OpenAI Announces GPT-4, the Company’s Latest Language Model

OpenAI Announces GPT-4, the Company’s Latest Language Model

As the first quarter of 2023 draws to a close, it’s clear that AI chatbots have stolen the show so far this year. Most of January and February was spent either debating their consistency, their moral implications, or more recently which one would reign supreme.  As of now, OpenAI has broken ahead of the pack.

In a move that keeps them ahead of their competitors, OpenAI has announced GTP-4, their most recent AI language model. According to OpenAI themselves, GPT-4 is their “most advanced system, producing safer and more useful responses.”

The newest version of their language model is able to solve difficult problems with greater ease and accuracy. They attribute greater problem solving abilities and a broader general knowledge as the reasons why.

In examples posted on their official website, GPT-4 formulates an impressive passage about Cinderella with the next letter of the alphabet required with each subsequent word, it recognizes ingredients from a photograph and lists many of the possible dishes that could be made from them, and it even explains why Rihanna’s Super Bowl halftime performance was so special.

Other examples of GPT-4’s superiority comes from its advanced reasoning capabilities, and ability to easily pass tests like the Uniform Bar Exam. OpenAI states that by leveraging more data and computations, GPT-4 is able to “create increasingly sophisticated and capable language models.”

One concern about ChatGPT earlier in the year were the reports that users were able to trick the chatbot into writing malicious code for them – essentially meaning that amateur cyber attackers could wreak havoc with AI-generated malware code.  OpenAI claims that GPT-4 is 82% less likely to respond to requests with poor intentions, such as the outrageous requests for it to write malicious code. The AI giant also disclosed that through training with human feedback, improvement from real-world use and GPT-4-assisted safety research, the new language model will act in a much more responsible way than its predecessors.

Currently, you can try GPT-4 on ChatGPT Plus. Users can also join the API waitlist. It’ll be interesting in the next few weeks to see the feedback from those using GPT-4, and to see if it’s as polarizing and divisive as it was at the beginning of 2023.

A Fake Version of ChatGPT is Spreading real Malware

A Fake Version of ChatGPT is Spreading real Malware

New Details in LastPass Breach Investigation Reveal Employee’s Home PC was Hacked

New Details in LastPass Breach Investigation Reveal Employee’s Home PC was Hacked