The GPT-4 language model is now available for ChatGPT Plus subscribers. Developers can join an API Waitlist.
For search engines and enterprise writing assistance, the main competitor is OpenAI, which yesterday announced the latest version of its language model, GPT-4.
GPT-4 is now available on ChatGPT Plus and as an API, for which developers can join a waiting list. This launches a new weapon in the AI war, in which organizations are jostling to provide the best and most flexible writing AI.
GPT-4 improves reasoning skills
OpenAI demonstrated the new natural language model with a challenge: “Explain the plot of Cinderella in a sentence where each word must start with the next letter of the alphabet from A to Z, without repeating any letter.” This is an interesting riddle to show that AI can perform reasoning while producing simple text, but what is it doing in the office?
Both creative and technical tasks are on the table for GPT-4, OpenAI said in its announcement. The new model is able to match a person’s writing style and take voice and tone instructions. Some of the GPT-4 answers in the demo are simpler, leading Occam’s Razor to scheduling issues. The new nuance comes in part from his training on a custom-built Microsoft Azure AI supercomputer.
Specifically, GPT-4 is considered a great multimodal deep learning model, which means it accepts image and text inputs and creates text outputs. Note the distinction between OpenAI’s different iterations of its product. ChatGPT is the popular chatbot based on GPT-3. Meanwhile, GPT-3.5 is an early draft of GPT-4 and began its formation about a year ago.
SEE: How Salesforce uses ChatGPT (TechRepublic)
The exact difference between the capabilities of GPT-3 and GPT-4 can be difficult to gauge. By OpenAI’s own admission, the difference is “subtle”. OpenAI tracked the progress of GPT-4 by subjecting it to both GPT-3 and a variety of academic tests, such as those administered at the end of AP high school courses or the Uniform Bar Examination, and GPT -4 generally got higher scores. You can find more information about this in the full technical report.
How OpenAI is working towards more “factual answers”
One of the common criticisms of natural language AI like this is that the results they produce tend to sound like human speech, but aren’t based on the actual facts of the content – they don’t verify accuracy. OpenAI seems aware of this, noting that GPT-4 is “40% more likely to produce factual answers than GPT-3.5 on our internal ratings.”
GPT-4 is also 82% less likely to “respond to requests for unauthorized content”. Prohibited content includes hate speech, obscenity, threats of harm, or other non-workplace conversational topics that the model may have picked up from the Internet text on which it was trained. “High-risk government decision-making” and law enforcement decisions are also officially prohibited.
To avoid them, OpenAI took advantage of comments submitted by ChatGPT users; it has also employed artificial intelligence experts in the areas of safety and security. However, OpenAI notes that one of the model’s main problems is still its tendency to spit out “social biases, hallucinations, and contradictory prompts.”
Interestingly, part of this process involved the AI itself.
“We used GPT-4 to help create training data for fine-tuning the model and iterating on the classifiers through training, evaluations, and monitoring,” OpenAI wrote.
OpenAI also releases open source code for OpenAI Evals, its AI performance evaluation framework, for anyone to review its criteria and report issues.
What does GPT-4 mean for business owners?
Business leaders may want to decide how much and which natural language AI service to allow or encourage their employees to use. So far, ChatGPT has sought feedback from various companies that use its products, such as language learning app Duolingo, visual accessibility app Be My Eyes, and wealth management firm Morgan Stanley.
Microsoft uses GPT-4 behind the scenes of the Bing search engine for about five weeks.
Is GPT-4 a big step forward or an old idea in new clothes?
Google and Microsoft are clashing when it comes to adding AI to search capabilities. Google’s Bard AI will soon be made available to a test group, while Microsoft’s Bing has partnered with ChatGPT from February. Anthropic, a generative AI company backed by Google money, has also entered the fray with a natural language model called Claude.
With Google leading the pack but also feeling the pressure, it’s hard to say where the AI trend will go next. Is it up to your business to get on board or wait? Maybe it depends on the use of the AI assistant. Where could he save time, or where could he get in the way? After all, some of these features are starting to look like just a more flexible and resource-intensive version of Microsoft’s lost and much-loved Clippy.