Would you trust an artificial chatbot to do your taxes? Well, Open AI — the company behind the revolutionary ChatGPT program — says its latest version can do that and so much more.
Last year, ChatGPT made a huge cultural splash as its text generator provided conversational, almost human-sounding responses to questions and prompts. This led to some concerns that AI will soon be eliminating the need for various human jobs … and GPT-4 is adding fuel to the fire.
Here are some of the advancements the latest version has over its predecessor:
- ChatGPT scored in the 10th percentile on the bar exam while GPT-4 scored in the 90th percentile.
- GPT-4 is significantly less likely to provide a response to an inappropriate prompt and more likely to provide trustworthy results.
- The chatbot can now generate entries of up to 25,000 words and create captions based on a submitted image.
Only paying ChatGPT Plus members will be able to access the newest platform for now, but OpenAI is incorporating it into products offered by partner companies including Duolingo and the Icelandic government.
Don’t expect perfection
There’s no doubt that GPT-4 is a powerful tool, but even its developer acknowledges some shortcomings that OpenAI plans to remedy.
For starters, it’s possible for the chatbot to “hallucinate,” which is what the cool kids call it when the program spews out a bunch of nonsense that might seem credible — at least at first glance.
The technology is also inherently vulnerable to those who seek to exploit its safety features and can reflect biases that it picks up from the reference material used to generate responses.
Nevertheless, it looks like chatbots are here to stay and will soon be integrated into all sorts of programs that we use online.