Four months after the groundbreaking ChatGPT was released, the company behind it announced GPT-4, its “safer and more aligned” successor.
ChatGPT’s creator, OpenAI, describes it as a model with the “best-ever results on capabilities and alignment,” and has spent six months developing this improved version, which promises more creativity and less likelihood of misinformation and biases.
However, the company warns that it is still susceptible to “hallucinations,” which refer to the chatbot’s tendency to make up facts or respond incorrectly.
Given that artificial intelligence (AI) bots learn by analyzing large amounts of online data, ChatGPT’s failures in some areas, as well as its users’ experiences, have contributed to making GPT-4 a better and safer tool to use.
“GPT-4 is more reliable, creative, and capable of handling much more nuanced instructions than GPT-3.5,” according to OpenAI.
What’s changed and what improved?
OpenAI asserts GPT-4 is more creative in terms of generating creative writings such as screenplays and poems, as well as song composition, with an improved ability to mimic users’ writing styles for more personalized results.
GPT-4 is also referred to as a “multimodal” model, which means that it can accept text and image inputs.
The new model can respond to images – providing recipe suggestions from photos of ingredients, for example, as well as writing captions and descriptions.
It can also process up to 25,000 words, about eight times as many as ChatGPT.
As a result of analyzing image components, it will be able to generate captions and provide responses.
Perhaps more impressively, OpenAI’s new system can now ace various standardized tests thanks to its new advanced reasoning abilities.
Previous versions of the technology, for instance, weren’t able to pass legal exams for the Bar and did not perform as well on most Advanced Placement tests, especially in maths.
Read more: The magic behind the machine – Artificial intelligence in the spotlight
Limitations
Although the improved version of the chatbot appears to have more features, GPT-4 is still plagued by “hallucinations” and is prone to making up facts.
While according to OpenAI, GPT-4 scores “40 percent higher” on tests that measure these hallucinations, the company admits “GPT-4 still has many known limitations that we are working to address, such as social biases, hallucinations, and adversarial prompts”.
Other limitations have included the inaccessibility of the image input feature until now. While it’s exciting that GPT-4 will be able to suggest meals based on a picture of the ingredients, this technology isn’t yet ready for public use.
Furthermore, GPT-4 will initially be available only to ChatGPT Plus subscribers, a premium service that costs $20 per month.
AI is everywhere, including crypto
Generative AI tools like ChatGPT and Midjourney have prompted debates for AI’s inclusion across business and technological spaces, and AI has even reached the crypto world. As a result, many AI-based crypto projects are now gaining investors’ attention.
The link to AI has triggered positive price action for several cryptocurrencies, indicating the rise of a new trend in the crypto market. Some investors believe that AI coins may reach the heights of previous popular trends like the DeFi boom, the explosion of Metaverse coins, or meme coins.
AI-token prices surge
AI coins or tokens are digital assets used for AI-based crypto projects. These coins allow artificial intelligence to be integrated into varying projects, like asset management, price predictions, AI-based DAOs, and fraud detection. Like most crypto tokens, AI tokens are also used to offer governance rights or rewards to users. In addition, these tokens can also be used for paying transaction fees.
For more on AI, click here.
