Ask it about Tiananmen Square.

Quote from: LifePo7 on January 27, 2025, 21:17:49This is far inferior to chatGPT outside the scope of these weak tests that they train the AI to excel at

Nah. It's better. Claude 3.5 is still better than deepseek in some cases, but deepseek has impressed me quite a bit.

I for one welcome our new chinese overlords. Well, don't really welcome them but it's not like we have much of a choice. I'd rather work under them than under the incestuous, corrupt cliques from California.

And it is cheaper because Chinese KGB will read my prompts and spy on my PC?

Quote from: Jonathan Swift on January 27, 2025, 22:15:40

Quote from: LifePo7 on January 27, 2025, 21:17:49This is far inferior to chatGPT outside the scope of these weak tests that they train the AI to excel at

There's a big difference to factor in: this company is achieving some relatively good-enough genAI on a fraction of the resources and budgets.

Western companies are sinking billions of dollars/euros and using insane amounts of power into something that is overhyped and that has yet to turn a profit (it's all promises and projections) - hence the stock market panic, even if DeepSeek is inferior to ChatGPT.

I agree. I've always thought the AI craze is just another astroturfed fad being forced on everyone by a man (Sam Altman) seeking relevance and fortune by stealing other's copyrighted material without compensation. What a scumbag.

Quote from: LifePo7 on January 27, 2025, 21:17:49This is far inferior to chatGPT outside the scope of these weak tests that they train the AI to excel at

There's a big difference to factor in: this company is achieving some relatively good-enough genAI on a fraction of the resources and budgets.

Western companies are sinking billions of dollars/euros and using insane amounts of power into something that is overhyped and that has yet to turn a profit (it's all promises and projections) - hence the stock market panic, even if DeepSeek is inferior to ChatGPT.

This is far inferior to chatGPT outside the scope of these weak tests that they train the AI to excel at