by
The GPT-4 language model is now available for ChatGPT Plus subscribers. Developers can join an API waiting list.
For search engines and business writing help, the leading contender 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. It throws a new weapon into the AI war, in which organizations are racing to provide the best, most versatile writing AI.
Jump to:
GPT-4 improves reasoning abilities
OpenAI introduced the new natural language model with a challenge: “Explain the plot of Cinderella in a sentence where each word must begin with the next letter of the alphabet from A to Z, without repeating any letters.” It’s a neat puzzle to show that AI can perform some reasoning along with producing plain text, but what does it do in the office?
Creative and technical work is on the table for GPT-4, OpenAI said in their announcement. The new model can match an individual’s writing style as well as take cues about voice and tone. Some of the GPT-4 answers in the demo are simpler, leading Occam’s Razor to programming problems. The new nuance comes in part from training it on a custom Microsoft Azure AI supercomputer.
Specifically, GPT-4 is referred to as a large multimodal deep learning model, meaning it accepts image and text inputs and generates text outputs. Note the distinction between the various iterations of OpenAI’s product. ChatGPT is the popular chatbot based on GPT-3. Meanwhile, GPT-3.5 is the first draft of GPT-4 and started training 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 GPT-4’s progress by putting both the GPT-4 and GPT-3 through a variety of academic tests, such as those administered at the end of high school AP courses or the Uniform Bar Exam, and the GPT- 4 generally scored higher. More information on this can be found in the full technical report.
How OpenAI works for more ‘real answers’
One of the common criticisms of natural language AIs like this is that the results they produce tend to look like human speech, but they aren’t based on the actual facts of the content – they don’t check for accuracy. OpenAI seems to be aware of this, noting that GPT-4 is “40% more likely to produce real answers than GPT-3.5 in our internal evaluations.”
GPT-4 is also 82% less likely to “respond to requests for disallowed content”. Disallowed content includes hate speech, profanity, threats of harm, or other conversational topics not appropriate for the workplace that the model may have picked up from the online text they were trained on. Also, high-risk government and law enforcement decisions are not officially allowed.
To avoid these, OpenAI took advantage of feedback submitted by ChatGPT users. it also employed AI experts in the fields of safety and security. However, OpenAI notes that one of the major problems with the model is still its tendency to spit out “social biases, illusions and counterintuitive prompts.”
Interestingly, part of this process involved AI itself.
“We used GPT-4 to help generate training data to refine models and iterate on classifiers throughout training, evaluations, and monitoring,” OpenAI wrote.
OpenAI is also releasing open source code for OpenAI Evals, the framework for evaluating AI performance, for anyone to review its benchmarks and report issues.
What does GPT-4 mean for business leaders?
Business leaders may want to decide how much and which natural AI service language to allow or encourage their employees to use. So far, ChatGPT has solicited feedback from a number of companies that use its products, including language learning app Duolingo, visual accessibility app Be My Eyes and wealth management firm Morgan Stanley.
Microsoft has been running GPT-4 behind the scenes on the Bing search engine for about five weeks.
Is GPT-4 a great leap forward or an old idea in new clothes?
Google and Microsoft are going head-to-head when it comes to adding artificial intelligence to search capabilities. Google’s Bard AI will soon be available to a test team, while Microsoft’s Bing has been working with ChatGPT since February. Anthropic, an artificial intelligence company backed by Google dollars, 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 board or wait? Perhaps it depends on what the AI assistant could be used for. Where could it save time or where could it hinder? After all, some of these features are starting to sound like a leaner, more resource-intensive version of Microsoft’s long-lost and multi-membered Clippy.
