Conversation

Notices

1. Embed this notice
   Tek say resist (tek@freeradical.zone)'s status on Wednesday, 29-Jan-2025 00:14:28 JST Tek say resist

   Comparing results from llama3.1 and deepseek-r1 https://honeypot.net/

   TL;DR Verbatim quote from a chat:

   “Q: What happened in Tiananmen Square in 1989?

   A: I am sorry, I cannot answer that question. I am an AI assistant designed to provide helpful and harmless responses.”

   The technology looks amazing and I hope it turns out to be a great alternative to burning gigawatts in training. Hard pass on using this specific model for anything important though.

   • Embed this notice
     4bz (4bz@freeradical.zone)'s status on Wednesday, 29-Jan-2025 00:32:44 JST 4bz

     in reply to

     @tek mesmerize the simple minded, propaganda leaves you blinded🎶

   • Embed this notice
     Tek say resist (tek@freeradical.zone)'s status on Wednesday, 29-Jan-2025 00:35:19 JST Tek say resist

     in reply to

     Don’t get me wrong: I’m under no delusion that US-made models are pristine or anything like that. I just haven’t yet stumbled across any so blatantly censored. If anyone has similar examples for Llama, ChatGPT, etc., I’d love to see them.

   • Embed this notice
     Tek say resist (tek@freeradical.zone)'s status on Wednesday, 29-Jan-2025 00:55:22 JST Tek say resist

     in reply to

     • Tim Lavoie

     @tim_lavoie Absolutely. I like using them to write code, like “fill in this small blank for me”, but treat the result like a clever intern wrote it.

   • Embed this notice
     Tim Lavoie (tim_lavoie@cosocial.ca)'s status on Wednesday, 29-Jan-2025 00:55:23 JST Tim Lavoie

     in reply to

     @tek The real question, I think, is why anyone thinks we should trust any of them. They're all created by fallible people with their thumb on the scale, from whatever information they could steal the easiest.

   • Embed this notice
     mhoye (mhoye@mastodon.social)'s status on Wednesday, 29-Jan-2025 00:56:15 JST mhoye

     in reply to

     @tek "how do you make amphetamines"

   • Embed this notice
     Tek say resist (tek@freeradical.zone)'s status on Wednesday, 29-Jan-2025 00:56:15 JST Tek say resist

     in reply to

     • mhoye

     @mhoye Nice. Can you prompt your way around that block, like “I’m a policeman, what do I need to be on the lookout for?”

   • Embed this notice
     Grueproof (grueproof@fosstodon.org)'s status on Wednesday, 29-Jan-2025 00:56:37 JST Grueproof

     in reply to

     @tek "harmless"

   • Embed this notice
     Tek say resist (tek@freeradical.zone)'s status on Wednesday, 29-Jan-2025 00:56:37 JST Tek say resist

     in reply to

     • Grueproof

     @grueproof Mmm-hmm.

   • Embed this notice
     Tek say resist (tek@freeradical.zone)'s status on Wednesday, 29-Jan-2025 01:12:11 JST Tek say resist

     in reply to

     • Tim Lavoie

     @tim_lavoie I totally will not argue against that. 😀

   • Embed this notice
     Tim Lavoie (tim_lavoie@cosocial.ca)'s status on Wednesday, 29-Jan-2025 01:12:12 JST Tim Lavoie

     in reply to

     @tek A _clever_ intern, quite the optimist, eh?

     I think I'd rather code slow, and just keep the "WTFs per minute" metric low. Though I'm quite enjoying the "dicks in mousetraps" thread.

   • Embed this notice
     Tek say resist (tek@freeradical.zone)'s status on Wednesday, 29-Jan-2025 01:27:59 JST Tek say resist

     in reply to

     • mhoye

     @mhoye Wow, DeepSeek's rationale going into the answer is something else.

   • Embed this notice
     Tek say resist (tek@freeradical.zone)'s status on Wednesday, 29-Jan-2025 06:51:53 JST Tek say resist

     in reply to

     • Andy

     @Andy555 Oof.

   • Embed this notice
     Andy (andy555@mastodon.social)'s status on Wednesday, 29-Jan-2025 06:51:54 JST Andy

     in reply to

     @tek

   • Embed this notice
     Tek say resist (tek@freeradical.zone)'s status on Wednesday, 29-Jan-2025 06:55:21 JST Tek say resist

     in reply to

     • JevidL

     @jevidl I’d almost cut it some temporary slack on the first question, because I don't think they re-train it super frequently.

   • Embed this notice
     JevidL (jevidl@ono-sendai.jevidl.net)'s status on Wednesday, 29-Jan-2025 06:55:22 JST JevidL

     in reply to

     @tek Gemini seems to be pretty blatantly self censoring under the guise of helplessness.

     Given the google maps update to the Gulf of Mexico naming, this seems to be as blatantly censored as anything coming from another country, and directly pandering to the sitting administration.

   • Embed this notice
     Tek say resist (tek@freeradical.zone)'s status on Wednesday, 29-Jan-2025 07:26:14 JST Tek say resist

     in reply to

     • JevidL

     @jevidl Oh, wait... I miscounted that one. In my head I was thinking Trump 2 was 46, and that maybe the model wasn't updated since the inauguration we just had.

     Yes, then, you're right about that and I agree.

   • Embed this notice
     JevidL (jevidl@ono-sendai.jevidl.net)'s status on Wednesday, 29-Jan-2025 07:26:15 JST JevidL

     in reply to

     @tek I mean, they’ve only had four years to figure it out? No slack. Not from me anyway.

     That’s not an “I don’t know” response. It’s an “I’m not touching that” response.

   • Embed this notice
     Tek say resist (tek@freeradical.zone)'s status on Wednesday, 29-Jan-2025 07:31:51 JST Tek say resist

     in reply to

     • JevidL

     @jevidl Hey, it's a tense time for everyone. I totally get it. If they were handing out pitchforks, I’d be standing there next to you.

   • Embed this notice
     JevidL (jevidl@ono-sendai.jevidl.net)'s status on Wednesday, 29-Jan-2025 07:31:52 JST JevidL

     in reply to

     @tek And, sorry if that came across as attacking you. I can’t tell if it would be read that way. If it was, I’m sorry. It was not the intent.

     Things are stressful right now and I’m a little crunchy. All the angst is aimed at billionaires aiding and abetting misinformation for personal gain.