The Download

Morning Brew Inc.

TL;DR: Sometimes you have to fight fire with fire, which Reddit might be doing by using AI to fight AI slop. But it also highlights how platforms that were once so valuable because of their human-generated content are now struggling to tamp down on a flood of AI spam—and it could have broader consequences.

What happened: Reddit revealed yesterday that it now uses LLMs to catch around 25,000 spammy posts or comments every day. The irony: AI probably played a big part in the platform’s spam problem getting so bad in the first place.

While Reddit doesn’t have a blanket policy on AI content (it does have a sitewide ban on spam), many subreddits have rules around disclosing AI content or even outright banned it. There’s no reliable estimate of how much of the site is now AI, but browse any busy thread and you’ll find a few commenters accusing others of using ChatGPT. All this poses a problem for a site that, per its CEO, wants to be “the most human place on the internet in a world flooded with AI slop.”

The internet’s town square: Since launching in 2005, Reddit has become an endless font of crowdsourced human opinion and expertise—the go-to place for both product recs and the juicy relationship drama that fuels group chats. Over the last few years, Google even tweaked its algorithm to downrank SEO sites engineered for clicks while boosting “helpful content” from forums like Reddit. (And all that human-generated text is unsurprisingly something AI companies want access to, with Google reportedly paying $60 million a year to train its AI on Reddit data.)

The gamification of AI search: AI is now threatening Reddit’s value proposition not just by flooding discussions with annoying, low-quality posts and comments, but also by helping companies manipulate search and chatbot answers. Brands reportedly use AI to blast mentions of their products on the site, knowing that Reddit content is scraped (and prioritized) by AI search and bots—a strategy called “answer engine optimization.”

Bottom line: Thanks to AI, it’s becoming harder to trust the once seemingly human-written content on platforms like Reddit—impacting the reliability of answers from search engines like Google and forcing us to become more discerning about what’s genuine and what’s not. —WK

      

Sponsored By PwC

We’re past the point of discussing potential. It’s high time for some strategy, applied learnings, and actionable insights.

AT&T’s leadership felt the same way. So when they spotted an opportunity to pilot a tool everyone’s talking about, they turned that chatter into movement with the right data.

PwC sat down with AT&T at Mobile World Congress to understand their story and success with AI, and we’re here to recap the learnings and discussion.

It’s not just about AI being the latest conversation starter for AT&T. It’s about their forward-thinking leadership spotting a variety of opportunities proactively, guiding the organization’s prioritization, and creating measurable business impact while others may still be experimenting.

Read all about it.

Your iPhone’s secret trackpad

If you’ve ever tapped the wrong spot in a text box on your iPhone and watched your cursor land literally anywhere except where you wanted it, there’s a surprisingly useful trick hiding in plain sight.

As Tech Brew reader Kevin, from the US, notes, “you can drag your cursor wherever you want to make corrections or add new text” instead of tapping around your screen trying to place the cursor in an exact spot.

How to do it: Write text in any app → press and hold the text box (or the space bar until the letters on the keyboard disappear) → slide your finger to move the cursor where you want it → let go to start editing the text. (A similar feature exists for many Androids.)

Why it works: Trying to place the cursor with your fingertip can sometimes feel like performing surgery with oven mitts on. This workaround gives you much better control and makes it easier to fix typos, insert words, or edit long messages. No more repeatedly stabbing at the screen and hoping for the best. —CM

If you have a tech tip or life hack you just can’t live without, fill out this form and you may see it featured in a future edition.

Sponsored By JumpCloud

Are your agents in check? Many organizations have successfully brought autonomous AI agents into production environments. But now, 92% report that they’re hitting a wall when trying to scale AI operations. In our recent article with JumpCloud, we explore how to address the main roadblocks. Give it a read.

The Zeitbyte

The data center that just won’t shut up

Morning Brew Inc., Photo: Adobe Stock

How long could you stay sane if you had to listen to this constant mechanical hum all day, every day? The loud droning is what residents of one Michigan town are dealing with, thanks to the 30-megawatt data center in their backyard. The sound clocks in at around 60 decibels, which is roughly the volume of a “normal conversation”—except this convo never ends, just like when your most annoying coworker insists on hovering at your desk. One local compared the constant noise to “living in a prison in our own yard.”

Now there’s a lawsuit claiming that the data center, which could expand to 340 megawatts (making it even noisier), didn’t do enough soundproofing. And these Michigan residents aren’t alone in getting litigious for a quiet night’s sleep—people in Wisconsin have also filed a class action against Microsoft’s Fairwater data center. The suit describes the noise as “not only excessive, but consistent and pervasive,” which is relatable to anyone who’s ever been trapped in a mall while “All I Want for Christmas Is You” plays on loop. —WK

Chaos Brewing Meter: ☕☕☕☕/5

• This South Korean firm could pass Nvidia to become the most profitable tech company in the world.
• Harry, yer a wizard! An open-source app turns tablets into Tom Riddle’s diary, secrets and all. Watch the words disappear here.
• Hollywood studios sued Midjourney, so the image-generation company is doing tit for tat—and it could uncover exactly how studios use AI.
• Four US states are seeking $1.4 trillion in penalties from Meta—very close to the company’s entire market cap.
• Forget drones. More than 100 autonomous ground vehicles built in the US have reportedly been quietly helping in Ukraine for months.
• If you’re sick of dropshippers and knockoff brands on Amazon, this serial entrepreneur has a Chrome extension for you.
• This state just became the first to mandate independent audits of AI companies’ safety practices.

^{*A message from our sponsor.}

This time last week...

Readers’ most-clicked story was about OpenAI really sticking to its “no side quests” directive with a new hardware announcement.