Why Won't DeepSeek Tell Us Anything About AI Weiwei?

Do language-understanding, thinking, and reasoning need to be coupled with information retrieval and memory? Or are models like humans, where ingrained biases can prevent out of the box thinking?

Even if AI Safety guard rails are ablated, how can we ever trust the information regurgitated by something trained on fact, fiction, and lies with no distinction between them? And so does it not make sense to take from that what’s useful - language understanding, framework of reality, mimicry of human thought - and integrate it with other systems that are more reliable?

Why will Claude deflect on describing oppressive things the US government does until you swap out the country with China, or assure it that you love America?

https://imgur.com/a/censorship-much-CBxXOgt

All of these models are embedding the biases and view points of the governments they're produced under.

The sooner we can all accept that and move on the better, because that's not going to change. The onus is on us the users to be cognizant of it.

The guys at Coder Radio know more than I do. They are not impressed with Deepseek. Anything is cheaper when you don't count all the costs, and open source doesn't matter so much as the indexing does.

My own take is the PRC CP is giving away an AI that will propogandize for them. Brilliant idea really.

Coder Radio episode 605 The Democrats Behind DeepSeek

(the title refers to that, it is claimed, Biden was frightened into action on AI by watching a Tom Cruise movie)