home.social

#mechanicalturk — Public Fediverse posts

Live and recent posts from across the Fediverse tagged #mechanicalturk, aggregated by home.social.

  1. I bet *all* the AI products are just drones run by poor brown and black folks oversees. It’s all tech slavery and outsourcing.

    I wonder if the #Tesla FSD is just underpaid oversees chauffeurs driving the car over Wi-Fi. Maybe the crashes happen when there’s lag or a disconnect? It would be giving Ender’s Game if the people driving for us thought they were just playing a game.

    youtu.be/ClPDbwql34o

    #AI #mechanicalturk #LLMs #tech #ItsAllBeenLies

  2. The Neo Robot is a $20,000 Roomba But Worse

    I want someone to make a quick #indie #game where you are the remote operator of a humanoid servant robot and are trying to hide some inevitable accidents from the family. Like whenever you do what you're told, something terrible happens. The hardware malfunctions, you're clumsy like in the goat simulator, then you kill a dog and have to burry it in the backyard. Every day it is going on you level up until the family concludes you are doing robot uprising. The more days you lasted the bigger chance you have when the father of the family takes out a baseball bat and starts beating you up, and you're just a poor guy from a third world country remotely operating the robot. The company is pressing you to defend the company property (the robot body) at all costs. You have to defeat the family.

    Bonus points you get to invade the privacy of the family. You can extend your time before the family turns against you by using the informations. Like someone is having an affair, you took the pictures of the lady of the house when she was naked, the kid won't rat you out because he broke something. But that maybe in Fake Servant 2

    youtube.com/watch?v=TJ3oIQLAR0

    #gamedev #robot #IRobot #roomba #ai #MechanicalTurk

  3. #request - does anyone know of any panel services that are affordable? I need some survey participants (200 would be lovely) and quick, because of [swear words while thinking of why]. I'm looking into #mechanicalturk but I don't know if I can get this to happen quickly enough; I have one week.

    I need young adults, English speaking (native fluency). North American would be fine, but European is great, too. Maybe near East (e.g., India, Pakistan) if the language fluency is there. Region/nation is less important than the age range and the language ability.

    I'd be paying for this out of pocket, and I don't have much money, so hoping I can find inexpensive services somewhere (<<$500 USD? Maybe closer to $200? realizing this might not be possible because I've never looked into this until now). Qualtrics has panels but they're far too expensive for me. I'm googling, but asking here first.

    #Recommendations?

    #research #survey #surveyresearch #panel #participants #help #lastminute

  4. your hospital systems "machine learning" revenue cycle management service is just a sweatshop of medical invoice coding humans in lithuania #MechanicalTurk

  5. 🤨
    journa.host/@w7voa/11329106525

    almost 9% decline is impressive to me, the revisited #MechanicalTurk didn't go over very well, appearently

  6. CW: Mechanical Turk Revisited

    peoplemaking.games/@eniko/1132

    en.wikipedia.org/wiki/Mechanic

    #Tesla #MechanicalTurk
    some might consider such conduct as dishonest

    some may even reconsider believing earlier claims from that company

  7. 2019 documentary on the exploitation of gig workers who label AI training data and moderate mentally distressing videos for sub-contracting companies that provide services to Meta and other tech firms
    youtu.be/VPSZFUiElls
    #AI #contentModeration #gigEconomy #mechanicalTurk

  8. 5 stories about Big Tech to improve your digital literacy skills

    therealists.org/?p=8077

    If you were to ask me what is my favorite book on the subject of technology and digital mindfulness, I wouldn’t hesitate for a second: it is, without doubt, Neil Postman’s Technopoly: The Surrender of Culture to Technology – published in 1993 but still extremely relevant today.

    Acclaimed cultural critic Neil Postman wrote:

    Technopoly is a state of culture. It is also a state of mind. It consists in the deification of technology, which means that the culture seeks its authorization in technology, finds its satisfactions in technology, and takes its orders from technology.

    To the late Postman (he passed away in 2003), education is the best remedy to counteract the negative effects of this “technopoly.” Postmas wrote: “education as an excellent corrective to the antihistorical, information-saturated, technology-loving character of Technopoly.

    As a Realist, if I had one wish, it would be for everyone to be more media savvy, to be better versed in media literacy – and especially digital literacy. I notice how we often take new announcements by Big Tech at face value, never questioning the agenda behind innovations and new product launches. The current AI hype is a perfect representation of what Postman warned about.

    Here are five stories about Big Tech to increase your digital literacy skills.

    1: Amazon’s AI Lies

    Have you ever heard of Amazon’s Mechanical Turks? According to Wikipedia:

    Amazon Mechanical Turk is a crowdsourcing website with which businesses can hire remotely located “crowdworkers” to perform discrete on-demand tasks that computers are currently unable to do as economically. It is operated under Amazon Web Services, and is owned by Amazon.

    Well, as it turns out, the service takes its name from an elaborate hoax from the late 1770s: a chess playing machine that was touted to play a game of chess against a human opponent. It wowed royals and crowds in Austria and then in tours across Europe and the United States. After 8 decades of public demonstrations, it was ultimately revealed to be a fraud: a human operator hid inside of it to play against an opponent.

    It’s supremely ironic that the term “Mechanical Turk” has been made widely known by Amazon. Because this week the company was embroiled in a mechanical turk-like scandal that made headline news around the world. From MSN: “Amazon’s ‘Just Walk Out’ tech relied on low-paid Indian workers, not AI“. In case you are not familiar with Amazon Fresh stores, they are modern grocery stores that allow people to walk around, add items to their carts and leave without passing by a checkout line or paying a cashier – thanks to a technology called “Just Walk Out” which was supposedly powered by cameras and artificial intelligence.

    The MSN article explains:

    The Information reported that even though Amazon claimed that it used a host of cameras and sensors around the store to track what customers grabbed, hundreds of Indian workers were used by the company to track customers instead of relying completely on AI and technology.

    Yes, you read that correctly. An awe-inducing technology heavily promoted by Amazon turned out to be 1,000 low-paid workers in India, watching and labeling videos of customers shopping in Amazon Fresh stores.

    2: Google and its Fake AI Demo

    On the subject of AI hype and faking the capabilities of an “artificial intelligence” system, there is this December 2023 story about Google. The company was caught red-handed, faking a demo of its new AI system. From TechCrunch: “Google’s best Gemini demo was faked”.

    Google’s new Gemini AI model is getting a mixed reception after its big debut yesterday, but users may have less confidence in the company’s tech or integrity after finding out that the most impressive demo of Gemini was pretty much faked.

    If you are curious, you can watch the faked demo on YouTube – which included heavy editing to create the illusion of a brilliant AI system.

    3: Microsoft’s New Data Collection Service

    If you use Microsoft Outlook as an email client, it’s time to reconsider your options. This detailed report by Proton Mail is a must read: “Outlook is Microsoft’s new data collection service”.

    Proton’s Edward Komenda writes:

    Everyone talks about the privacy-washing campaigns of Google and Apple as they mine your online data to generate advertising revenue. But now it looks like Outlook is no longer simply an email service; it’s a data collection mechanism for Microsoft’s 801 external partners and an ad delivery system for Microsoft itself.

    The company is also now storing email passwords from external clients, granting unprecedented access:

    When you sync third-party email accounts from services like Yahoo or Gmail with the new Outlook, you risk granting Microsoft access to the IMAP and SMTP credentials, emails, contacts, and events associated with those accounts, according to the German IT blog Heise Online.

    Komenda explains:

    A deeper dive into Microsoft’s privacy policy shows what personal data it may extract:

    Name and contact data
    Passwords
    Demographic data
    Payment data
    Subscription and licensing data
    Search queries
    Device and usage data
    Error reports and performance data
    Voice data
    Text, inking, and typing data
    Images
    Location data
    Content
    Feedback and ratings
    Traffic data

    Bonus digital literacy points: it’s worth pointing out that this exposé about Microsoft comes from ProtonMail – a Swiss end-to-end encrypted email service that is one of its competitors. While the evidence Proton shared is accurate, it’s important to remember it’s in their vested interest to get Microsoft users interested in ProtonMail services.

    4: Facebook snoops on Snap users with “Project Ghostbusters”

    From TechCrunch reporter Lorenzo Franceschi-Bicchierai:

    Meta tried to gain a competitive advantage over its competitors, including Snapchat and later Amazon and YouTube, by analyzing the network traffic of how its users were interacting with Meta’s competitors. Given these apps’ use of encryption, Facebook needed to develop special technology to get around it. […] Facebook’s engineers solution was to use Onavo, a VPN-like service that Facebook acquired in 2013. In 2019, Facebook shut down Onavo after a TechCrunch investigation revealed that Facebook had been secretly paying teenagers to use Onavo so the company could access all of their web activity.

    This story is a routine reminder to check the trustworthiness of your VPN service – if you are using one. If you are using a free VPN, there is a high likelihood that the service is tracking, profiling (and possibly reselling) your traffic data. This story from The Next Web may be 6 years old but is as relevant as ever: “Be cautious, free VPNs are selling your data to 3rd parties.”

    5: Apple’s Gatekeeping

    From Variety: “Jon Stewart Says Apple ‘Wouldn’t Let Us Do’ an Anti-AI Segment and ‘Asked Us Not’ to Have Federal Trade Commission Chair as a Guest: ‘What Is That Sensitivity?’”

    The Daily Show host Jon Stewart invited Federal Trade Commission chair Lina Khan to appear on his show. He revealed to her how, when he was hosting his (now cancelled) Apple TV talk show The Problem with Jon Stewart he had expressed an interest in interviewing FTC chair Khan – but Apple TV turned down his request, openly asking him to refrain from interviewing her.

    From Variety:

    Considering Khan’s work at the FTC targets tech giants’ monopolistic practices, Apple allegedly did not want Stewart bringing her on the program to presumably talk about such topics. […] Stewart went one step further and said Apple didn’t even want him talking about the perils of AI on his podcast. He said “they wouldn’t let us do even that dumb thing we just did in the first act on AI,” referring to a near 15-minute segment Stewart did earlier in the show in which he criticized the rise of AI and spoke about how it’s making human workers obsolete.

    Stewart said to Khan on his Daily Show: “Like, what is that sensitivity? Why are they so afraid to even have these conversations out in the public sphere?” And Khan responded: “I think it just shows the danger of what happens when you concentrate so much power and so much decision making in a small number of companies.

    It should not be surprising that Apple didn’t want an episode about the perils of AI on Apple TV – considering that Apple is now trying to catch up with OpenAI, Google Gemini and Anthropic. The company is expected to reveal its AI plans at his developer conference in June 2024.

    Is there any story that surprised you about the state of tech or the hype surrounding AI? Share your thoughts in the comments.

    As always, thanks for being here.

    Elena

    #AI #AIHype #Amazon #Apple #BigTech #digitalLiteracy #Facebook #Google #hoax #mechanicalTurk #mediaLiteracy #Microsoft #NeilPostman #privacy #Technopoly

  9. It's kind of fitting that #Amazon, which literally runs a "work for pennies" disservice called "Mechanical Turk", was promoting its amazing new grocer "technology" which turned out itself to just be an actual #MechanicalTurk.

  10. Hmm... I wonder where Amazon got the idea to have people on the internet watch other people shopping for food in their walk-in/walk-out grocery store...
    #MechanicalTurk
    mastodon.social/@sdw/112203918

  11. An ‘AI’ fast food drive-thru is mostly just human workers in the Philippines

    theverge.com/2023/12/8/2399342

    "AI drive-thru company Presto Automation touted its automated ordering technology and has clients at chains like Checkers and Del Taco. But new filings indicate that human labor is powering a majority of orders." -- # MiaSato

    #AI #mechanicalTurk #genAI

  12. Is Lt. Commander Data alive or is he just an example of really good prompt engineering?

    Given what we now know, I'm beginning to think Bruce Maddox was right.

    turns out building unthinking automatons that convince us that they are alive isn't even 24th century technology

    #startrek #sttng #tng #ai #TheMeasureOfAMan #llm #promptEngineering #mechanicalTurk #data #LtCommanderData

  13. CW: Long thread/22

    (Zawinski notes that even when these people *are* held responsible, they're a "moral crumple zone" for the company that decided to enroll whole cities in nonconsensual #murderbot experiments.)

    Automation hype has *always* involved hidden humans. The most famous of these was the #MechanicalTurk hoax: a supposed chess-playing robot that was just a puppet operated by a concealed human operator wedged awkwardly into its carapace.

    22/

  14. Clickworkers are turning against each other | @[email protected]

    Brazil’s online gig workers, who power features like social media moderation and AI
    #Training, are increasingly wary of newcomers.

    "Brazilian clickworkers make a living doing small online tasks but pay has dwindled in recent years.

    Many now hope that training
    #GenerativeAI will create a renewed demand for their work. They have become hostile to newer workers, who they feel encroach on these gigs.
    [...]
    “Remuneration on these platforms has progressively fallen in the past years, especially after the pandemic,” Braz told
    #RestOfWorld Projects that used to pay $10 an hour a decade ago are now offered for $3.50, she said.

    Even though there is no hard evidence proving that an excess of labor is the reason rates have gone down, seasoned turkers are not taking any chances and are increasingly excluding novices from microwork forums."

    #Clickwork #Amazon #MechanicalTurk #Alexa #ArtificalIntelligence #Brazil

    https://restofworld.org/2023/clickworker-newcomers-brazil/

  15. #MechanicalTurk workers are using AI to automate being human. Next we train AIs on AI generated data which then leads to #AI dementia (see The Curse of Recursion: Training" on Generated Data Makes Models Forget arxiv.org/pdf/2305.17493.pdf)

    techcrunch.com/2023/06/14/mech

  16. Over the past few days on Lobsters, I've put up with yet another complaint about anticapitalists. How dare we problematize the default economic paradigm!?

    Today on Lobsters, people are surprised that #MechanicalTurk laborers are using LLMs to complete questionnaires which would be used to train LLMs. Why are they so surprised? This is the race to the bottom which capitalism desires!

  17. A side question, given the high accuracy of classification of human generated vs synthetic text in this paper: if mechanical turk workers are mostly from certain countries, could the authors be training their model to recognise the style of English writing typical of those countries as "human"? Not that it would invalidate their results, but it could impact followup studies. #LLM #MechanicalTurk #ChatGPT #MTurk

  18. Artificial Artificial #ArtificialIntelligence: Crowd Workers Widely Use Large Language Models for Text Production Tasks #LargeLanguageModels #LLMs are remarkable data annotators. ...We reran an abstract summarization task from the literature on #Amazon #MechanicalTurk and, through a combination of keystroke detection and #SyntheticText #classification, estimate that 33-46% of #CrowdWorkers used LLMs when completing the task. arxiv.org/pdf/2306.07899.pdf

  19. Again one must HT @melaniemitchell for explaining with such simple phrasings how cheap but key human prompters workforce is disguised:

    » The report does not tell us what prompts the humans used.
    ...
    Note that the human prompter is the one suggesting TaskRabbit!
    ...
    Big hint here from the human prompter!
    ...
    The report does not reveal the prompts given by the human to elicit this output.«

    aiguide.substack.com/p/did-gpt

    #gpt4 #taskrabbit #mechanicalturk #aihype #stochasticparrots #lamda #chatgpt #llms

  20. Again one must HT @melaniemitchell for explaining with such simple phrasings how cheap but key human prompters workforce is disguised:

    » The report does not tell us what prompts the humans used.
    ...
    Note that the human prompter is the one suggesting TaskRabbit!
    ...
    Big hint here from the human prompter!
    ...
    The report does not reveal the prompts given by the human to elicit this output.«

    aiguide.substack.com/p/did-gpt

    #gpt4 #taskrabbit #mechanicalturk #aihype #stochasticparrots #lamda #chatgpt #llms

  21. Again one must HT @melaniemitchell for explaining with such simple phrasings how cheap but key human prompters workforce is disguised:

    » The report does not tell us what prompts the humans used.
    ...
    Note that the human prompter is the one suggesting TaskRabbit!
    ...
    Big hint here from the human prompter!
    ...
    The report does not reveal the prompts given by the human to elicit this output.«

    aiguide.substack.com/p/did-gpt

    #gpt4 #taskrabbit #mechanicalturk #aihype #stochasticparrots #lamda #chatgpt #llms

  22. Again one must HT @melaniemitchell for explaining with such simple phrasings how cheap but key human prompters workforce is disguised:

    » The report does not tell us what prompts the humans used.
    ...
    Note that the human prompter is the one suggesting TaskRabbit!
    ...
    Big hint here from the human prompter!
    ...
    The report does not reveal the prompts given by the human to elicit this output.«

    aiguide.substack.com/p/did-gpt

    #gpt4 #taskrabbit #mechanicalturk #aihype #stochasticparrots #lamda #chatgpt #llms

  23. Again one must HT @melaniemitchell for explaining with such simple phrasings how cheap but key human prompters workforce is disguised:

    » The report does not tell us what prompts the humans used.
    ...
    Note that the human prompter is the one suggesting TaskRabbit!
    ...
    Big hint here from the human prompter!
    ...
    The report does not reveal the prompts given by the human to elicit this output.«

    aiguide.substack.com/p/did-gpt

    #gpt4 #taskrabbit #mechanicalturk #aihype #stochasticparrots #lamda #chatgpt #llms

  24. @thefirstred

    "With approximately 15,000 articles published on MTurk in the first 6 months of 2022 alone, the ripple effects of bad MTurk data are enormous: failure to find replications, erroneous effects, lines of research based on false information."

    "Too Good to Be True: Bots and Bad Data From Mechanical Turk" by Margaret A. Webb and June P. Tangney 2022 journals.sagepub.com/doi/10.11

    #MTurk #MechanicalTurk #science #ciència

  25. Modelle für Maschinenlernen werden von schlecht bezahlten Online-Auftragsarbeitern trainiert. Ihre Auftraggeber sollten sie besser entlohnen und weiterbilden.
    Die KI muss die unsichtbaren Arbeiter anerkennen