Jetpack AI – your writing assistant on WordPress.com
We have introduced 2 new blocks for WordPress.com customers in partnership with OpenAI:
- `AI Paragraph` will auto-complete your post. Insert the block to generate a new paragraph from your existing content, or craft an example post from only a title
- `AI Image` will let you create new images based on the text prompt.
You can find those blocks by searching for
This is an experimental feature.
AI models and tools are a bit unpredictable. We want to be mindful of how those tools behave, what impact they have on your creativity, and how we can make them support you.
At the same time, we are still learning and adapting, so please keep in mind that these features can change or disappear anytime.
Once we announce them, they are most likely be paid extra (on top of existing plans) depending on how much we have to pay OpenAI for their services.
Your feedback is appreciated.
Please share any feedback in this thread. You have the opportunity to shape the future of this feature and as I said – I would love this feature to be as much help to you as possible.
I think it’s fantastic, but it doesn’t allow you to amend or add words.
Who owns the rights to the images created?
it doesn’t allow you to amend or add words.
The generated content can be edited and tweaked by transforming the block into a regular paragraph. The transforms menu is accessible by clicking the block icon on the left side of the block menu bar:
We were considering making the block outright editable, and I see that we may need to create a more intuitive editing experience.
Who owns the rights to the images created?
Currently, it is a beta feature and I did not consult our lawyers, so the following is not a binding legal agreement:
A formal policy will be formulated once this feature is commercially available and we start charging for the API use. We will most likely pursue the wording that assigns all copyright to the site owner. We will most likely treat these images in a similar way as those uploaded by the site owner and may reserve some legal rights for compliance in certain regions. This would be a policy similar to one adopted by OpenAI:
OpenAI hereby assigns to you all its right, title and interest in and to Output. OpenAI may use Content as necessary to provide and maintain the Services, comply with applicable law, and enforce our policies. You are responsible for Content, including for ensuring that it does not violate any applicable law or these Terms.
I loved what it put in, but I wasn’t sure what the generated content was being based on. From your post here, I see it is based on something just previous, like maybe a section? I’m still not exactly sure.
Also, it seemed to peeter out, as in just stop mid-sentence without finishing its thought.
but I wasn’t sure what the generated content was being based on. From your post here, I see it is based on something just previous, like maybe a section? I’m still not exactly sure.
AI Paragraph will try to grab all the data available in the editor to create an AI prompt:
- Categories and tags
- Content of the post.
For AI it’s overall its very good, sometimes it comes up with images that are a bit wobly and other times it’s spot on. If you don’t like it try it again.
It would be better if it offered more images
This censorship is unacceptable.Latino America
When I type in a description for AI image and it brings up the 3-4 images, there doesn’t seem to be any way to start over without removing the entire block and re-adding it with a new set of parameters. What if I don’t want to select any of the pictures shown and want to try again? It doesn’t seem very smooth to have to add the block all over again. Am I missing something?
On the AI paragraph, there needs to be something that is MUCH more intuitive for changing it to a paragraph block to be editable. In fact, it should be editable from the get go. Since it doesn’t necessarily even finish a sentence, as seen on this post:
The regular font is the AI paragraph and the italics are human. You can see that the AI simply stops midway through the sentence, so needing to edit is obvious.
Hey y’all! If you want customer feedback on this experimental feature, wouldn’t it be good to make this a sticky post in the forums? #justsaying
Nice feature! I love it, but would be great if it allowed to refresh the prompt and regenerate images.
As a user of OpenAI, I was really delighted to see this feature added and immediately recognized it as their work. However, what protections are in place to prevent abuse and stay in line with OpenAI’s mission to make AI safer to use? That is, are these connections between WP and OpenAI straight to the pre-existing AI, and thus, for example, inappropriate prompts will not be generated? Also, part of OpenAI’s research involves users giving feedback about incorrect generations or bias that hasn’t been removed yet. Will this feedback channel be available as well? Lastly, I did a test of the AI image generator, and noticed there was no mention of the AI generation in the caption or alt-text generated. That is part of the transparency discussion occuring with AI (and part of OpenAI’s terms, I believe), and it would be good to have generated images also include that critical information (in the description and alt-text for images, and as a byline for paragraphs). Are there plans, if this test feature works out well, to have that added?
Again, delighted to see this feature, but would like to know about the nuanced things and how it relates to OpenAI’s mission, and how we can contribute to growth of more ethical, more accessible AI. Cheers!
what protections are in place to prevent abuse and stay in line with OpenAI’s mission to make AI safer to use?
Thank you for this comment! We actually put a lot of thought into that issue:
- We construct the prompt in Gutenberg from post content to prevent prompting the model explicitly. That limits a number of scenarios where API could be abused
- We filter all the prompts through our own system of abuse detection.
- Before we submit the prompt to GPT, we run it through OpenAI Moderation API, to make double-sure the prompt is inspected
- OpenAI is independently including chatGPT feedback data to the model we are using
- Once you publish the post, we have automated systems in place to catch abuse in published content.
Still, this is cat-and-mouse game, and I am sure there is a scenario where somebody can generate disturbing content, just as with a human contractor.
As for the societal impacts of OpenAI’s mission, it’s hard for me to judge. As our mission is to “democratize publishing and make the internet a better place”, we feel like that is enough for now :).
are these connections between WP and OpenAI straight to the pre-existing AI
Yes, in case of the paragraph block we are using completions API with the
text-davinci-003model, which is a little different from chatGPT.
part of OpenAI’s research involves users giving feedback about incorrect generations or bias that hasn’t been removed yet. Will this feedback channel be available as well?
That channel indeed exists in chatGPT. However, OpenAI does not provide such a channel in the completions API, and we have no way to funnel this feedback back to OpenAI.
That is part of the transparency discussion occuring with AI (and part of OpenAI’s terms, I believe), and it would be good to have generated images also include that critical information (in the description and alt-text for images, and as a byline for paragraphs). Are there plans, if this test feature works out well, to have that added?
I personally absolutely agree that in the future, we will need ways to distinguish AI-generated content and wrote about this before. I do hope that, over time, the industry can settle on a web standard to do so. In preparation for this future, we do store the information about the AI origins of the content in the database, so we can surface it on the frontend if needed.
Tried it at https://www.mymaringop.org/
Neither one work in Edge, Brave or Chrome on a Windows 10 box, nor Safari on an M1 Mac Studio box.
I want my AI! How do I get it working?
@phraingck What do you mean “neither one work”? Does the block not appear when you try to select it or does it do nothing? Or something else?
Okay, it’s not for post or page viewers. I now know that input is needed while creating a post or page. The AI image works okay with simple input, as does the AI paragraph. Having both in the same post or page doesn’t work out that well.
I’m really impressed with this functionality, got it working without any problems whatsoever, shared my experiences in this post which then became a bit of a critique. But not on the technology per se – on that front, top job and well done 💪Looking forward to incorporating this into my blogging in future.