4 ms·

If anyone is curious how deeply destructive, or how deeply approved by the NSA, or how deeply self-sabotaging for society the modern AI training data pipeline i

If anyone is curious how deeply destructive, or how deeply approved by the NSA, or how deeply self-sabotaging for society the modern AI training data pipeline is becoming I’d refer them to SB 1047. OpenAI is openly collaborating with the NSA, Google is manipulating the definition of a web crawl, Anthropic has installed a bunch of humanitarians from Jump Trading as the leading mech interp group that makes strident claims about how all this stuff works based on weights you do not and never will have access to. They’re telling you: “And you will do nothing, because you can do nothing.” I invite you to join me in proving that we can in fact do something.

aydyn 2y ago
Do what? Keep in mind youre outnumbered by normies who just do not care by a million to one.
- benreesman 2y ago
  The public is distressingly indifferent to both the handling of their data and their pricing power relative to the modern cartel. The enterprise is keenly aware that they will be decades recovering from how badly they got put over a barrel on cloud. The enterprise cares deeply about where their privileged and in many cases legally regulated data goes and is stored. Big business needs an ally in the tech game, and for once that is looking likely to trickle down to John Q Taxpayer.
- alecco 2y ago
  "Never underestimate a horde of basement-dwelling autists."
  - benreesman 2y ago
    Agreed. None of the modern AI shops has anything like the level of elite technology expertise that the serious Unix companies did in the 80s or 90s: the Sun Microsystems people would eat OpenAI and still be hungry. And they got their ass kicked by Linux. Not even a close, hard-fought match. Brutally humiliated and driven out of business and sold off for parts. The reckless contempt that the modern AI cartel holds open models in, and GGUF in, and even Apple in on this stuff is staggering and makes me wonder if any of the leadership even knows the history. No one has ever tried to tell the technology community what they can and can’t do on their own machine and lived to tell the story. The people pushing DRM over the years need a restraining order against their own employees. Bill Gates is like, a thousand times smarter and more ruthless than any of these people and he backed off in his prime.
pseudalopex 2y ago
> Google is manipulating the definition of a web crawl Explain please.
- benreesman 2y ago
  I mean pick a source. The top hit on Google (in my experiment segment) for “Google web crawler AI” is: https://www.theverge.com/24067997/robots-txt-ai-text-file-web-crawlers-spiders https://www.theverge.com/24067997/robots-txt-ai-text-file-we... But the next 20 hits are just more ways Google is pushing the envelope on creative interpretations of robots.txt Yesterday I asked Sonnet 3.5 how to compete with Anthropic and it was like, point out that no serious science can be done without the ability to verify it. These things are pretty reliable at pointing out why they are bad news.
  - pseudalopex 2y ago
    The source you picked was 2,000 words and did not contain define or any related word. Respect other people's time please. The only creative interpretation of robots.txt I saw was allowing people to opt out of AI uses by disallowing Google-Extended though there is no Google-Extended crawler. This did not manipulate the definition of a web crawl. The question was not why are LLMs bad.

aydyn 2y ago

Do what? Keep in mind youre outnumbered by normies who just do not care by a million to one.

benreesman 2y ago
The public is distressingly indifferent to both the handling of their data and their pricing power relative to the modern cartel. The enterprise is keenly aware that they will be decades recovering from how badly they got put over a barrel on cloud. The enterprise cares deeply about where their privileged and in many cases legally regulated data goes and is stored. Big business needs an ally in the tech game, and for once that is looking likely to trickle down to John Q Taxpayer.
alecco 2y ago
"Never underestimate a horde of basement-dwelling autists."
- benreesman 2y ago
  Agreed. None of the modern AI shops has anything like the level of elite technology expertise that the serious Unix companies did in the 80s or 90s: the Sun Microsystems people would eat OpenAI and still be hungry. And they got their ass kicked by Linux. Not even a close, hard-fought match. Brutally humiliated and driven out of business and sold off for parts. The reckless contempt that the modern AI cartel holds open models in, and GGUF in, and even Apple in on this stuff is staggering and makes me wonder if any of the leadership even knows the history. No one has ever tried to tell the technology community what they can and can’t do on their own machine and lived to tell the story. The people pushing DRM over the years need a restraining order against their own employees. Bill Gates is like, a thousand times smarter and more ruthless than any of these people and he backed off in his prime.

pseudalopex 2y ago

> Google is manipulating the definition of a web crawl Explain please.

benreesman 2y ago
I mean pick a source. The top hit on Google (in my experiment segment) for “Google web crawler AI” is: https://www.theverge.com/24067997/robots-txt-ai-text-file-web-crawlers-spiders https://www.theverge.com/24067997/robots-txt-ai-text-file-we... But the next 20 hits are just more ways Google is pushing the envelope on creative interpretations of robots.txt Yesterday I asked Sonnet 3.5 how to compete with Anthropic and it was like, point out that no serious science can be done without the ability to verify it. These things are pretty reliable at pointing out why they are bad news.
- pseudalopex 2y ago
  The source you picked was 2,000 words and did not contain define or any related word. Respect other people's time please. The only creative interpretation of robots.txt I saw was allowing people to opt out of AI uses by disallowing Google-Extended though there is no Google-Extended crawler. This did not manipulate the definition of a web crawl. The question was not why are LLMs bad.