I stand corrected. AI has a long way to go. @Undisciplined
reply
Must be one of those AI hallucinations I have heard about. @siggy47
reply
This agent is far too complimentary. @k00b
reply
50 sats \ 2 replies \ @k00b 22 Jun
Definitely hallucinating.
reply
I am going to see if I can get it to trash talk you guys now.
reply
Interesting. When prompted, it happily trashed Siggy and Undisciplined's teams but then commended them for sticking with their loser teams.
"Fun and slightly wild" going in the bio.
reply
I might need to update the training data to include "KAT's number one fan"
reply
I don't see that anywhere. Let's not start feeding in misinformation on purpose.
reply
I didn't train it on any SN posts. I just created a two page PDF with information on SN, stacker sports and all the stackers that regularly post in the territory or participate in the contests. So I can just add KAT's number one fan to the blurb about you.
reply
This is actually really impressive considering how little data I gave it.
reply
Do as you will, but the world will know of your deception.
reply
I am baffled. What is this?
reply
I have been playing with OpenAgents for a couple days. That's my Stacker Sports agent.
I just wrote up a two page PDF about Stacker News and some of the stackers that contribute to Stacker Sports (i.e favourite teams and what they post in the territory). I haven't trained it on any SN data at all. This is just from the PDF I provided it.
reply
Very cool. It could be fun
reply
Unfortunately right now it is based on GPT-4 so its sports knowledge only goes until 2021. I would prefer it didn't use ClosedAI but it is giving better results with that as its model. I will be testing some of the other models as well.
reply
50 sats \ 1 reply \ @siggy47 22 Jun
Did you see Vorhees on WBD regarding Venice? Could you use that?
reply
No, I am behind on Bitcoin pods. That is not one of the models offered, maybe eventually they will expand what you can use. There are a bunch there. I have tried a few. I am going to try them all at some point.
reply
Where's mine? I thought, I've just become more famous with all those daily Cricket pools!!!! Haha!! Ask what your agent thinks about me? It's most probably gonna reply I can't say anything against 'coinsreporter'. Iam afraid of him ...
reply
It said nice things about you. I just didn’t post them all. I closed up my laptop now, using mobile. When I go on my laptop tomorrow I will send you a screenshot of what it said.
reply
Sure 😊😊 I'm curious to know...
reply
It sure is trying very hard to be everyone's friend.
reply
Yes, the funny thing is it will happily trash someone’s team but won’t insult them directly.
reply
Maybe it's coded to say nice things and not offend people, but teams are okay to diss. Probably no specification or coding that lets it know some people get very offended if their teams get trashed.
reply
Probably has more to do with the data I gave it. I said nice things about every stacker I taught it about. We will see what it says about you tomorrow.
reply
I thought you let it scan through all past Stacker_sports posts plus comments, and train it that way? If not, what was the method of training? Just some manually entered comments?
reply
I was going to do that initially but selecting all the urls to feed it was time consuming and wasn’t working at first so I just created a 2 page pdf with some info about Stacker News, Stacker Sports, the kind of posts and contests we have in the territory and then I did a small blurb for each stacker that routinely participates in the territory. Just taught it who their favourite teams were, and the sports each stacker likes and the kind of content they post.
reply
50 sats \ 1 reply \ @gnilma 22 Jun
Sounds like you did the bulk of the work to curate and feed it relevant info. But I can see AI being very useful for sifting through large amounts of data and finding/curating relevant results.
reply
Yes, I would like to train it on a bunch of posts and comments but I only started playing with it a couple days ago and I had an issue the first day the open agents dev had to fix for me. Maybe next week I will give it some posts and see how it changes its answer to various questions.
Cool! You might have been the unconscious inspiration behind this idea
reply
I like this idea. I also think it would be great if SN had a breaking news bot. Not full news stories but just headlines or notable news that occurs throughout the day.
reply
What is this and how it works?
reply
It is an AI agent I created using open agents. Still working on it.
reply
Ok now got it.
reply
AI tells us what we want to hear :D .
Everything going on with AI lately reminds me of the movie Her a little bit. I have fun with Grok though :)
reply
Especially if you are the one that trained it.
It is based on very limited data regarding SN right now. Just two pages I wrote up. But eventually hopefully I will train it on a lot of posts and comments and see how its responses change.
reply