uk1000

If it's trained on lots of AI generated content that might contain incorrect statements then it probably will get worse the more it gets trained on incorrect stuff. I'm guessing normal hallucinations can be reduced by changing things like "temperature, top_p, and top_k," (though I haven't used chatGPT - and they might be only in the API - according to perplexity.ai) where temperature controls how random it is. So reducing the random part of it might make it less creative but might also reduce incorrect statements/hallucinations They could be doing more to validate it to make sure it's trained with accurate info and especially not AI generated misinformation, but also checking the content created after the prompt is entered against proper sources.

They can scan more stuff in the message itself (eg. new buyers telling people to "click this QR Code" or tell people "Your account is currently suspended. To restore your account...". There will be quite a few words/phrases they can add to their list to check for. If there's no URL in the message but there is a QR code image, Fiverr could also check the images and if any contain QR codes they could check what they contain and what URL they might point to. There are sites which analyse a QR code (without someone having to scan it). I'm sure Fiverr could do the same with some code/using an API. This is a bit of code perplexity.ai suggested for Python. Fiverr could use something like this to check images which might contain a QR code and see what URL they point to (I haven't checked them so don't know how correct they are): Another way it suggested:

And while it's now been fixed on the Fiverr site (it no longer gets put in the page source), nothing has been done about the Fiverr seller pages stored in archive.org from when the profiles were changed to the new format until before it was fixed (in late November last year). Maybe Fiverr could contact archive.org (or any other sites like that) to ask them to remove those captures if necessary.

They did though for a while when the new profile format got displayed. They shared the "full name" field in the page source, even though it's in the "settings" menu and it doesn't show that that will be in the public part of the profile (above the screen that shows the "full name" and email it says "Need to update your public profile? Go to My Profile) - indicating the "full name" field wouldn't get displayed/accessed publicly (and that for changig what will be shown/accessed publicly you'd need to go to a different place, the "My Profile" page to change them), even though, before I told CS about it, it was). Fiverr also automatically enabled the option to make seller plus members see your past average purchase info and most frequently bought category. That was enabled for all users without informing them. You had to go to the bottom of the profile page to turn it off. Now there's no way to turn it off (so it's giving out what could be private info with no option to disable it), so all seller plus members you contact will be able to see a user's average purchase info and most frequently bought subcategory without their full knowledge or consent (sellers might be able to guess that that might be shown by the advert for seller plus features, but buyer-only users probably aren't told about it). Also there's a lot of tracking info on the site that the user doesn't really opt in to. Using Fiverr Neo or when the Grammarly option that was there might be/have been sending info to certain other companies (through APIs). The privacy policy says they send stuff to 3rd parties (we have no Fiverr option over which 3rd parties they send out info to). There's a "do not track" option in some browsers. Fiverr say "We do not honor browser requests not to be tracked online". Why would a company say/do that?

by not doing more to block people putting their display name as "Fiverr Support" (when it's happened again today (edit: or maybe 3 days ago, which is still a long time after someone else reported the same display name being used for a scam), when it's already been stated to them that the fraudsters are putting their display name as that), attaching Fiverr logos to their messages and asking new users to "click on" (or scan) QR codes they're not doing the best they can to avoid people getting contacted scammers trying to take their info, and they're not really educating people enough about this in case their systems can't detect those users. Fiver already scans links, surely an added check could be made to see if it's a url that starts with "fiverr." but goes to a totally different site than Fiverr (which is probably a scam site so Fiverr's system could flag the message and not show it to other Fiverr users unless checked and accepted by the trust & safety team).

If someone used Midjourney or similar to create images for their delivery, how is that affected by the community standards page: https://www.fiverr.com/community/standards/ai-generated-content where it says: According to perplexity.ai: So based on that, the seller wouldn't have the copyrights to what they create from creating prompts in MidJourney (though the AI generated content section of the community standards says sellers must have those rights (all rights to what they create, including copyrights). Though they would probably have copyright of the other parts (but not the AI generated parts as far as I know), at least if there wasn't enough of human authorship after creating the images. Also is the risk of legal problems from using image generators like MidJourney (given that they, like Stability AI were sued for copyright infringement) negligible so is it not a risk really for any Fiverr seller from that?

When I checked the review it it doesn't mention "buggy" or bugs (the review from a week ago). Though maybe it's been changed since then. If it did lie then it might be covered by the following link https://www.fiverr.com/community/standards/violence-unlawful but it might really be about the sections they mention further on in that link, about spreading fake news etc.

Not all of them. And it's changed over the years. eg. HTML will stay as HTML if you enter it like that. so will CRM, UI and UX and SEO and KDP. Also VA stays capitalized, so does HQ, UK, US, USA, VR, CSS, HD and SAS, BBQ, PPC. PSD, CV, AWS, HR and PHP also stay capitalised. Maybe it's all the ones where it's not just the 1st letter that's capitalised stay capitalised. The word "I" also can stay capitalised (after the 1st I). It seems like British used to stay capitalized up until sometime in 2013 (or about then) but since then they're all lower-case. You could suggest it to Fiverr customer support (for it to keep the formatting, at least of a list of words like "English") since it does look bad, especially in the writing & translation category.

It's a scam. You should report the message to Fiverr by clicking on the report option above the message. Also they say "Click on the QR code" - wouldn't someone scan a QR code not click on it? Maybe they've changed their wording to try to get through Fiverr's filters. It's been posted by others on the forum that this is a scam too. It actually goes to a site that starts with "activate-" that is not Fiverr and is not created by Fiverr.

It wasn't suggesting it for Fiverr though. He was asked: And he talked about a medical AI that he worked on before ChatGPT was around and that was simpler, and used decision trees, regression etc. he said (according to Whisper AI):

Probably it had to be manually checked because the quoted part had the word that starts with "spe" and ends with "ll" which certain users used to promote stuff. But if each time that is used it gets flagged then it's harder to talk about normal spellling checking of documents etc. I didn't see it live but it seemed okay from recording of it. It could have shown more detail on things like showing doing AI stuff. I think was mostly discussing things instead and using pre-done info screens. eg. no realtime stuff done using AI. I don't think they talked much/showed much about open source AI either (their prompts shown were for ChatGPT mostly I think). You can watch the recording of it here: https://help.fiverr.com/hc/en-us/articles/24321964769169-How-to-make-AI-your-business-partner-Webinar And it talked a bit about their search engine results (gig list) ranking using AI for ranking it for each user (buyer). Though Fiverr's community standards page says: So I'm not sure if that would be allowed if it was ChatGPT pulling the leads from social media (wouldn't that also be scraping?).

I think each person who wants gig advice is supposed to create their own thread for it but here's a few suggestions/comments about your gig: In the gig description I'm not sure about the way most of the first part is highlighted. Maybe too much? The main gig image says "WORK ON MORE CLIENTS". That might be better rephrased. Another seller says "Works on all major email clients" (don't copy that but maybe some other phrasing could be used). You say "Unlimited Revisions" (and on the gig image "unlimited revision") but offering unlimited revisions can lead to buyers constantly asking for revisions (much too many) but you aren't getting any more $ for it. So it might be better to offer a fixed amount of revisions instead. In the FAQ section: Maybe reword "What I Need To Start The Project?" eg. "What do I need to..." One question says "Will you Host the Images?" And the answer is "Yes, I will host your Image." (they asked about "images" and you replied about an "image" (singular)). Though I assume there could be some limitations if you're hosting them. In the question "Will this Signature works on iPhone?" I'd change it a bit so it reads better grammatically.

If Fiverr added a "send draft" option that stopped the clock until the buyer responded it might be helpful. If you send drafts and they don't respond for days it could make your delivery go late in the current Fiverr system or they might keep asking for lots of revisions on the drafts sent (before the actual delivery is sent) which could also cause the delivery to go late (unless an extensions is made, which might not be accepted in time to stop it going late). Though it would also be better if Fiverr stopped the revisions making an order late if they keep asking for revisions (it didn't used to work like that. It used to be if the initial delivery was on time it wouldn't be counted as late if they kept revising it). So I communicate enough (which seems to increase success score even if other things are lower), including for things in the inbox if an order has completed but they still wanted revisions and still had free ones, but I don't send a lot of drafts.

These are my suggestions/comments: Gig: I will do data entry data mining excel and data extraction In the gig title you say "I will do data entry data mining excel and data extraction". If you are going to say that it would be better with some punctuation (commas). In the gig description: Maybe change "i offer" (capitalisation) Maybe change "more! . I am" (punctuation/spacing) It might better not to say "24/7 availability" as it sounds too unrealistic. Maybe change the spacing in "or ongoing support,I am here". You have 2 search tags. Maybe some more could help the gig be found (though it doesn't always show all the ones your gig is using on the page). Your gig video: It has an image with an "alamy" watermark on it. It would be better if it didn't. The gig video thumbnail says "EXCEL DATA CLEAING". I'd change that

If the answer to some of the questions could be something you parents/family might know (and you haven't written them down anywhere) could you ask them?

Remember how there were disagreements on the forum about whether there was a ranking system in Fiverr for search results/gigs. eg. and in the latest Fiverr webinar (How to make AI your business partner), Fiverr say: See: https://help.fiverr.com/hc/en-us/articles/24321964769169-How-to-make-AI-your-business-partner-Webinar from 50:49 in that webinar. So this helps prove that their is a search results (which is the gigs list) ranking system in Fiverr.

The community standards page at https://www.fiverr.com/community/standards/prohibited-services says this: Your gig was about managing their Forex trading account and the community standards page says do not offer "Opening, managing, or implementing changes to a financial account of any kind".

I assume it should be okay for any forum member to post their disagreements about AI as long as it's done in an okay way and being careful how it's said - nothing against any forum rules etc.

Try pressing Enter after each search tag you enter (if that's the section you mean).

This is what the terms of service at https://www.fiverr.com/legal-portal/legal-terms/terms-of-service says: So if it was definitely required for the order (eg. for it to be put onto a business card that the seller was designing) then it would be allowed if the buyer entered it on the order page).

If Fiverr wants to empower AI journeys shouldn't they add a relevant AI section on the forum where there's FAQs and a section where staff would also answer questions related to AI and Fiverr? Maybe text resources could be put there too. Fiverr did create an AI community club (though I don't think it had much resources etc. but it could have added it) and then deleted every club on the forum and now says "we want to empower your AI journey" but the AI section/club has been deleted and there's no replacement section with the resources related to AI and Fiverr (apart from the videos on the "Sellers AI hub" page) - no text info. edit: there are a few text links on that page too but not much and not much about any open source stuff. eg. Fiverr could go into the copyright related stuff (like has been mentioned - with the community standards saying we need to have copyright in the AI generated output but but other places say AI generated output can't be copyrighted - at least as they are) and copyright related to the AIs training data.

Isn't a problem with that that the client's device might not be on or they might have no connection to it? Plus if it was constantly streaming it couldn't take bandwidth? And it might occasionally have the wrong time on their device. If the buyer has turned their device(s) off their other methods (eg. using the geolocation/IP address/saved country settings (though some countries have >1 timezone) could work (eg. it could still say the time in the buyer's timezone even if the buyer's device(s) were turned off).

You could do the poll based on Fiverr level or orders per month, success score or monthly Fiverr earnings or some other non-country thing and make the results private so it could be more valid.

Couldn't it be useful in game development though? According to perplexity.AI which referenced zipdo.co in their "ai in video games section":

You said you could demonstrate it, so if you showed proof that in some system what you said could be the case and it was accurate proof then it should be believable. Maybe there could be some universe where 8 of something in one could be equal to 4 of something in another. For the Pi thing, if you had some number system (not the standard ones) that started at 0 and counted up in half Pis it could be that Pi (in our system) could equal 2 in that system. So yes you could in theory create some number system in a programming language that counted in half Pis (eg. if it made things easier, like using degrees or radians in them might make some thing easier).