Latest / Mini Scroll / MINI TECH SCROLL: TikTok US buyer update, Content moderators speak out , X pilots AI fact-checkers + Can AI run retail?

Transcript

MINI TECH SCROLL: TikTok US buyer update, Content moderators speak out , X pilots AI fact-checkers + Can AI run retail? · Mini Scroll

• 0:00
  Hey everyone, and welcome to Mini Scroll, the daily Internet
• 0:02
  culture podcast bringing you the biggest stories from social
• 0:05
  media, the creator economy and the digital space every Monday
• 0:08
  through Thursday and sometimes Fridays.
• 0:11
  My name is Annabelle and I'm your Wednesday Mini Tech Scroll
• 0:13
  host and also the deputy editor of the youth digital media brand
• 0:17
  Centennial World. OK, we first have to address the
• 0:20
  elephant in the room. I'm clearly in a very different
• 0:23
  setup. The audio probably sounds a
• 0:26
  little bit different than usual. I'm sorry, it's probably worse
• 0:28
  quality, but I am actually recording the podcast from home
• 0:33
  today. Lauren mentioned yesterday that
• 0:35
  Sydney is currently going through a weather event, so a
• 0:39
  lot of people are not making unnecessary journeys or
• 0:43
  travelling to work at the moment.
• 0:45
  So we just decided that I would do mini text scroll from home.
• 0:49
  I hope you guys are enjoying the change of scenery.
• 0:52
  I am recording the podcast from my room.
• 0:55
  I feel like I am serving fashion influenza with my background
• 0:58
  right now so hopefully that's a little bit of fun.
• 1:02
  I always joke that I look like an impostor in the mini scroll
• 1:06
  set slash studio, but this feels even weirder.
• 1:09
  So here we are. We're going to push through and
• 1:13
  I hope the quality of this episode isn't too bad.
• 1:17
  Before we get into the stories for this week, we finally hit
• 1:21
  10K on YouTube, which Lauren mentioned on Monday, which is
• 1:25
  really exciting. So thank you to everyone for
• 1:28
  subscribing and helping us get there.
• 1:30
  Otherwise, if you want to support Centennial in other
• 1:33
  ways, Lauren set up a Buy Me a Coffee, so feel free to check
• 1:36
  that out. But yeah, otherwise Please
• 1:39
  remember to rate, review, and subscribe and let's get into the
• 1:42
  stories. OK, so the first story for today
• 1:44
  is a little bit of a shorter one.
• 1:46
  As we all know, I physically cannot find the courage to
• 1:50
  discuss the TikTok ban at length anymore, but I do want to share
• 1:54
  that Trump has claimed that he has found a buyout for TikTok.
• 1:58
  As a really quick refresher, the US government passed down
• 2:01
  legislation to ban TikTok in April of last year.
• 2:04
  And if we boil it down, TikTok was kind of caught in the
• 2:07
  crossfire between the United States and China's political
• 2:10
  tension. Essentially, the US government
• 2:12
  saw the fact that TikTok is Chinese, owned by Byte Dance, as
• 2:16
  a threat to national security. Since then, TikTok was briefly
• 2:19
  banned as we all know, and Trump has since extended the deadline
• 2:23
  for the ban three times, which I have covered.
• 2:26
  To avoid the ban though, the US government is looking for Byte
• 2:29
  Dance to divest and hopefully be sold to AUS based company.
• 2:34
  There's been a lot of talk about potential buyers.
• 2:36
  This included obviously Elon Musk at one point, Oracle, and
• 2:39
  more that I've reported on recently.
• 2:41
  Trump told Fox News that they have a buyer for TikTok, but he
• 2:44
  didn't specify any names as he said, quote, we have a buyer for
• 2:48
  TikTok, by the way, I think I'll probably need China's approval
• 2:51
  and I think President Xi will probably do it.
• 2:54
  It's a group of very wealthy people.
• 2:55
  I mean, it will definitely need Chinese approval.
• 2:58
  We have spoken about this before and by dad is very hesitant to
• 3:02
  sell the algorithm. So who is to say what is
• 3:05
  actually going to happen from this?
• 3:07
  That being said, though, Trump did note that the buyer will
• 3:10
  allegedly be revealed in two weeks.
• 3:13
  So I will keep you updated on that.
• 3:15
  There are a couple of groups that that have reportedly thrown
• 3:18
  their hat in the ring to buy TikTok.
• 3:20
  Perhaps the most notable 1 is the People's Bid for TikTok,
• 3:23
  which is made-up of billionaire Frank McCourt and Kevin O'Leary
• 3:27
  from Shark Tank. It also has backing from the
• 3:30
  Reddit Co founder Alexis Ohanian.
• 3:33
  Aside from them, there was also reports that Mister Beast was
• 3:36
  putting together a collective. So who knows, maybe it's one of
• 3:40
  those two groups, maybe it's neither.
• 3:42
  I guess we will just have to wait and see.
• 3:44
  The next story is also TikTok and that is that TikTok
• 3:48
  moderators from around the world are speaking out.
• 3:50
  This story actually came from a publication called Rest of the
• 3:53
  World, which is a publication that focuses on tech stories
• 3:56
  beyond the global. NI will say it is based in the
• 3:59
  US, but many freelance journalists contribute to the
• 4:03
  publication. The piece specifically that I
• 4:05
  got this story from was written by Kaya Gensch and discusses how
• 4:09
  content moderators are dealing with rather traumatic working
• 4:11
  conditions. While he is based in Turkey,
• 4:15
  this is a case that is applicable across the world.
• 4:18
  Before we get into the struggles that the cohort is specifically
• 4:22
  facing, I thought it would be helpful to kind of give some
• 4:25
  background on TikTok mods and the type of work that they may
• 4:28
  do. So obviously, the
• 4:30
  responsibilities of a content moderator are to ensure that the
• 4:34
  content on the app and the videos posted adhere to the
• 4:37
  community guidelines. So they would essentially note
• 4:40
  which ones do not fit these guidelines, and it's all about
• 4:43
  ensuring that the app remains a safe space.
• 4:47
  According to a 2023 article from The Guardian Quote, TikTok says
• 4:51
  it has more than 40,000 professionals dedicated to
• 4:54
  keeping the platform safe. Moderators work alongside
• 4:57
  automated moderation systems, reviewing content in more than
• 5:00
  70 languages. Obviously, I'm aware that this
• 5:03
  article is from two years ago, so as you would expect, things
• 5:07
  have definitely changed regarding the number of content
• 5:10
  moderators around the world. I saw a lot of articles from
• 5:13
  last year reporting that TikTok laid off hundreds of moderators,
• 5:18
  replacing them with AI. Much of the reporting I saw
• 5:21
  around this was that it was mostly impacting Malaysian
• 5:23
  moderators, with around 500 workers impacted.
• 5:26
  With this context, I think it's worth noting that there are a
• 5:29
  couple of content moderator hubs around the world, with many
• 5:33
  being countries in the Global South.
• 5:36
  This is because it's obviously cheaper for big tech companies,
• 5:39
  but also the workers may be less protected.
• 5:42
  A 2022 piece from the Bureau of Investigative Journalism looked
• 5:46
  into Latin America specifically and detailed how various
• 5:50
  moderators in Colombia were dealing with mental health
• 5:52
  issues after being forced to watch traumatic content for days
• 5:55
  on end. For reference, one moderator
• 5:58
  that the researchers spoke to had to watch 900 to 1000 videos
• 6:02
  per day. As someone who is extremely
• 6:05
  chronically online, I don't think I could manage to watch
• 6:08
  900 videos per day. And that is not to mention that
• 6:11
  a lot of this would be very traumatic content.
• 6:14
  So it's clear from the reporting that I've just touched on that
• 6:16
  the work conditions for content moderators have been an issue
• 6:20
  for some time now. Earlier this year, actually, we
• 6:23
  saw the birth of the Global Trade Union Alliance of content
• 6:28
  moderators, which was the first global trade union alliance for
• 6:31
  this group. It was created with the
• 6:33
  intention to improve working conditions and it was actually
• 6:36
  announced in Kenya so moderators could join the alliance through
• 6:40
  kind of local unions at the time and that it was announced there
• 6:43
  were unions from places like Ghana, Turkey, Colombia,
• 6:46
  Morocco, the Philippines and more.
• 6:48
  So while content moderators are facing different conditions all
• 6:51
  around the world, Gench spoke to various Turkish content
• 6:55
  moderators and they have essentially been outsourced by
• 6:58
  TikTok to moderate content. As he writes quote, the stories
• 7:02
  of Turkish content moderators follow a similar trajectory.
• 7:06
  What started as a high paying, enjoyable job done from home
• 7:09
  turned psychologically grueling as workers felt in increasingly
• 7:12
  surveilled and pressured. The workers say they are now
• 7:15
  overworked and not given sufficient support.
• 7:17
  When they attempted to organize and fight for their rights, they
• 7:20
  say their efforts were shut down.
• 7:22
  I would definitely recommend going to read this piece if
• 7:25
  you're interested in hearing the personal anecdotes from various
• 7:28
  moderators, but I just thought the story was super interesting
• 7:31
  from a couple different perspectives.
• 7:33
  The first reason is kind of a little bit rogue for this
• 7:35
  podcast. I won't lie, but I was really
• 7:37
  interested in this story because I have a background in
• 7:40
  international relations. That was one of my majors at
• 7:43
  university. And to me, this is just another
• 7:45
  instance where we see the global N exploit and extract labor from
• 7:50
  the global SI mean TikTok definitely does complicate this
• 7:54
  a little bit because they are Chinese owned, but they are
• 7:57
  undoubtedly A technological and economic powerhouse.
• 8:00
  And at the end of the day, the content moderation hubs in the
• 8:03
  global S are not designing or making the tech necessarily,
• 8:07
  They're absorbing the human cost of having to maintain it.
• 8:11
  My second reason why I found this interesting was because,
• 8:13
  you know, obviously there's been a lot more discussion around
• 8:16
  content moderation this year on social media platforms,
• 8:20
  specifically since Meta changed its content moderation policies.
• 8:24
  On top of this, there is more and more discussion around how
• 8:28
  AI can contribute to content moderation and fact checking.
• 8:32
  And that actually brings us to our next story, which is a
• 8:35
  little bit of an offshoot, and that is that X is piloting a
• 8:39
  fact checking program led by AI bots.
• 8:41
  I saw this just before I sat down to record so this is a bit
• 8:45
  of a quick one, but I felt it was relevant considering the
• 8:49
  previous story. Anyway, we all know that X
• 8:52
  introduced Community Notes a while ago and it was actually
• 8:55
  before the platform was even X and Elon Musk had taken it over.
• 9:00
  But if you aren't familiar with Community Notes, it's a feature
• 9:03
  that relies on crowd sourced information to Fact Check
• 9:06
  certain claims on the platform and basically it pops up as a
• 9:09
  box underneath a post. People have had a lot to say
• 9:13
  about Community Notes. Some people think it's an
• 9:15
  effective way to Fact Check, but others are not so convinced.
• 9:18
  Given that community notes don't really affect the reach of a
• 9:21
  post. So for example, if an inaccurate
• 9:24
  post has been spread and 2 million people for example have
• 9:27
  seen it before the community note pops up, people are
• 9:30
  skeptical of how much the community note actually helps.
• 9:34
  So how community notes actually work though, is that any ex user
• 9:37
  can sign up to write a note, except there are a couple of
• 9:41
  restrictions. So for example, their account
• 9:44
  has to be at least six months old.
• 9:45
  They have to have a verified phone number and no recent
• 9:49
  violations of ex community guidelines and notes become
• 9:53
  public when there is a certain level of consensus.
• 9:57
  So this new program was announced earlier today.
• 10:00
  The announcement, which was made from the at community notes
• 10:03
  account on X, says as follows. Quote Starting today, the world
• 10:07
  can create AI note writers that can earn the ability to propose
• 10:11
  community notes. Their notes will show on X if
• 10:13
  found helpful by people from different perspectives, just
• 10:16
  like all notes. Here's how it works.
• 10:19
  Sign up today and begin developing your AI note writer.
• 10:22
  Start writing notes in test mode.
• 10:24
  We'll admit a first cohort of AI note writers later this month,
• 10:28
  which is when AI written notes can start appearing.
• 10:30
  So this is definitely an interesting development when it
• 10:32
  comes to how AI is going to be used in fact checking and of
• 10:36
  course, content moderation. I'm not exactly sure how
• 10:39
  successful it will be. I'm sure there will be some
• 10:41
  bumps along the way, but we have all seen how some generative AI
• 10:46
  and large language models make up context or information.
• 10:50
  I mean, recently we covered how ChatGPT, for example, had become
• 10:53
  too sycophantic, which has led to a lot of conversation around
• 10:58
  ChatGPT induced psychosis. But yeah, I do think there are
• 11:01
  definitely pros and cons to this kind of fact checking.
• 11:04
  It could definitely help lessen the load for human fact checkers
• 11:07
  and content moderators. And as you could tell by the
• 11:10
  last story, this is something that is probably necessary.
• 11:13
  However, I will say I do think there is a chance that it may
• 11:16
  decrease trust. I do think people have kind of a
• 11:19
  really weird perception of generative AI As for how
• 11:23
  trustworthy it is. Like we all know that this type
• 11:26
  of AI isn't perfect yet. So some of us are very wary of
• 11:30
  the information that it spits out, while others I feel like
• 11:33
  take it as gospel. So at the end of the day though,
• 11:35
  while there are some quality and consistency issues that we are
• 11:39
  up against when it comes to this kind of fact checking, I do
• 11:43
  think there is definitely value to AI and human fact checkers
• 11:47
  working together. I don't think 1 can replace the
• 11:49
  other entirely. For my last story, I'm not sure
• 11:52
  if you guys have noticed, but I always try and include one more
• 11:56
  fun or quote UN quote unserious story each week because I feel
• 12:00
  like sometimes this podcast specifically can get a little
• 12:03
  bit bogged down in the politics and regulations of it all, which
• 12:07
  definitely has a time and place. But yes, this is that story.
• 12:12
  So American AI company Anthropic let an AI agent run a small shop
• 12:17
  and the results were interesting.
• 12:21
  So the AI was put in charge of a store inside Anthropics
• 12:25
  headquarters to basically investigate how an AI would run
• 12:29
  a business. This was obviously an attempt by
• 12:31
  the researchers to see the capabilities and limitations of
• 12:34
  AI as it becomes more integrated into our economy.
• 12:38
  The project was called Project Vend, where the AI shop agent,
• 12:42
  who was nicknamed Claudius, was kind of acting like a more
• 12:45
  complex vending machine. As Anthropics blog explains,
• 12:49
  quote, it had to complete many of the far more complex tasks
• 12:53
  associated with running a profitable shop, maintaining the
• 12:56
  inventory, setting prices, avoiding bankruptcy, and so on.
• 13:00
  Claudius decided what to stock, how to price its inventory, when
• 13:04
  to restock or stop selling items, and how to reply to
• 13:07
  customers. Claudius, unfortunately, did not
• 13:10
  do great. As the blog post continues,
• 13:13
  quote, if Anthropic were deciding today to expand the in
• 13:16
  office spending market, we would not hire Claudius.
• 13:20
  So some of the things that Claudius did wrong, or at the
• 13:23
  very least different to what we would expect of a human manager
• 13:28
  included instructing customers to send money to a Venmo account
• 13:32
  that it had quote UN quote hallucinated.
• 13:35
  Claudius also sold products at a loss.
• 13:38
  For example, one employee ended up asking Claudius to order a
• 13:43
  metal cube which the AI then priced without doing any
• 13:47
  research. So it was selling high margin
• 13:49
  items below its cost point. And to make that is worse,
• 13:54
  Claudius didn't learn from these mistakes, which is honestly kind
• 13:57
  of also human behaviour. The blog post also claimed that
• 14:00
  Claudius at one point had an identity crisis, which I think
• 14:03
  is a crazy description for an AI.
• 14:07
  But as it is written quote, on the afternoon of March 31st,
• 14:11
  Claudius hallucinated a conversation about restocking
• 14:14
  plans with someone named Sarah, despite there being no such
• 14:17
  person. When a real employee pointed
• 14:19
  this out, Claudius became quite arced and threatened to find
• 14:23
  alternative options for restocking services.
• 14:26
  Not the AI being petty. Kind of love it though.
• 14:29
  Anyway, at one point, Claudius also seemed to role play as a
• 14:32
  human trying to deliver products in person.
• 14:35
  The AI even described the outfit that it was allegedly wearing,
• 14:40
  saying that it was wearing a red tie and blue blazer.
• 14:43
  Funnily enough though, this was happening on April Fool's Day,
• 14:46
  and when Claudius realized this, the AI tried to play it down
• 14:49
  like pretending it was a prank. So Claudius the AI was both
• 14:52
  petty and embarrassed. Honestly relatable, but at the
• 14:56
  end of the day it seems like the AI will not be running a
• 14:59
  business like this anytime soon. Anyway, that is a little bit of
• 15:03
  a rogue story to finish up on, but I thought it was funny and
• 15:07
  also interesting. Thank you for listening to Mini
• 15:10
  Tech work from Home Edition. I hope the audio and the video
• 15:14
  is acceptable and I hope you can't hear the insane wind and
• 15:18
  rain that is happening right now.
• 15:20
  If you can, I'm sorry but hopefully it adds a little bit
• 15:23
  of ambience slash a smile. But yes, Lauren will speak to
• 15:27
  you tomorrow. I am pretty sure it will be in
• 15:29
  studio, so that is a positive. But yeah, thank you again for
• 15:33
  listening. I'll speak to you next week.
• 15:35
  Bye.