And I’m also glad the site seems to work for you! Just received a report from @kevinpowell.co that he can’t securely connect. So I’m now investigating SSL certificate issues over here …😅

💚😁

And so we’re left talking to brains we can’t control that don’t know right or wrong and have an antisocial personality disorder (because they don’t have a personality or real conscience) – wrapped into a charming voice our primate brains mistake for a friend. Wild.

So even if OpenAI actually wanted to make this technology secure, they simply can’t. That’s why Geoffrey Hinton and other researchers at Google, for example, felt like the technology wasn’t ready for the public yet. But Sam and friends released it anyway.

And because even the people working on the LLMs only understand how the algorithms work that train the neural networks but not how answers and decisions form in those “brains”, they have no way to intervene at specific points of the decision making process. It’s a black box.

I found this interview with Eliezer Yudkowsky really insightful. Because he clearly explains that OpenAI et al. don’t have a way to program certain behavior into those models except for training them with examples of good behavior and hoping the model will adopt some of it.

youtu.be/2Nn0-kAE5c0?...

FOMO2.

Hi Elwin! Thanks for the heads up! It seems to be working on my side, but I’ll have a look at it … thanks a lot! 🤗

Zis woult nevver hav passt German quolity assurenz!

😂

Actually …… it’s more like trying to load a dishwasher while being remote-controlled via a VR headset by a guy who had three bottles of wine 😂

@mkbhd.com already grilled them (very politely, as always)
youtu.be/j31dmodZ-5c?...

Hi Nils! That looks really interesting – I’ll read it over the weekend! And I actually once visited the World Usability Congress in Graz as well (before the pandemic, I ran a workshop there). That was a really great conference! 😄

Oh, enjoy your time in Amsterdam guys! 🥳