OpenAI is making its flagship conversational AI available to everyone, even those who have never bothered to create an account. However, it won't be exactly the same experience. Of course, all chats are included in your training data unless you opt out.
Starting today in some markets and gradually rolling out to other parts of the world, you will no longer be prompted to log in when you visit chat.openai.com, but you will be able to log in if you wish. It is possible. Instead, you can immediately join his ChatGPT conversation, which uses the same model as the logged-in user.
You can chat to your heart's content, but keep in mind that you won't have access to the exact same feature set as users with accounts. You will no longer be able to save or share chats, use custom instructions, or other features that would normally be associated with a permanent account.
That said, you still have the option to opt out of your chat being used for training (which some wonder if it undermines the whole reason the company does this in the first place). Just click on the little question mark in the bottom right corner, click on “Settings” and disable the feature there. OpenAI includes the following useful gifs.
Image credit: OpenAI
More importantly, this additional free version of ChatGPT comes with a “slightly more restrictive content policy.” What does that mean? When I asked, I received a verbose but mostly nonsensical response from a spokesperson.
Signed-out experiences benefit from existing safety mitigations already built into the model, such as refusing to generate harmful content. In addition to these existing mitigations, we have also implemented additional safeguards specifically designed to address other forms of content that may be inappropriate for signed-out experiences.
Based on our understanding of the functionality of GPT-3.5 and the risk assessment we completed, we considered ways in which logged out services could be used in inappropriate ways.
So…actually, I have no clue what exactly these more restrictive policies are. We have no doubt we'll find out soon enough, as Randos will be hitting the site avalanche to try out this new product. “We recognize that additional iterations may be required and welcome feedback,” the spokesperson said. And they will have it in abundance!
In that regard, I also asked whether there is a plan for how to deal with attempts to exploit and weaponize the model on what will almost certainly be an unprecedented scale. please think about it. It is the use of the platform that causes billionaires to lose money. After all, inference is still expensive, and even sophisticated low-lift GPT-3.5 models require power and server space. People are going to slam it as hard as they can.
They did not respond to this threat with any subtlety.
We're also thinking carefully about how to detect and stop abuse of the Signed Out experience, and the team responsible for exploit detection, prevention, and response has been involved throughout the design and implementation of this experience and will continue to do so. will also provide information on its design. We are moving forward.
Notice that there is nothing that looks like concrete information. They probably understand less than anyone else about how people are affected by this issue, and will be forced to be reactive rather than proactive.
It's not clear which regions or groups will be the first to get access to super-free ChatGPT, but it starts today, so check back often to see if you're among the lucky ones. please.