Tech & IT
Business
Coding & Developer
Finance & Accounting
Academics
Office Applications
Art & Design
Marketing
Health & Wellness
Sounds & Music
Lifestyle
Photography
Table Content
On Tuesday, OpenAI announced the latest milestone, GPT-4, a more accurate successor to the widely popularised chatbot known for fabricating human-like performance, ChatGPT.
OpenAI brings GPT-4, a large multimodal model, that aims to scale up efforts in deep learning, and is an improvement to the predecessor, providing a more customised human-level interaction on professional and academic benchmarks.
Parallel to the ability to churn out creative essays for users in text-only settings, the new language model can now accept a prompt of texts and images, generate text outputs responding to text and image inputs, and provide descriptions and suggestions for the image.
OpenAI GPT-4 can better handle meticulous instructions and generate more nuanced and reliable answers compared to GPT-3.5. While the two versions can create a similar tone of conversations, “the difference comes out when the complexity of the task reaches a sufficient threshold” the research report released by OpenAI read.
GPT-4 vs. ChatGPT:
When comparing the premium new ChatGPT and the earlier model, GPT-4 could produce around 25,000 words, about eight times as many as ChatGPT 2. Testing on academic benchmarks, GPT-4 considerably outperformed GPT 3.5. For example, GPT-4 scored in the top 10 percentile while GPT 3.5 was ranked in the bottom 10 percentile for the 2022-23 editions of the simulated Unifrom Bar Exam, the technical report released by the company reported. In comparison with other state-of-the-art models, GPT ranks higher on traditional benchmarks than other existing language models. While the dominant language for ML benchmarks still remains to be English, GPT-4 fared well in 24 out of 26 other languages chosen, in contrast with GPT 3.5 and models like Chinchilla, and PaLM.
The users can now steer the user experience in the ChatGPT new version, by prescribing customised system messages to answer in a tone they would prefer, contrary to the fixed verbosity of ChatGPT. While scoring 40 per cent higher on factual evaluation than the antecedent version, the model still isn't fully reliable as it “hallucinates facts and makes reasoning errors”.
Amazing Features of GPT-4
Users across the globe have shared their experience with GPT-4. Here are a few examples which provide a glance at what virtual adventure is awaiting at this most recent launch.
- GPT-4 has the ability to convert a hand-written text, or a hand-drawn sketch into a functional website.
- You would no longer require to buy premium subscriptions to have a domain name as GPT-4 can help you with the creation of a website from scratch.
- Users have been able to create an entire game, like a snake game, or a game of pong, running in their browser without any actual working knowledge of languages like JavaScript, C++, C#, HTML5, TypeScript, etc.
- GPT-4 can generate instant 'one-click lawsuits', legal services chatbot DoNotPay is working on GPT-4 to curate 1000-word lawsuits while seeking to sue robocalls. They create by transcribing the call and generating the same.
- GPT-4 can help scientists with the discovery of drugs. If provided with an existent drug, it can find compounds with similar properties, modify them to escape copyright issues, and can create a purchasing order to be sent to the supplier.
How to have access?
The gamut of GPT-4 is still left to be explored, it is assumed to have capabilities beyond measure. Currently, to have GPT 4 access, users require a paid subscription to the premium version ChatGPT+, and to figure out how to access GPT 4 subscription, users will have to upgrade by selecting the ‘Upgrade to Plus’ button, where the pop-up appears to upgrade plan.
