5 ms·

Public web unravels in AI-driven storm

16 points by benjamg 3y ago

13years 3y ago
Bedlam incoming from the unforeseen consequences of an unpredictable disruptor with broad capability. The "AI clone wars" arriving soon. "Anything you create can be created without investment cost while also being unique in design as well as delivering the same function or experience" From - https://dakara.substack.com/p/ai-and-the-end-to-all-things https://dakara.substack.com/p/ai-and-the-end-to-all-things
- bob-09 3y ago
  I stopped seriously reading at their personal interpretation of the "Twitter Files." Some individuals serving in the public sector might want to "control your thoughts" with this new scary technology, but I think that's FUD and you're much more likely to encounter the private sector trying to control your thoughts. The author mentions how large corporations will be very slow to adopt because of "endless processes and approvals," but fails to mention government bureaucracy. Yet "government" is mentioned in the story twice as much.
  - 13years 3y ago
    Some individuals serving in the public sector might want to "control your thoughts" with this new scary technology, but I think that's FUD and you're much more likely to encounter the private sector trying to control your thoughts. They are not mutually exclusive. However, there is significant precedent for government interest in controlling population narratives. We have agencies that exist for that purpose alone. Additionally, Twitter Files just recently got another personal interpretation from the court which has now just blocked the government from such previous actions.
    - bob-09 3y ago
      I understand that, and the public narratives being spun by the private sector about “Twitter Files” in an attempt to prevent government intervention in future cases where coordinated disinformation is used to attack governments and control its citizens. Scary government is disproportionately represented in the story you linked, that’s all I’m observing here.
JimtheCoder 3y ago
Maybe having something like confirmed identities and actually retaining some ownership over what you contribute online might be a good idea...
- ImPostingOnHN 3y ago
  the article is referring to read access
binarymax 3y ago
A conclusion made in the article - "these sites are trying to shut their technical gates so others can't gobble up troves of data for AI models to study." Isn't supported by anything. Maybe it's true? But it may just as well be speculation from the author. Is there any evidence to this?
- hotpotamus 3y ago
  > And so, by restricting usage, Elon Musk is hoping to prevent its data from being pulled into LLMs. He wants to negotiate with AI companies and get paid for the content that they take. https://www.bbc.com/news/technology-66093324#:~:text=On%20Saturday%20Twitter%20began%20to,verified%20users%2010%2C000%20per%20day https://www.bbc.com/news/technology-66093324#:~:text=On%20Sa....
- Terr_ 3y ago
  I've never seen any good evidence for it, it's just what operators of the sites (i.e. Reddit, Twitter) claim as a justification. IMO "following the money" indicates their disruptive wholesale access-restrictions are driven by much blander profit-seeking reasons.
  - hactually 3y ago
    I mean, it's known that GPT3 was trained on Reddit data with 3 or more upvotes. The training data is still available to download.
    - Terr_ 3y ago
      Perhaps, but what Reddit chose to do is not consistent with its stated reasons. A simple change to the terms-of-service and/or some rate-limits would deter those large-scale (and sue-able) AI companies, until Reddit could introduce some sort of "machine-learning tier" of API access and charges. Instead they announced punishing fees on everyone and everything with barely any warning. That looks a lot more like a general push for blanket monetization. Their "nasty AI learning made us do it" statement looks more like a convenient excuse to deflect user-anger while simultaneously getting their company name in the news right next to hot buzzwords.
ProllyInfamous 3y ago
Wait until Us Masses begin realizing that these companies are using both our public-facing AS WELL AS PRIVATE MESSAGES within their dataset trainings.
- nomel 3y ago
  I'm sure you're right, but reference?
  - ProllyInfamous 3y ago
    I do not ever wish to be sued for libel, again; so instead, I will encourage you to learn about GPT2's specific training dataset, primarily upon one specific social media's contents (public and/or private. I'll let you decide).
    - nomel 3y ago
      Ah, it's fairly recent: https://www.businessinsider.com/openai-chatgpt-generative-ai-stole-personal-data-lawsuit-children-medical-2023-6 https://www.businessinsider.com/openai-chatgpt-generative-ai...
      - ProllyInfamous 3y ago
        Thank you for this citation, as it led me to the recent 160+ page PDF/lawsuit.

16 points by benjamg 3y ago

Bedlam incoming from the unforeseen consequences of an unpredictable disruptor with broad capability. The "AI clone wars" arriving soon. "Anything you create can be created without investment cost while also being unique in design as well as delivering the same function or experience" From - https://dakara.substack.com/p/ai-and-the-end-to-all-things https://dakara.substack.com/p/ai-and-the-end-to-all-things

bob-09 3y ago
I stopped seriously reading at their personal interpretation of the "Twitter Files." Some individuals serving in the public sector might want to "control your thoughts" with this new scary technology, but I think that's FUD and you're much more likely to encounter the private sector trying to control your thoughts. The author mentions how large corporations will be very slow to adopt because of "endless processes and approvals," but fails to mention government bureaucracy. Yet "government" is mentioned in the story twice as much.
- 13years 3y ago
  Some individuals serving in the public sector might want to "control your thoughts" with this new scary technology, but I think that's FUD and you're much more likely to encounter the private sector trying to control your thoughts. They are not mutually exclusive. However, there is significant precedent for government interest in controlling population narratives. We have agencies that exist for that purpose alone. Additionally, Twitter Files just recently got another personal interpretation from the court which has now just blocked the government from such previous actions.
  - bob-09 3y ago
    I understand that, and the public narratives being spun by the private sector about “Twitter Files” in an attempt to prevent government intervention in future cases where coordinated disinformation is used to attack governments and control its citizens. Scary government is disproportionately represented in the story you linked, that’s all I’m observing here.

JimtheCoder 3y ago

Maybe having something like confirmed identities and actually retaining some ownership over what you contribute online might be a good idea...

ImPostingOnHN 3y ago
the article is referring to read access

binarymax 3y ago

A conclusion made in the article - "these sites are trying to shut their technical gates so others can't gobble up troves of data for AI models to study." Isn't supported by anything. Maybe it's true? But it may just as well be speculation from the author. Is there any evidence to this?

hotpotamus 3y ago
> And so, by restricting usage, Elon Musk is hoping to prevent its data from being pulled into LLMs. He wants to negotiate with AI companies and get paid for the content that they take. https://www.bbc.com/news/technology-66093324#:~:text=On%20Saturday%20Twitter%20began%20to,verified%20users%2010%2C000%20per%20day https://www.bbc.com/news/technology-66093324#:~:text=On%20Sa....
Terr_ 3y ago
I've never seen any good evidence for it, it's just what operators of the sites (i.e. Reddit, Twitter) claim as a justification. IMO "following the money" indicates their disruptive wholesale access-restrictions are driven by much blander profit-seeking reasons.
- hactually 3y ago
  I mean, it's known that GPT3 was trained on Reddit data with 3 or more upvotes. The training data is still available to download.
  - Terr_ 3y ago
    Perhaps, but what Reddit chose to do is not consistent with its stated reasons. A simple change to the terms-of-service and/or some rate-limits would deter those large-scale (and sue-able) AI companies, until Reddit could introduce some sort of "machine-learning tier" of API access and charges. Instead they announced punishing fees on everyone and everything with barely any warning. That looks a lot more like a general push for blanket monetization. Their "nasty AI learning made us do it" statement looks more like a convenient excuse to deflect user-anger while simultaneously getting their company name in the news right next to hot buzzwords.

ProllyInfamous 3y ago

Wait until Us Masses begin realizing that these companies are using both our public-facing AS WELL AS PRIVATE MESSAGES within their dataset trainings.

nomel 3y ago
I'm sure you're right, but reference?
- ProllyInfamous 3y ago
  I do not ever wish to be sued for libel, again; so instead, I will encourage you to learn about GPT2's specific training dataset, primarily upon one specific social media's contents (public and/or private. I'll let you decide).
  - nomel 3y ago
    Ah, it's fairly recent: https://www.businessinsider.com/openai-chatgpt-generative-ai-stole-personal-data-lawsuit-children-medical-2023-6 https://www.businessinsider.com/openai-chatgpt-generative-ai...
    - ProllyInfamous 3y ago
      Thank you for this citation, as it led me to the recent 160+ page PDF/lawsuit.