ChatGPT took the world by storm upon launching in November, injecting artificial intelligence into the public zeitgeist and upending just about every industry it could touch. Recently though, some of the fire appears to have worn off.
The chatbot's traffic has begun to slide, dropping 10% in June amid a slew of rising online complaints about ChatGPT's flagging performance; a recent study found that, between March and June, the AI's accuracy has swung hard, dipping by close to 90% in certain categories.
DON'T MISS: The Company Behind ChatGPT Is Now Facing a Massive Lawsuit
In the midst of this, OpenAI rolled out new enhancements to its flagship product that are intended to allow users the option of personalization.
The new update -- custom instructions -- was made available in beta with OpenAI's Plus plan Thursday and will broaden out to all users "in the coming weeks."
Absolutely stunning instability.
— Gary Marcus (@GaryMarcus) July 19, 2023
Who in their right mind would rely on a system that could be 97.6% correct on a task in March and 2.4% correct on same task in June?
Important results.
Anyone planning to rely on LLMs, take note. https://t.co/tGF2xZgJGC pic.twitter.com/P2p4kMecAU
"Custom instructions allow you to add preferences or requirements that you’d like ChatGPT to consider when generating its responses," OpenAI said.
The feature comes, according to the company, as the result of lengthy conversations with users who have cited "friction" around starting each conversation with the AI model completely fresh. After a user inputs their custom instructions, the AI will "consider the instructions every time it responds."
"For example, a teacher crafting a lesson plan no longer has to repeat that they're teaching 3rd-grade science," OpenAI said.
The company went on to warn users that, specifically while this feature is in its beta testing period, the AI model may misinterpret or overlook instructions. ChatGPT, OpenAI said, "can also refuse or ignore instructions that lead to responses that violate our usage policies."
OpenAI disclosed that, unless it is manually disabled, it may collect the data from users' custom instructions in an effort to improve the performance of the model.
More Business of AI:
- Here's the Startup That Could Win Bill Gates' AI Race
- Meet Your New Executive Assistant, A Powerful AI Named Atlas
- Here's the Steep, Invisible Cost Of Using AI Models Like ChatGPT
"Custom instructions is just one small step towards more personalized AI, but a surprisingly fun/useful one," Sam Altman, OpenAI's CEO said.
OpenAI recently announced a $5 million partnership with the American Journalism Project to examine how the company's product can be used to help elevate local newsrooms.
custom instructions is just one small step towards more personalized AI, but a surprisingly fun/useful one. available for plus users now, and all users soon:https://t.co/kwHX80Zc6u
— Sam Altman (@sama) July 20, 2023
Altman has previously been a prominent voice in the case for AI regulation, citing fears over future models that might be outside of humanity's ability to control. Many have disputed the veracity of these claims, especially considering the fact that, despite Altman's apocalyptic fears, he is intent on producing artificial general intelligence (an AI equal to or greater than human intelligence) regardless.
