General news
OpenAI’s ChatGPT crosses 1 million users, CEO says they might have to monetize this at some point
Sam Altman
OpenAI’s ChatGPT, the AI-powered chatbot that has gone viral, has surpassed 1 million users in less than a week since it was officially made available to the public. ChatGPT became available for public testing last Wednesday. Open CEO Sam Altman confirmed this via a tweet. His post also drew questions about whether the company plans to keep ChatGPT free forever, to which he replied that they “will have to monetize it somehow at some point,” adding that the compute costs of running this are “mind-boggling.”
Twitter CEO Elon Musk
Twitter CEO Elon Musk also asked Altman what the average cost per chat is for OpenAI. Altman said this is “probably single digit cents per chat; trying to figure out more precisely and also how we can optimize it.”
Interestingly, Musk also tweeted about how OpenAI uses the Twitter database for training and that he has put this on hold for now. According to him, OpenAI used to be a non-profit, open source company that has changed and “needs to understand more about the governance structure and revenue plans going forward.”
ChatGPT was launched on Wednesday. today crossed 1 million users!
— Sam Altman (@sama) December 5, 2022
we will have to monetize it somehow at some point; computing costs are staggering
— Sam Altman (@sama) December 5, 2022
the average is probably single digit cents per chat; trying to figure out more precisely and also how we can optimize it
— Sam Altman (@sama) December 5, 2022
Computing and Language Lab
Meanwhile, others have also pointed out that ChatGPT is not without its flaws and continues to fall victim to old racist and sexist biases. Steven T Piantadosi, a UC Berkeley professor who runs the ‘Computing and Language Lab (colala)’ at the university, wrote that while ChatGPT is amazing, it’s not without its biases.
He posted a thread showing the problems with the chatbot. By the way, Altman also responded to the larger thread and asked users to “thumbs down on these” replies which are offensive and help improve the AI.
Check out his tweet and Altman’s response below.
Yes, ChatGPT is amazing and impressive. No, @OpenAI hasn’t come close to addressing the bias issue. Filters seem to get bypassed with simple, superficially masked tricks.
And what lurks within is appalling. @Abebab @sama
two racism, sexism. pic.twitter.com/V4fw1fY9dY
—Steven T. piantadosi (@spiantado) December 4, 2022
please thumb down on these and help us improve!
— Sam Altman (@sama) December 4, 2022
What is ChatGPT and how do you sign up?
ChatGPT is the OpenAI conversational chatbot that can respond almost like another human would; many say that ChatGPT could soon be writing emails, articles, code, how-to guides, and even college essays. Once a user signs up to ChatGPT, they can use the chatbot to have a conversation, and the expectation is that it will give you reasonably intelligent responses in essay form. Some have also used it to write fiction, although, in our experience, this is one of the limitations for now.
Try Now
A user can simply go to the OpenAI website and tap on the Try Now option next to the ChatGPT banner right at the top. Or you scroll down to ChatGPT and tap on it. You will need to register and create an OpenAI account.
Please note that you may not be able to register at this time. We tried to sign up with a new account and it said “ChatGPT is currently overloaded. Please check back later.”
If you manage to log in, you will be able to see the interface of the chatbot. Getting started, ChatGPT has some examples, capabilities, and caveats listed for you. For example, examples include the type of questions you might ask, such as “explain quantum computing in simple terms” or “do you have any creative ideas for a 10-year-old’s birthday?”
ChatGPT can write entire essays based on simple instructions.
The ‘Capability’ tab mentions that the chatbot can remember what the user “said earlier in the conversation” and allows a “user to provide follow-up corrections”. It is also trained to reject inappropriate requests. Its limitations are that it can occasionally generate incorrect information, occasionally produce harmful instructions or biased content, and limited knowledge of the world and events after 2021.
Reinforcement Learning
According to OpenAI, this model was trained with “Reinforcement Learning from Human Feedback (RLHF)” and is similar to a previous model they created called InstructGPT.
ChatGPT and GPT
“We trained an initial model using supervised fine-tuning: AI human trainers provided conversations in which they played both sides: the user and an AI assistant. We gave trainers access to template-written hints to help them draft their responses. The blog post adds: “ChatGPT is fine-tuned from a model of the GPT-3.5 series, which finished training in early 2022. ChatGPT and GPT 3.5 were trained on an Azure AI supercomputing infrastructure.”
The post also acknowledges that there are limitations with the chatbot. This means you could write “plausible-sounding but incorrect or nonsensical answers,” and this remains a challenge the company hopes to address. It is also “sensitive to adjustments to the input wording or trying the same prompt multiple times. For example, given a question phrase, the model may claim it doesn’t know the answer, but given a slight rephrasing, it may answer correctly.” You can also abuse certain phrases due to “biases in training data…”
