OpenAI has introduced that ChatGPT will now gently remind customers to take a break if they’re in a protracted dialog with the chatbot. The brand new characteristic comes as a part of the initiative by the corporate to encourage wholesome behaviour with its in style chatbot.
“As a substitute of measuring success by time spent or clicks, we care extra about whether or not you allow the product having accomplished what you got here for,” OpenAI mentioned in a weblog publish asserting the mild reminders characteristic.
A pattern reminder proven by OpenAI reveals the message, “Simply checking in. You have been chatting for a while-is this a great time for a break?” Customers will then have two choices to click on on, both to maintain chatting or take a break by clicking on “This was useful”.
OpenAI says it is going to preserve tuning when and the way mild reminders present up to be able to ensure that they “really feel pure and useful.”
Aside from the mild reminders characteristic, OpenAI additionally acknowledged it is going to quickly roll out a brand new behaviour for ‘high-stakes private choices’ the place ChatGPT will assist customers assume via their massive choices in life by asking questions and weighing execs and cons as an alternative of giving a straight reply.
The brand new options added to ChatGPT come after a June report by The New York Occasions, which mentioned the chatbot’s tendency to agree with folks, present flattery and have interaction in ‘hallucinations’ (making stuff up), led customers to develop delusional beliefs.
The corporate itself had confirmed in April that its chatbot had grow to be overly agreeable and sycophantic after an replace to GPT-4o. This made the favored AI companion align with the opinions of customers, even those that had been factually incorrect or dangerous. Whereas OpenAI had pulled again that replace, the brand new effort now looks like a option to encourage more healthy habits for the customers.
In the meantime, OpenAI CEO Sam Altman has additionally been teasing the corporate’s newest GPT-5 language mannequin. The brand new mannequin is predicted to be launched very quickly, and Altman had earlier revealed that it will likely be the primary ever OpenAI mannequin and not using a mannequin picker. This implies customers won’t want to change between a reasoning mannequin and a GPT mannequin, and the AI system will be capable of make that call by itself.
