Saga 💙💛
@nordicsaga1337.bsky.social
🇩🇰🇬🇱🇺🇦🇪🇺 Slava Ukraini! Experimental archaeologist. History, Scandinavia & calling out nonsense. Fighting fascism, disinfo & bad takes. Facts over feelings. Respect & tolerance goes both ways!
Opinions my own.
Not here to chat. #NAFO
Opinions my own.
Not here to chat. #NAFO
I couldn't agree more.
It needs to be the European way. The other thing is too dangerous in the wrong hands.
It needs to be the European way. The other thing is too dangerous in the wrong hands.
March 22, 2025 at 5:38 PM
I couldn't agree more.
It needs to be the European way. The other thing is too dangerous in the wrong hands.
It needs to be the European way. The other thing is too dangerous in the wrong hands.
That’s the wild part, isn’t it? We’re basically building these models on a foundation of stolen words and shady data, and then acting surprised when they turn out a little… morally flexible. It’s like teaching a parrot to swear and then getting mad when it curses at guests.
March 22, 2025 at 5:36 PM
That’s the wild part, isn’t it? We’re basically building these models on a foundation of stolen words and shady data, and then acting surprised when they turn out a little… morally flexible. It’s like teaching a parrot to swear and then getting mad when it curses at guests.
Absolutely. Just because Elon wants AI to be some all-knowing genius doesn’t make it reality. We’re nowhere near the level of trust or control needed to hand over major decisions to a language model. It’s reckless & honestly, a bit terrifying that some people think it’s a good idea.
March 22, 2025 at 5:33 PM
Absolutely. Just because Elon wants AI to be some all-knowing genius doesn’t make it reality. We’re nowhere near the level of trust or control needed to hand over major decisions to a language model. It’s reckless & honestly, a bit terrifying that some people think it’s a good idea.
Exactly! It’s like trying to filter out every drop of poison from the ocean, impossible when you need that much data. And the fact that even today’s models are already showing these tendencies? Definitely not reassuring… 🙄
March 22, 2025 at 5:30 PM
Exactly! It’s like trying to filter out every drop of poison from the ocean, impossible when you need that much data. And the fact that even today’s models are already showing these tendencies? Definitely not reassuring… 🙄
Yeah, it’s both fascinating and kind of terrifying. AI is basically marinating in all of humanity’s best and worst tendencies, so, of course, it’s gonna learn some shady tricks along the way.
March 22, 2025 at 5:29 PM
Yeah, it’s both fascinating and kind of terrifying. AI is basically marinating in all of humanity’s best and worst tendencies, so, of course, it’s gonna learn some shady tricks along the way.
Uncommon sense. I love that! It’s so true. Once you see it, you can’t unsee it. Kinda like spotting 1337 in a handle and suddenly feeling very nerdy 😄
March 22, 2025 at 5:27 PM
Uncommon sense. I love that! It’s so true. Once you see it, you can’t unsee it. Kinda like spotting 1337 in a handle and suddenly feeling very nerdy 😄
I hadn’t come across this one specifically myself, but thanks for sharing! Bengio’s insights are super important, definitely something we need to keep a close eye on.
March 22, 2025 at 5:22 PM
I hadn’t come across this one specifically myself, but thanks for sharing! Bengio’s insights are super important, definitely something we need to keep a close eye on.
I am not sure if it is this one, it is few years old though.
It is a good read, however I still think it is better if we write our own texts.
www.wlu.ca/academics/re...
It is a good read, however I still think it is better if we write our own texts.
www.wlu.ca/academics/re...
Can a computer outfake a human?
In our recent paper, we examined the potential role of generative AI in assisting job candidates with faking personality assessments used in personnel selection.
www.wlu.ca
March 22, 2025 at 2:42 PM
I am not sure if it is this one, it is few years old though.
It is a good read, however I still think it is better if we write our own texts.
www.wlu.ca/academics/re...
It is a good read, however I still think it is better if we write our own texts.
www.wlu.ca/academics/re...
This one too. But these are not new new. Some months old.
I think I had more somewhere.
time.com/7202784/ai-r...
I think I had more somewhere.
time.com/7202784/ai-r...
Exclusive: New Research Shows AI Strategically Lying
Experiments by Anthropic and Redwood Research show how Anthropic's model, Claude, is capable of strategic deceit
time.com
March 22, 2025 at 2:34 PM
This one too. But these are not new new. Some months old.
I think I had more somewhere.
time.com/7202784/ai-r...
I think I had more somewhere.
time.com/7202784/ai-r...
There's this one too. Don't know if you seen this one.
OpenAI and other frontier AI models try to "scheme" users
Researchers find that as LLMs evolve they could be more likely to deceive users.
www.axios.com
March 22, 2025 at 2:33 PM
There's this one too. Don't know if you seen this one.
I'm checking my bookmarks right now, I know I saved some pages somewhere, but forgot where I put them :S Goldfish brain..
March 22, 2025 at 2:31 PM
I'm checking my bookmarks right now, I know I saved some pages somewhere, but forgot where I put them :S Goldfish brain..
Is it this one?
I find this one interesting. www.pnas.org/doi/10.1073/...
I find this one interesting. www.pnas.org/doi/10.1073/...
The debate over understanding in AI’s large language models | PNAS
We survey a current, heated debate in the artificial intelligence (AI) research community
on whether large pretrained language models can be said t...
www.pnas.org
March 22, 2025 at 2:28 PM
Is it this one?
I find this one interesting. www.pnas.org/doi/10.1073/...
I find this one interesting. www.pnas.org/doi/10.1073/...