Latest / Connecting the Dots / Scam Ads, Market Moves, and a Hollywood Tragedy

Transcript

Scam Ads, Market Moves, and a Hollywood Tragedy · Connecting the Dots

• 0:00
  Today's tech news moves fast, but understanding
• 0:03
  it shouldn't be hard. That's why Snarfl Solutions
• 0:06
  Group brings you Connecting the Dots, where we
• 0:09
  break down what's happening in tech and why it
• 0:11
  matters to you. Now, let's dive in with your
• 0:14
  hosts. Welcome to Connecting the Dots. I'm Alex.
• 0:18
  And I'm Morgan. And a very good morning to you.
• 0:20
  We're glad you're with us. It is Monday, December
• 0:23
  15th, 2025. And we're coming to you from a...
• 0:27
  Well, a very foggy Sacramento, California. Yeah,
• 0:31
  definitely foggy. The high today is barely going
• 0:33
  to hit 50 degrees. So it's a cool start to the
• 0:36
  week. For sure. And if you're out east of Lake
• 0:38
  Syria and Ontario, you're dealing with that heavy
• 0:40
  lake effect snow. So please stay warm. And to
• 0:44
  everyone in the Pacific Northwest, that atmospheric
• 0:46
  river is no joke. heavy rain, gusty winds, keep
• 0:51
  an eye out for potential flooding. Definitely.
• 0:53
  Okay, let's take a quick look at the markets.
• 0:55
  They closed on Friday with, I'd say, a mixed
• 0:57
  but pretty resilient feel. Right, the Dow Jones
• 1:00
  was down just a little 0 .17 % but... The weekly
• 1:04
  trend was still positive, which is good to see.
• 1:06
  And the S &P 500 was pretty similar, closing
• 1:08
  down about 0 .30 % for the day. But then you
• 1:10
  have Bitcoin. It sort of bucked that trend. It
• 1:12
  did. It closed up almost a full percent at $89
• 1:15
  ,000 and 17 cents. Still down for the week. But
• 1:18
  hey, we'll take that daily momentum. We will.
• 1:21
  So before we jump into the two big stories we've
• 1:23
  been tracking for you, we want to thank our sponsor.
• 1:26
  That's right, Old Glory. If you have a techie
• 1:29
  friend who's also into, you know, sports or music,
• 1:32
  this is the perfect place for holiday gifts.
• 1:35
  They're the iconic merch store for fans. It's
• 1:38
  a great spot. You can head over to OldGlory .com
• 1:41
  and when you check out, use the promo code SNARFL
• 1:43
  to get 15 % off your entire purchase. Alright,
• 1:47
  let's connect some dots. We've been going through
• 1:49
  the SNARFL Tech article report, what we call
• 1:51
  the Morningstar, and today we're looking at two
• 1:54
  huge stories. They really show two different
• 1:56
  sides of the tech economy, don't they? They really
• 1:58
  do. First, we're going to break down the frankly,
• 2:02
  shocking financial incentives behind Meta's advertising
• 2:04
  platform. And then we'll pivot to NVIDIA and
• 2:07
  its really aggressive new strategy to dominate
• 2:10
  the open source AI market. It's a tale of two
• 2:12
  giants. One seems to be prioritizing profit over
• 2:15
  user safety, and the other is trying to redefine
• 2:18
  the whole game through transparency. Exactly.
• 2:21
  Let's start with Meta. So we're talking about
• 2:23
  Facebook, Instagram, WhatsApp. New internal documents
• 2:26
  have come out and they paint a really troubling
• 2:28
  picture. Traveling is putting it lightly. These
• 2:31
  documents suggest Meta is making billions, billions
• 2:34
  with a B from fraudulent ads. The numbers are
• 2:37
  just staggering. I mean, internal projections
• 2:39
  showed Meta expected to make about 10 % of its
• 2:41
  2024 revenue from ads for scams and banned goods.
• 2:45
  Wait, let's put a dollar figure on that. That's
• 2:47
  roughly $16 billion. $16 billion. And you know,
• 2:51
  the exposure for users is just as mind -blowing.
• 2:54
  The platforms show an estimated 15 billion high
• 2:57
  -risk scam ads. Every single day 15 billion a
• 3:01
  day. That's a that's a tidal wave of fraud and
• 3:03
  it is hitting people's wallets Hard fraud losses
• 3:06
  in the US shot past twelve and a half billion
• 3:08
  dollars in 2024 That's a 25 % jump in just one
• 3:12
  year and social media is the main pipeline for
• 3:14
  this, right? It really is. Scams that started
• 3:17
  on social media platforms accounted for almost
• 3:20
  $2 billion in losses last year. And Meta's own
• 3:24
  internal analysis confirmed their products were
• 3:27
  involved in about a third of all successful scams
• 3:29
  in the U .S. OK, so that leads to the obvious
• 3:31
  question. If the problem is this huge, why aren't
• 3:35
  they stopping it? Well, this is where it gets
• 3:38
  really cynical. It all comes down to the internal
• 3:41
  mechanics of their ad system. It's the profit
• 3:43
  motive laid bare. Their system is apparently
• 3:46
  set up to only ban an advertiser if there's a
• 3:48
  95 % certainty of fraud. 95%. Think about how
• 3:53
  high that bar is. Anything below that, the bad
• 3:55
  actor just gets to stay on the platform. But
• 3:57
  they don't just stay, it's worse than that. It
• 4:00
  is. Instead of removing them, meta. Well, they
• 4:02
  penalize them by charging them higher ad rates.
• 4:04
  So they're still profiting. In fact, they're
• 4:06
  profiting more from the sketchy advertiser than
• 4:08
  a legitimate one. Exactly. You just hit the nail
• 4:10
  on the head. They've made fraud a more expensive,
• 4:13
  but still totally viable, cost of doing business
• 4:16
  for scammers. So just to be crystal clear, instead
• 4:19
  of protecting their users from harm, they're
• 4:22
  just charging the source of that harm a premium.
• 4:25
  That's right. And it creates this perverse incentive.
• 4:28
  The internal documents even showed that their
• 4:30
  trust and safety teams had specific limits on
• 4:33
  how much revenue they were allowed to lose by
• 4:36
  kicking off suspicious accounts. Unbelievable.
• 4:38
  And it gets even worse for the person who's just
• 4:40
  scrolling through their feed, doesn't it? Oh,
• 4:42
  yeah. That's what they call the worsening trap.
• 4:45
  If you accidentally click on one of those scam
• 4:47
  ads, maybe you're just curious, the algorithm
• 4:49
  tags you. It tags you as interested. Yep. And
• 4:52
  then, because the whole system is designed to
• 4:54
  maximize engagement and revenue, it starts feeding
• 4:57
  you even more similar fraudulent ads. It's a
• 5:00
  feedback loop from hell. It basically helps the
• 5:03
  scammers find their most vulnerable targets.
• 5:05
  Precisely. Now, in fairness, we have to note
• 5:07
  that a Metta spokesman, Andy Stone, he did push
• 5:10
  back. He said the $16 billion estimate was rough
• 5:13
  and overly inclusive. Okay. And he also pointed
• 5:16
  out that they've reduced user reports of scam
• 5:18
  ads globally by 58 % over the last 18 months.
• 5:22
  Which is a real number, and we should include
• 5:23
  that context. But the fact that these internal
• 5:27
  revenue guardrails for safety teams exist, to
• 5:31
  me, that speaks louder than any public statement.
• 5:33
  It does. It raises that huge question about platform
• 5:36
  responsibility. So until that's sorted out, the
• 5:39
  responsibility really falls on you, the listener.
• 5:41
  What is the single most important thing someone
• 5:44
  can do to protect themselves? OK. First, just
• 5:47
  be vigilant. You know, if a price looks too good
• 5:50
  to be true, it is. Look for brand new accounts
• 5:53
  with no history, no followers. Those are all
• 5:55
  red flags. But the real line of defense is how
• 5:58
  you pay. Credit card, not a debit card. Say it
• 6:01
  again. Always. Always use a credit card for any
• 6:04
  purchase you make through a social media ad.
• 6:06
  Why is that so critical? Because with a credit
• 6:08
  card, your maximum liability for fraud is legally
• 6:11
  capped at $50, and most companies just waive
• 6:13
  it entirely. Yeah. They investigate, they handle
• 6:15
  the charge back, and your actual money stays
• 6:18
  safe. Whereas with a debit card, that cash is
• 6:20
  gone from your checking account instantly? Instantly.
• 6:23
  And getting it back is a nightmare. The protections
• 6:26
  are much weaker. So protect your cash. Use a
• 6:29
  credit card. And if you do get scammed, contact
• 6:31
  your bank or credit card company immediately,
• 6:33
  then report it to the FTC at reportfraud .ftc
• 6:36
  .gov. That helps them track these operations.
• 6:39
  That is such a critical takeaway. It's frustrating
• 6:42
  that we have to be this defensive, but it's the
• 6:44
  world we live in. It is. OK, let's shift gears
• 6:46
  completely from consumer risk to the cutting
• 6:50
  edge of enterprise technology. Our second story
• 6:53
  is about Nvidia, right? So Nvidia. Everyone knows
• 6:56
  them for their chips, their GPUs that power basically
• 7:00
  the entire AI world. Well now they're getting
• 7:02
  into the software side. They've just launched
• 7:04
  a new family of open source AI models called
• 7:07
  Nemotron 3. And this puts them in direct competition
• 7:10
  with models like Meta's llama, right? Likely.
• 7:12
  It is, I have to say, a brilliant strategic hedge.
• 7:16
  NVIDIA isn't just content being the arms dealer
• 7:18
  for the AI revolution anymore. They want to be
• 7:20
  one of the architects. So they're protecting
• 7:22
  themselves in case rivals like Google. start
• 7:25
  making their own chips that are just as good.
• 7:27
  Exactly. And at the same time, they're trying
• 7:29
  to define what the next generation of AI for
• 7:32
  businesses should even look like. Okay, so Nimotron
• 7:34
  3 comes in different sizes. Nano, super, ultra.
• 7:37
  But the real story here is the technology underneath
• 7:40
  it all. That's where the breakthrough is. They're
• 7:43
  using something called a hybrid latent mixture
• 7:45
  of experts architecture. We can just call it
• 7:47
  MoE. MoE, okay. And why is that so important?
• 7:50
  because it's the key to building something called
• 7:53
  agentic AI. Agentic AI. That sounds very sci
• 7:56
  -fi. Bring that down for us. So think about today's
• 7:59
  AI. It's usually one giant model, like a single
• 8:01
  brain trying to do everything. Agentic AI is
• 8:04
  different. It's more like a team of specialized
• 8:06
  AI agents that all work together on complex,
• 8:09
  multi -step tasks. So instead of one brain, it's
• 8:12
  a team of expert brains. You got it. So instead
• 8:14
  of asking one AI to say, plan a marketing campaign.
• 8:18
  You'd have one agent do the research, another
• 8:19
  write the copy, a third agent handle the budget.
• 8:23
  You get the idea. I do. And this MOE architecture
• 8:26
  is what makes that team work efficiently. That's
• 8:28
  the whole point. With a traditional model, when
• 8:31
  you ask a question, the whole giant brain has
• 8:34
  to wake up and process it. With MOE, only the
• 8:37
  specific experts needed for that task get activated.
• 8:40
  So it uses way less computational power. Which
• 8:42
  means it costs way less money. And cost is measured
• 8:45
  in tokens, right? Token throughput and reasoning
• 8:47
  tokens. This is the accounting of AI. It is.
• 8:50
  And it matters hugely to a business's bottom
• 8:53
  line. Every little piece of information an AI
• 8:56
  generates is a token. Complex tasks can take
• 8:59
  hundreds of AI calls, and those token costs just
• 9:02
  explode. So how much of a difference does this
• 9:05
  new architecture make? It's huge. The Nemetron
• 9:08
  3 Nano model gets up to four times higher token
• 9:11
  throughput, so four times faster processing,
• 9:13
  and it cuts the cost of generating reasoning
• 9:16
  tokens by up to 60%. Wow. For a big company running
• 9:19
  millions of AI processes, that's millions of
• 9:22
  dollars in savings. It's a game changer for costs.
• 9:25
  So NVIDIA is solving the biggest barrier to adopting
• 9:27
  next -gen AI, which is the price. A very smart
• 9:30
  move. It is. And they're also making a big deal
• 9:32
  about transparency. Yeah, this is another key
• 9:34
  part of the strategy. They're not just releasing
• 9:36
  the models. They're releasing the trillions of
• 9:39
  tokens of training data they use to build them.
• 9:42
  They're going for true open source status. Which
• 9:45
  is a direct shot at Meta. It is. Meta's llama
• 9:49
  models have kind of stagnated a bit. And there
• 9:52
  are all these signals that their next big model
• 9:55
  codenamed Avocado might be closed source. So
• 9:59
  as Meta potentially pulls back from being open,
• 10:02
  NVIDIA is rushing in to fill that vacuum. And
• 10:05
  businesses are noticing. Major players like Accenture,
• 10:08
  CrowdStrike, Deloitte, they're already integrating
• 10:10
  Nematron 3. They need that trust and efficiency.
• 10:13
  So it sets up this fascinating contrast for the
• 10:15
  next year. You have Meta focusing on proprietary
• 10:18
  consumer tech and NVIDIA championing open, cost
• 10:21
  -effective enterprise tech. And what's interesting
• 10:23
  is how it will affect you, the listener. NVIDIA
• 10:26
  making AI cheaper and more efficient means the
• 10:28
  automated tools you use every day, from customer
• 10:30
  service bots to your software, are about to get
• 10:32
  a lot faster and a lot smarter. Very cool. Okay,
• 10:34
  before we wrap things up, we need to give a quick
• 10:36
  shout out to our other sponsor. That's Scony,
• 10:39
  the premium golf shoe store. If you're a golfer
• 10:42
  or know someone who is, you really have to check
• 10:45
  out their stuff. You do. Head over to scony .com
• 10:47
  and just like before, use the promo code SNARFL
• 10:50
  at checkout for 15 % off your entire purchase.
• 10:53
  Okay, so a quick recap of the two big takeaways
• 10:55
  from today. First, the economics of Meta's ad
• 10:59
  platform are troubling. It means you have to
• 11:02
  be vigilant and please use a credit card, not
• 11:05
  a debit card. It's your best defense. And second,
• 11:08
  Nvidia is making a huge strategic play to lead
• 11:11
  enterprise AI. They're launching the super -efficient
• 11:14
  Nimrotron 3 models to make that next generation
• 11:17
  of agentic AI much faster and, crucially, much
• 11:21
  cheaper to run. The thread connecting both stories
• 11:23
  is really about incentives, isn't it? It really
• 11:25
  is. Whether it's the incentive to profit from
• 11:28
  fraud or the incentive to dominate a new market
• 11:30
  through efficiency, understanding what drives
• 11:32
  these companies is key. We're here to help at
• 11:34
  Snarful, so reach out on our website if you have
• 11:36
  any more questions. That's a wrap on today's
• 11:42
  top stories. Remember, tech doesn't have to be
• 11:45
  complicated and we're here to help connect the
• 11:47
  dots. If you've got questions or want to learn
• 11:50
  more, visit us at snarful .com. Thanks for listening
• 11:54
  and we'll catch you next time.