utilising OpenAI?s GPT-4 on Bing Search Engine
The latest model of GPT-4 has a whopping 25,000-word capacity for input and output. This capacity will allow you to provide it with much more detailed input prompts, enabling it to answer questions more efficiently.
Additionally, GPT-4 is multimodal and can process and respond to texts as well as images. It can also accept a much wider range of languages than the previous versions.
This is a huge development and opens up a lot of possibilities for developers. It allows you to feed the chatbot with much longer inputs and answers, and even allow it to synthesize outputs from multiple images and texts into one.
However, it must be noted that GPT-4 is not perfect and can still make reasoning errors. It may be prone to share disinformation, so it is important to use it in a carefully thought-out manner.
It is multimodal
A German news report claims that Microsoft Confirms It Is Utilising OpenAI?s GPT-4 on Bing Search Engine
Microsoft is using the latest version of OpenAI’s ChatGPT model on its Bing search engine. The chatbot is used to explain and answer questions people type into the search engine, with responses provided in a more conversational manner.
The GPT-4 model is multimodal, meaning it can accept and process input in text as well as images. This is a major advancement, as previous versions of GPT could only process text.
This feature will allow GPT to respond to queries that include a combination of image and text inputs, such as asking the chatbot to explain a joke involving a decades-old computer cable plugged into a modern iPhone’s tiny Lightning port.
This update comes after the previous iteration of GPT, GPT-3, suffered from a number of problems, including “hallucination” and generating biased, false, or hateful text. OpenAI has made improvements to the system, but it still has a long way to go.
It can process and respond to languages
GPT-4 can process and respond to natural language text for a variety of tasks, including language translation, text summarization and question answering.
The model can also accept and interpret images, such as documents that combine text and images, hand-drawn sketches and screenshots. This is a significant advance, as it makes it easier to train the AI to understand and respond to different kinds of input.
Another major advance in GPT-4 is that it can caption relatively complex images, such as a picture of a bulky decades-old computer cable plugging into the Lightning port of a modern iPhone.
The AI can also self-upgrade its abilities, which means that it learns and improves based on training data. This is a huge step forward in AI tech, and one that could transform how we use it.
It can accept images as input
Microsoft Confirms It Is Utilising OpenAI?s GPT-4 on Bing Search Engine
In a blog post today, Microsoft confirmed that its botchatbot ChatGPT has been using the latest version of OpenAI’s language model, GPT-4, from day one. It’s a multimodal model that can accept images as input and generate captions, classifications and analyses for them.
It also exhibits human-level performance on various professional and academic benchmarks. For instance, it can pass a simulated bar exam with a score that’s in the top 10 percent of test takers.
It also responds to language in a more creative and collaborative way than its predecessor, ChatGPT-3 and 3.5. It’s 82% less likely to give inappropriate content and follows guardrails better for sensitive topics like medical advice and self-harm.