ChatGPT was already considered one of the most powerful AI bots in the world, but users could only give it instructions or ask questions via text.
Now the new version of the chatbot has reached a new level: processing images.
This is considered by experts as a breakthrough.
In the new version, called GPT-4, a user can, for example, take a photo with ingredients available in their kitchen and ask the chatbot to offer recipe options and instructions.
But the robot can also provide solutions to more complex tasks. If a user writes on a piece of paper the data that they would like to see on a new web page, GPT-4 can read that and create all the code necessary to generate a complete website.
Millions of people have used the chatbot since its launch in November 2022.
The chatbot answers questions or commands using a natural language, very similar to human, and can also imitate styles of artists, such as writers or composers.
However, its knowledge database is limited to the information that was on the internet until 2021.
Experts have expressed concern that it may one day take over many jobs that are currently performed by human employees.
OpenAI said it had been improving security features for GPT-4 for six months and had trained it with human feedback.
However, he cautioned that it may still be prone to offering unverified data or misinformation to user queries.
«More Advanced Reasoning»
GPT-4, like ChatGPT, is a type of generative artificial intelligence. It uses algorithms and predictive text to create new content based on instructions given to it.
The potential of the tool depends to a large extent on the type of tasks that are requested of it.
Among the most demanded functions of the robot so far is the composition of songs, poems, texts for advertising, software coding and even help with homework, although the latter is something highly rejected by teachers.
In a live demo hosted by OpenaAI, the chatbot generated an answer to a tricky tax question, though there was no way to check the veracity of its answer.
Among the improved features, the chatbot will now be able to process up to 25,000 words, about eight times more than the previous version (ChatGPT 3.5).
The new version has «more advanced reasoning abilities» than ChatGPT, according to OpenAI.
For example, a user asked him to tell him the plot of the classic story «Cinderella», but that each new word began with a letter of the alphabet from A to Z. This type of challenge was not capable of doing the version former.
OpenAI also announced new partnerships with the language-learning app Duolingo and with Be My Eyes, an app for the visually impaired, to create AI-powered robots that can help their users use natural language.
However, like its predecessors, OpenAI cautioned that GPT-4 is not yet completely reliable and can «hallucinate,» a phenomenon in which AI makes up facts or makes reasoning errors.
«GPT-4 still has many known limitations that we are working to address, such as social bias, hallucinations, and conflicting indications,» OpenaAI explained.
The new version will initially be available to ChatGPT Plus subscribers, who pay US$20 per month for a premium service.
The chatbot is already powering Microsoft’s Bing search engine platform. The tech giant has invested $10 billion in the OpenAI firm.
Presentational gray line
