Latest / Mini Scroll / MINI TECH SCROLL: AI headlines on Google, OpenAI axes Sora, Tubi x TikTok & the accounts profiting off racist AI slop

Transcript

MINI TECH SCROLL: AI headlines on Google, OpenAI axes Sora, Tubi x TikTok & the accounts profiting off racist AI slop · Mini Scroll

• 0:00
  Hey everyone and welcome to Mini Scroll, the daily Internet
• 0:03
  culture podcast bring you the biggest stories from social
• 0:06
  media, the creator economy and the digital space every Monday
• 0:09
  through Thursday and sometimes Fridays.
• 0:11
  My name is Annabelle and I'm your Wednesday Mini Tech Scroll
• 0:14
  host and also the deputy editor of the Australian youth digital
• 0:18
  media brand Centennial World. Welcome back to mini Tech
• 0:21
  scroll. We are officially back from our
• 0:23
  one week break and honestly, that one week was necessary.
• 0:27
  I feel so much more refreshed because as my start to 2026 has
• 0:31
  honestly been quite cast and chaotic as I'm sure it honestly
• 0:35
  has been for a lot of people. So having that one week off was
• 0:39
  honestly super refreshing. I actually went down to
• 0:41
  Melbourne, which was really fun, and I also watched the
• 0:44
  Manosphere documentary on Netflix.
• 0:47
  I am actually really curious to hear what you guys think about
• 0:50
  it. It's honestly funny that I
• 0:51
  watched that on my break because it is kind of adjacent to my
• 0:53
  job, so I guess I can never truly log off.
• 0:56
  But yes, definitely share your thoughts about the documentary.
• 0:59
  I feel like it has been quite polarizing.
• 1:03
  Aside from that, I've also very much been enjoying the Hannah
• 1:06
  Montana anniversary or anniversary, I should say,
• 1:09
  content that has emerged over the last day or so that has
• 1:13
  truly uplifted my spirits. But yes, Please remember to
• 1:18
  rate, review, and subscribe. Feel free to check out our
• 1:21
  Patreon as well if you want more behind the scenes content of
• 1:24
  Centennial World. So yes, let's get into it.
• 1:28
  The first story that I wanted to mention today is that Google
• 1:31
  Search is now rewriting headlines using AII believe.
• 1:35
  This was first reported by The Verge a couple of days ago, but
• 1:38
  the publication found that its headlines were being rewritten
• 1:41
  in Google search results, presumably by AI.
• 1:45
  So the piece that I am pulling from was written by Sean
• 1:48
  Hollister and he goes through some examples of the headlines
• 1:53
  in the article. He actually references a
• 1:55
  headline by The Verge about Cluley, which is the startup we
• 1:58
  actually covered in my last mini text scroll episode 2 weeks ago.
• 2:02
  As he writes quote, Google reduced our headline.
• 2:04
  I use the cheat on everything AI tool and it didn't help me cheat
• 2:08
  on anything to just 5 words Cheat on everything AI tool.
• 2:12
  It almost sounds like we're in resourcing a product we do not
• 2:14
  recommend at all. So this feature, interestingly
• 2:17
  enough, is actually something that Google seems to be working
• 2:21
  towards. It was reported earlier in the
• 2:23
  year that Google Discover, which I believe is the personalized
• 2:26
  content feed on the Google app, started rewriting headlines with
• 2:31
  AI. And this was also reported by
• 2:33
  The Verge back in January. At the time of recording, it
• 2:37
  does seem like the Google Search AI written headlines are only a
• 2:41
  small percentage of the headlines that are published.
• 2:44
  But the very fact that this has extended from Google Discover to
• 2:48
  Google Search indicates to me that there is a chance that this
• 2:52
  feature will be more widely instituted moving forward.
• 2:56
  And this is something that Hollister really tries to get
• 2:58
  across in his piece as well as he writes.
• 3:01
  Quote While Google says this is an experiment, you shouldn't
• 3:04
  assume that means the company won't roll it out more widely.
• 3:07
  Because Google originally told us it's AI headlines and Google
• 3:10
  Discover we're an experiment too.
• 3:12
  A month later, it told us those AI headlines are now a feature,
• 3:16
  one that performs well for user satisfaction.
• 3:18
  So the Verge also did reach out to Google to try and gather more
• 3:22
  information on why they've started rolling out this
• 3:25
  feature. And a Google spokesperson said
• 3:27
  it was a way to try and better match a link to a user's query
• 3:31
  in the search bar, which will in theory, they say lead to higher
• 3:36
  engagement. And it was also confirmed that
• 3:38
  these headline changes are not limited to news sites.
• 3:42
  As you probably would have guessed, there has been quite a
• 3:44
  strong reaction to these changes and I have a lot of thoughts.
• 3:48
  So we have to get into them for market is a feature like this
• 3:52
  completely disrupts everything we know about SEO search engine
• 3:55
  optimization. And this is already been
• 3:58
  changing quite dramatically with the rise of AI chat bots and how
• 4:01
  people get their information from a BLOB of text these days
• 4:05
  instead of clicking on links. So this seems to maybe just be
• 4:09
  another nail in the coffin, if you will.
• 4:12
  But for media, which is probably where I feel more comfortable
• 4:15
  talking about, there are so many issues at play here.
• 4:19
  First, it seems like there isn't really an option for publishers
• 4:23
  to opt out of this feature at this stage, which ultimately
• 4:26
  gives these publications less control over their editorial
• 4:30
  tone, vision, and voice. That isn't even to mention the
• 4:34
  biases that are present within generative AI tools that can be
• 4:38
  represented in these headlines, which can be harmful for these
• 4:41
  publication because as we've kind of already touched on, it
• 4:45
  can lead to inaccurate headings that don't represent the content
• 4:48
  of the article. Not only is this kind of unfair,
• 4:52
  but it can also foster further distrust in media at a time when
• 4:56
  that's already extremely high. I also don't think this bodes
• 4:59
  very well in our current media landscape with the rise of kind
• 5:03
  of click bait journalism and green screen style storytelling.
• 5:08
  Today a lot of people consume a new story through a green screen
• 5:13
  format on TikTok on Instagram where people are talking face to
• 5:16
  camera, often with an article in the background where you can
• 5:19
  probably just see the headline. It's already difficult enough to
• 5:22
  make the effort to go and find this article if you want to go
• 5:26
  read more because it can't be LinkedIn an immediately
• 5:29
  accessible way. Arguably, an update like this
• 5:32
  might only make it more confusing to find said article.
• 5:37
  On top of this, when we consider the people delivering these
• 5:40
  stories, there are some of them of course not all of them, who
• 5:44
  aren't operating with good intentions or the same fact
• 5:47
  checking and editorial processes as publications.
• 5:51
  And they might use a headline to verify the argument they are
• 5:56
  making. And here, AI generated headlines
• 6:00
  might not be totally accurate to the argument of the article.
• 6:04
  I know we're kind of talking in hypotheticals here, but I think
• 6:08
  the key take away is how it might be harder to gather
• 6:11
  accurate information with an update like this.
• 6:14
  As far as consumers, I have already seen quite a lot of
• 6:17
  discourse saying that people are jumping ship and moving to
• 6:20
  different search engines. DuckDuckGo, which we've
• 6:23
  mentioned before, seems to be the favorite as you can opt out
• 6:26
  of AI generated content as one user put it on blue sky quote.
• 6:31
  This put me over the edge. I just officially switched my
• 6:34
  default search engine to DuckDuckGo.
• 6:36
  Google as a search engine is the opposite of why it was so
• 6:39
  amazing when it started the in certification of the Internet
• 6:42
  continues, but even when these users switch it is definitely
• 6:45
  worth noting that Google remains still so widely used.
• 6:49
  So a lot of people, if they continue rolling out this update
• 6:53
  will be likely consuming inaccurate headlines and just
• 6:58
  running with that because these days, a lot of people just read
• 7:00
  headlines and that is their understanding of the news.
• 7:03
  And I think my last point that I want to mention about this story
• 7:06
  is that I think this update is symptomatic of a widespread
• 7:10
  issue that we are seeing across tech, social media, and Internet
• 7:13
  culture at the moment where introducing a feature like this
• 7:16
  is objectively unnecessary. Like people in these companies
• 7:20
  have already decided on the headline that best displays
• 7:23
  their argument that they think people will click on.
• 7:26
  They've done the work. AI just like is isn't necessary
• 7:30
  here. I think it's just an example of
• 7:31
  these companies trying to find uses for AI and creating an
• 7:35
  issue that just doesn't need to be fixed.
• 7:38
  Not everything needs to be AI. And I know I'm probably
• 7:40
  approaching to the choir here, but I'm definitely curious to
• 7:42
  hear your thoughts. OK, let's move on to the next
• 7:45
  story. And if you were watching, you
• 7:47
  will have noticed that probably everything about the set and me
• 7:51
  looks different from about a second ago.
• 7:54
  And that is because I'm jumping back on the mic with another
• 7:57
  story that I only saw after recording the full mini tech
• 8:01
  episode. So this is actually hours later,
• 8:04
  but I wanted to quickly jump back on the mic.
• 8:06
  And the story in question is that Open AI has officially
• 8:09
  discontinued Sora. So I'm sure we're all familiar
• 8:12
  with Sora. But in case you've missed my
• 8:14
  previous coverage on the topic, Sora is a text to video model
• 8:18
  developed by Open AI. And in September last year, Sora
• 8:22
  2 was rolled out along with an iOS app.
• 8:25
  And it was a massive deal. It went super viral because of
• 8:28
  all the AI. I generated clips with Jake Paul
• 8:31
  in them, which I have discussed at length, it feels like on this
• 8:34
  podcast. And this was a result of the
• 8:36
  Cameo feature, which Jake had actually opted into to kind of
• 8:40
  promote the app. But basically it allowed users
• 8:43
  to drop themselves or others straight into any Sora scene.
• 8:47
  This morning, it was reported that Open AI is planning to pull
• 8:50
  the plug on Sora. This was first reported by the
• 8:54
  Wall Street Journal. And as they wrote, quote, the
• 8:56
  move is one of a number of steps Open AI.
• 8:59
  AI is taking to refocus on business and coding functions
• 9:03
  ahead of a potential initial public offering as soon as the
• 9:06
  fourth quarter of this year. Sora's X account also issued a
• 9:09
  statement reading quote. We're saying goodbye to Sora.
• 9:12
  To everyone who created with Sora, shared it and built
• 9:15
  community around it. Thank you.
• 9:17
  What you made with Sora matted. And we know this news is
• 9:21
  disappointing. We'll share more soon, including
• 9:23
  timelines for the app and API and details on preserving your
• 9:26
  work. So it's clear that Open AI are
• 9:29
  kind of now moving to focus on other research areas.
• 9:33
  But alongside this, the company is rolling more of its
• 9:36
  capabilities into a single super app that is planning to merge
• 9:40
  ChatGPT, the Codex AI coding app, as well as the Atlas AI
• 9:44
  powered browser into one. Which kind of reminds me of what
• 9:47
  X was trying to do when it was trying to be the everything app.
• 9:51
  But what is interesting about the closure of Sora is that it's
• 9:55
  also complicated the deal between Disney and Open AI.
• 9:59
  So the companies announced a deal towards the end of last
• 10:02
  year where Disney made a $1 billion equity investment in
• 10:05
  Open AI, where Sora was then able to generate video drawing
• 10:10
  on 200 Disney, Marvel, Pixar and Star Wars characters.
• 10:14
  And it was set to be a three-year deal.
• 10:16
  From reports, it kind of seems like Urban AI's decision to
• 10:19
  close Sora was a little bit of a rug pull, quote, UN quote, for
• 10:22
  Disney, at least according to a source who spoke to writers.
• 10:26
  All in all, though, I would argue that the closure of Sora
• 10:31
  doesn't necessarily bode well for the future of short form AI
• 10:34
  video because of the way it was so viral, but the way other apps
• 10:39
  continue to pop up and the internet's continued obsession
• 10:42
  with AI content. AI fruit, Love Island, I'm
• 10:44
  looking at you. I don't think that we can take
• 10:48
  this as a bellwether or an indicator that this kind of
• 10:52
  content is going to die, but we can dream that the bubble is
• 10:56
  bursting. OK, let's get back to the rest
• 10:59
  of the episode Now. The next story I want to cover
• 11:02
  is a little bit more Internet culture esque.
• 11:04
  I wanted to talk about this collaboration between Tiktok and
• 11:08
  Tubi. Also, I don't know if it's Tubi
• 11:11
  or Tubi. My instinct as an Australian is
• 11:14
  to say To be, but I feel like that just sounds wrong so I
• 11:17
  might just start switching between them.
• 11:19
  Anyway, last week it was announced that To Be, which is a
• 11:22
  free streaming service that is also owned by Fox, is working
• 11:27
  with TikTok on a quote UN quote creative US incubator which is
• 11:31
  set to help creators develop original longer form shows.
• 11:35
  I will read a little bit of the press release so you can get
• 11:38
  more context as it reads. Quote Creator shows will
• 11:41
  Premiere X exclusively on Tubi, giving creators the opportunity
• 11:45
  to reach Tubi's more than 100 million monthly active users.
• 11:50
  Selected creators will develop original series across a variety
• 11:54
  of genres in both scripted and unscripted formats, with
• 11:57
  creative direction shaped by the creator and supported by Tubi.
• 12:01
  TikTok will help identify creators for consideration in
• 12:04
  the program, with to be sent to announce an initial cohort of
• 12:07
  participants later this summer. So this is not the first time
• 12:11
  that the streaming platform has collaborated with TikTok
• 12:13
  creators. Keelan Campbell's infamous
• 12:16
  character Terry Joe actually has a film on the streaming platform
• 12:19
  called Missionary in Miami that was released early this year.
• 12:23
  I think this collaboration is just the latest example of
• 12:27
  creator LED shows becoming increasingly legitimized in the
• 12:30
  pop culture and entertainment space.
• 12:33
  Last year for example, I think I may have covered it in mini tech
• 12:36
  scroll. There was a lot of discussion
• 12:37
  about how YouTube is seemingly leaning into these interview
• 12:41
  style late night shows. They actually ended up hosting
• 12:45
  an event which was kind of like a Film Festival style showcase
• 12:49
  where they ended up platforming creators like Julian Shapiro
• 12:52
  Barnum who is the host of Recess Therapy, as well as Brittany
• 12:56
  Broski who is of course the host of Royal Court.
• 12:58
  Between this and the Tubi TikTok incubator, it's really
• 13:03
  interesting to see creators afforded these opportunities and
• 13:06
  backed by a massive platforms. But it also kind of reminds me
• 13:10
  of like the mid twenty 10s when we saw these creator focused
• 13:13
  studios like pumping out content.
• 13:16
  I think Awesomeness TV is probably one of the most iconic
• 13:20
  examples. It was launched back in 2012 and
• 13:23
  was like a multi channel network aimed at young audiences who
• 13:26
  were kind of moving away from TV and film.
• 13:29
  One of the most iconic shows I remember was Foursim which was
• 13:32
  like a romantic comedy that had Jen McAllister AKA Gen.
• 13:36
  X pen in it. But I also remember Guidance
• 13:38
  which I feel like I know no one else in the world who watched
• 13:41
  that show. But it featured Michelle
• 13:43
  Trachtenberg who was Georgina in Gossip Girl as well as Amanda
• 13:47
  Steele, also known as makeup by Mandy 24 who I absolutely loved
• 13:51
  back in that era. To me, the Tubi slash TikTok
• 13:55
  deal definitely feels like an evolution from these Alia
• 13:58
  creator Focus studios, but it also has the added element of a
• 14:01
  creator who is a custom to making shorter form content move
• 14:06
  to longer form content, which definitely isn't impossible.
• 14:08
  We've seen that with so many TikTok creators launching
• 14:11
  podcast, but of course, it may be hard for some creators and
• 14:15
  also audiences may not want to see certain creators in longer
• 14:18
  form formats, but I'm definitely curious to see which creators
• 14:22
  are selected. This brings us to our last story
• 14:25
  for today, and this one is actually a little bit of a
• 14:28
  confronting 1, and it's referencing a recent
• 14:31
  investigation about how black women are being represented by
• 14:35
  AI and the rise of digital blackface.
• 14:38
  So we have covered the rise of racist and discriminatory AI
• 14:42
  generated content against black women specifically before,
• 14:46
  particularly after the launch of Google's VO3, which was a text
• 14:51
  to video generator that was launched last year.
• 14:54
  After its release, there was a flood of racist TikTok content
• 14:58
  that fed into disgusting stereotypes about black women.
• 15:02
  And to make matters worse, there weren't really any comments that
• 15:05
  I saw at the time that condemned this kind of content or content
• 15:10
  warnings on the platform itself. As I said then, and I will say
• 15:13
  it again now, obviously I don't know what it feels like to see
• 15:18
  that kind of content as a Black woman and have that likeness
• 15:21
  used in harmful and stereotypical ways.
• 15:24
  I can really imagine how confronting, emotionally taxing,
• 15:28
  and just tiring it would become because it is just so constant.
• 15:32
  So it's definitely worth continuing to platform and speak
• 15:35
  about in hopes that something can change.
• 15:37
  And that's why I wanted to cover this story today.
• 15:39
  I specifically wanted to talk about this investigation that
• 15:43
  I've been seeing making the rounds online.
• 15:46
  It was a collaboration between AI literacy creator Jeremy
• 15:49
  Carrasco, student researcher Angel Nulani, and BBC News
• 15:54
  Arabic. They ended up analyzing over 100
• 15:58
  social media accounts that depicted the likeness of black
• 16:00
  women through AI generated characters and personas.
• 16:04
  And these accounts were across both Instagram and TikTok.
• 16:08
  And actually most of them were created relatively recently, in
• 16:12
  the last eight months or so. Some of these accounts actually
• 16:15
  used content from real people, of course without their consent,
• 16:19
  and used AI tools to edit them. Meanwhile other accounts were
• 16:23
  just completely AI generated scenes.
• 16:26
  So I thought I would go through some of the findings but if you
• 16:29
  want to read the pieces yourself they are LinkedIn the
• 16:34
  description one of them is from the BBC and another one is from
• 16:37
  the riddance dot AI sub stack. But yes, let's get into some of
• 16:40
  the findings. Probably one of the most
• 16:42
  confronting findings from this investigation was that a
• 16:46
  majority of these accounts, and they counted 68, promoted
• 16:50
  monetization on sites that they link to which sell explicit AI
• 16:54
  generated media of the characters in question.
• 16:58
  Which unfortunately is just emblematic and just another
• 17:01
  example of the constant fetishization and exploitation
• 17:05
  of black women. As the Riddance dot AI piece
• 17:08
  writes quote, by depicting black women they perpetuate a long
• 17:12
  history of exploiting black people using caricatures and
• 17:15
  self hating characters to attract morbidly curious or
• 17:19
  racist uses. In the piece they suggest that
• 17:22
  the accounts that are not linking out to websites that
• 17:26
  have explicit AI generated media, they probably just aren't
• 17:30
  big enough to monetize it yet and it probably will be coming,
• 17:33
  which is really grim. Another key finding that I
• 17:36
  thought was worth mentioning was that most of the AI generated
• 17:39
  accounts express anti blackness, with many of them making
• 17:42
  comments about wanting to be European.
• 17:45
  There is a long history of forcing black people to adhere
• 17:48
  to white slash western standards in an attempt to enforce and
• 17:53
  maintain white supremacy. Just one historical example
• 17:57
  based here in Australia was the Stolen Generation, which was the
• 18:00
  whitewashing of indigenous children as they were forcibly
• 18:03
  removed from their families and communities throughout the 20th
• 18:07
  century up until about 1970. And to me, the dialogue from
• 18:11
  these AI generated characters making these anti black
• 18:15
  statements just feels like the most modern example of upholding
• 18:19
  these norms, while at the same time they're profiting off this
• 18:22
  demographic, fetishizing them. And of course we continue to see
• 18:26
  AAVE, for example, and black beauty standards as another
• 18:29
  example, being completely Co opted and embraced within other
• 18:33
  aspects of the digital landscape and pop culture.
• 18:36
  This racism is furthered by the way that many of these AI
• 18:39
  avatars are presented as well. The researchers noted the rise
• 18:43
  of AI generated characters that have what they describe as quote
• 18:47
  UN quote impossibly black skin, meaning that there is virtually
• 18:50
  no undertone to these AI generated character skin.
• 18:54
  Part of the reason that this is so harmful is because
• 18:56
  historically racist portrayal of black people, like in minstrel
• 19:00
  shows, had people paint their face just literally black.
• 19:04
  And the researchers make the argument that the people who are
• 19:07
  creating these AI characters with quote UN quote, impossibly
• 19:11
  black skin are actually doing it on purpose.
• 19:14
  As the report reads, quote, if it's not obvious, AI generated
• 19:18
  characters will have normal skin tones by default.
• 19:21
  Human creators have to deliberately try to make
• 19:24
  impossibly black characters. In testing, we have confirmed
• 19:27
  that prompting Flux 2 for very dark pitch black skin with no
• 19:32
  undertones will yield impossibly black skin.
• 19:35
  And this conversation has actually come up before within
• 19:38
  different sects of Internet culture.
• 19:40
  For me, the one I remember most was probably around the Euphoria
• 19:44
  Foundation shade 600 that completely left the digital
• 19:48
  world in shock. Beauty creators, notably Gloria,
• 19:52
  described the color of the foundation as jet black paint or
• 19:56
  minstrel show black, rather than an actual like usable foundation
• 20:00
  for darker skin tones. Of course, with the history of
• 20:04
  minstrel shows like I've mentioned and other racist
• 20:07
  depictions of black people, it's obviously why this was cast as
• 20:11
  discriminatory and racist. Since this is an issue and an
• 20:14
  experience that disproportionately impacts
• 20:16
  people of color, I did want to include the perspective of a
• 20:20
  Black woman to really express how harmful this content is.
• 20:23
  This quote is pulled from Angel Nulani, who is the student,
• 20:27
  researcher and editor that contributed to this piece.
• 20:31
  And this is part of what she wrote.
• 20:33
  Quote. What upsets me most is not that
• 20:35
  these characters are self hating, but that there is no
• 20:38
  self for the majority of the people who are behind these
• 20:41
  accounts. The only black people they know
• 20:43
  are the women they generated. They were not born black, they
• 20:47
  chose to be black, and yet they spend so much time distancing
• 20:50
  themselves from it. The appeal is not simply a black
• 20:53
  woman desiring white men. The appeal is in a black woman
• 20:57
  desiring a complete absorption into whiteness.
• 21:00
  The men they attract are not suitors, they are saviours.
• 21:03
  I've tried so long to dissociate.
• 21:05
  As a black woman that's been online since I was young, I
• 21:08
  assumed I saw it all. Nothing could really hurt me at
• 21:11
  first, all I wanted was to be objective and rational and done
• 21:14
  with it, but this cuts me down to my core.
• 21:16
  I would definitely recommend reading the sub stack piece if
• 21:19
  you want to read her full quote because obviously it's always
• 21:23
  important to hear from the groups that are affected by
• 21:26
  these issues. Anyway, I thought this was a
• 21:28
  really important story and investigation to touch on,
• 21:32
  especially because a lot of these accounts are still active
• 21:35
  and that just tells me that we need to keep talking about these
• 21:37
  stories and also reporting these accounts if you ever come across
• 21:41
  them. That brings us to the end of
• 21:43
  Mini Tech Scroll for today. I really hope you guys enjoyed
• 21:47
  this weeks episode. I will see you next Wednesday
• 21:50
  and Lauren will speak to you tomorrow.
• 21:52
  But there is also the last official episode of Touch Grass
• 21:56
  out today, so you should definitely go watch that.
• 22:00
  But yeah, thank you so much for listening.
• 22:02
  Bye. Centennial World Podcast Network
• 22:04
  operates on Gadigal land. We would like to acknowledge the
• 22:07
  Gadigal of the Eora Nation, the traditional custodians of this
• 22:10
  land, and pay our respects to elders both past and present.