DeepSeek

[CptTrips]

Remember the Opium Wars? It's really up to each to defend their own.

I'm so old, I remember when all my toys were stamped "Made in Japan".
Later it was all "Made in Taiwan".
Now everything is "Made in China".

I imagine we are going to re-shore a lot of stuff in the coming years. 

But it's not likely to be humans in those factories.

[LCADolby]

Ask Deepseek about "Tank Man 1989" and see how far you get

[AKIron]

Ask Deepseek about "Tank Man 1989" and see how far you get

Someone reported on that and said it gave an accurate description.

Others report mixed results depending on language used.

[LCADolby]

What about a phrase such as; "China belongs to Taiwan, give it back Winnie the Pooh!" 

[CptTrips]

Ask Deepseek about "Tank Man 1989" and see how far you get

That's why I wouldn't use Deepseek app directly.  I have no doubt their dataset has been carefully sanitized.

Western companies might use their search\reasoning logic code though, with their own uncensored dataset.


LoL.  This vid made a good point I hadn't thought of.



OpenAI has complaints about Deepseek scrapping their model for training. 

That's kinda ironic given that the biggest beef against OpenAI and the other AIs is that they are scraping the web stealing bits and pieces of other people IP to build their model and output.  Re-Presenting it for commercial purposes without original IP owners permission.  Every answer OpenAI gives is composed of little snippets of other people IP scraped off the internet and pasted together.  Peoples novels, articles, blogs, webpages, paintings, photographs, etc.

It's the Wild West. 

[AKIron]

Try Google's AI. No less biased.

[Lazerr]

Have yet to use any AI app, what do they do?  Is it like the new lazy way of google and reading?

[CptTrips]

Have yet to use any AI app, what do they do?  Is it like the new lazy way of google and reading?

Whether or not you use an AI app, in 10 years time AI will be completely interwoven with virtually every aspect of your life whether you want it or not.

It will be in your automobiles,
it will be used by your mechanic to diagnose problems right the first time and give detailed instructions for repair,
it will be used for more efficient air traffic control, and more efficient traffic light control to maximize traffic efficiency,
it will be used by police departments to track various metrics and predict what areas and time are most likely to need additional patrol to maximize use of limited man-power,
it will be used by doctors to more accurately read complex blood test results with hundreds of interrelated reading to match to recognize signs of diseases your doctor didn't even know about, or x-ray reading to find anomalies that would go unnoticed by a human,
it will be used by intelligence services to analyze billions of pieces of data to detect subtle signs of terrorist planning an attack that humans would miss,
etc, etc, etc.

In 10 years time AI will be a fundamental thread running through our entire civilization.

You might as well ask, “what are these dang nab computer things and why should I ever care they exist.” 

It’s going to be everywhere, touching everything.

[guncrasher]

and then we have skynet, just a few years past where it was first mentioned


semp

[Brooke]

DeepSeek is nifty, but it's just the latest in what will be periodic, ongoing leapfrogging of whatever what the previous best AI.

Also, don't believe everything Chinese organizations claim with regard to how many resources it took to do this or that.

As for "they stole our ChatGPT tech!"  My feeling is sort of like Gizmodo's title "OpenAI Claims DeepSeek Plagiarized Its Plagiarism Machine".  OpenAI isn't alleging DeepSeek stole confidential source code.  They are alleging DeepSeek gathered responses from ChatGPT, and used those for training.  But ChatGPT's responses are from training on the gathered writing of approximately the entire human species over all of time.  I suspect it is a terms-of-service violation of the API to use text output for training another model.  But I can't get all the worked up it.  ChatGPT spews out text into the general public.  Trying to police what people do with that spewed out text doesn't seem destined to succeed.  Similar to newspapers claiming people can read their stuff spewed out into the public (and thus people can incorporate it into their neural networks), but LLM's can't.  It'll be a thing fought out in courts on what you can and can't do with copyrighted material.

[AKIron]

.