Latest / Connecting the Dots / Connecting the Dots - 6-17-2025

Transcript

Connecting the Dots - 6-17-2025 · Connecting the Dots

• 0:00
  Today's tech news moves fast, but understanding
• 0:03
  it shouldn't be hard. That's why Snarful 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. We're part of the team here at
• 0:20
  SNARFL Solutions Group in Sacramento, California.
• 0:22
  Thanks for tuning in. Today is Tuesday, June
• 0:24
  17th, 2025. And if you're here in Sacramento
• 0:27
  with us, hope you staying cool. It's looking
• 0:28
  sunny out there. Heading for a high of about
• 0:31
  93 degrees Fahrenheit today. Definitely warm.
• 0:35
  Quick look at the markets before we jump in.
• 0:36
  The Dow Jones closed yesterday at 42 ,394 .79.
• 0:42
  And the S &P 500 finished the day at 6 ,006 .71.
• 0:47
  So some movement there. All right. So today we're
• 0:50
  diving into our internal snarfle tech article
• 0:53
  report. We've picked out the top three tech stories
• 0:56
  that caught our eye. Yeah. And our goal, as always,
• 0:58
  is to break them down, figure out what's really
• 1:00
  going on, and connect the dots to see how this
• 1:03
  stuff might actually affect your everyday life.
• 1:04
  Exactly. Let's get started with the first big
• 1:07
  one. It's about Adobe. Seems like they're making
• 1:09
  a pretty significant move in the AI space. specifically
• 1:12
  on mobile. Right. They've just launched the Adobe
• 1:15
  Firefly mobile app. It's out now for both iOS
• 1:18
  and Android devices. Okay, Firefly, that's their
• 1:21
  generative AI suite, right? Yeah. The stuff like
• 1:23
  generative fill where you can type to edit images.
• 1:26
  Precisely. It includes tools like generative
• 1:28
  fill, generative expand, basically letting you
• 1:31
  add, remove, or replace things in photos, or
• 1:34
  even realistically extend the background all
• 1:36
  just using text prompts. Pretty powerful stuff.
• 1:39
  And now that's on your phone. Yep. right in your
• 1:41
  pocket. The app itself is a free download. Adobe's
• 1:45
  really pushing for convenience here, making those
• 1:47
  web features available in a mobile -friendly
• 1:50
  way. So you can tinker with projects or make
• 1:53
  quick edits wherever you are. Exactly. And importantly,
• 1:56
  anything you start on the mobile app syncs up
• 1:58
  with your Adobe Creative Cloud account. Oh, okay.
• 2:01
  So you could like start an idea on the train
• 2:04
  and then finish it up later on your desktop.
• 2:06
  That's the idea. It makes Firefly feel much more
• 2:08
  integrated with their other mobile tools like
• 2:10
  Lightroom or Adobe Express. You get that continuity.
• 2:13
  Makes sense. And it has those key features like
• 2:15
  generative fill, generative expand. Those are
• 2:18
  definitely highlights. Plus, you can also generate
• 2:21
  entirely new images or even short video clips
• 2:24
  just from text prompts using their latest Firefly
• 2:27
  AI models. Cookie, but here's where it gets really
• 2:29
  interesting based on the report. Adobe isn't
• 2:32
  just using its own AI models in Firefly anymore.
• 2:35
  No, and this is a big shift. They're still using
• 2:36
  their own models like Image Model 4 and Ultra,
• 2:39
  but they're opening up Firefly to host models
• 2:42
  from other AI companies too. Oh, wow. Like who?
• 2:45
  Well, some major players. They've partnered with
• 2:48
  OpenAI, Google bringing in Image in 3 and 4 for
• 2:50
  images and VO2 and 3 for video. Also Black Forest
• 2:54
  Labs, Ideogram, Luma AI, Pica, Runway. It's quite
• 2:59
  a list. There's a lot of different AI engines
• 3:00
  under one roof. What's the benefit for the user?
• 3:03
  Flexibility, Manly, you get this huge ecosystem
• 3:06
  of different AI models to choose from, all within
• 3:08
  the same Firefly interface. Maybe one model is
• 3:10
  better for photorealism, another for artistic
• 3:12
  styles, another for video. So you can pick the
• 3:14
  right tool for the job. Exactly. Adobe's VP of
• 3:18
  Generative AI, Alexandra Costin, was quoted saying
• 3:21
  they want it to be the ultimate one -stop shop
• 3:23
  for creative experimentation, his words, letting
• 3:26
  you explore different models, styles, media types,
• 3:30
  all in one place. That's actually pretty smart,
• 3:32
  giving creators more options without having to
• 3:34
  jump between different apps or platforms. Right.
• 3:37
  And Costin emphasized that adding more partner
• 3:39
  models gives creators even more flexibility to
• 3:42
  experiment, iterate, and push their ideas further.
• 3:46
  And they're still including those content credentials,
• 3:48
  the metadata saying AI was used. Yep, that's
• 3:51
  crucial. Regardless of which model you use Adobe's
• 3:54
  or a partner's war, the output will include that
• 3:57
  metadata for transparency. Good to know. What
• 3:59
  about cost? Is the app free? The download is
• 4:01
  free, yeah, but using the AI features burns through
• 4:04
  what they call generative credits. They have
• 4:06
  different membership tiers, free, standard, pro,
• 4:09
  premium, that give you a certain number of credits
• 4:11
  each month. So you can try it out for free, but
• 4:14
  heavy use will require a subscription. Pretty
• 4:16
  much. The free plan lets you test the waters.
• 4:18
  Gotcha. The report also mentioned something called
• 4:21
  Firefly Boards. What's that about? Ah, yes. That's
• 4:24
  a newer platform more focused on collaborative
• 4:27
  brainstorming. Think of mood boards or story
• 4:30
  boards, but AI powered. OK. And now they've added
• 4:33
  video generation to boards. You can apparently
• 4:36
  remix video clips or generate new ones using
• 4:39
  models from Adobe, Google, Luma AI or Pica right
• 4:43
  within your board. Interesting for ideation.
• 4:45
  Yeah, and you can also do AI edits on images
• 4:48
  in the boards using text prompts, tapping into
• 4:50
  models from Fluxpot One context or OpenAI. Kostin
• 4:54
  called it like an AI -powered creative jam session.
• 4:57
  A creative jam session, I like that. Okay, so
• 4:59
  let's connect the dots here. What does this Adobe
• 5:01
  News mean for you, the listener? Well, fundamentally
• 5:04
  it means incredibly powerful AI creative tools
• 5:08
  are now super accessible right on your phone.
• 5:10
  Like imagine needing to resize a photo for Instagram
• 5:13
  quickly. Generative expand could probably do
• 5:15
  that in seconds or removing that annoying sign
• 5:17
  from the background of a vacation pic while you're
• 5:19
  still, you know, on vacation. Exactly. We're
• 5:23
  just playing around with AI image generation
• 5:24
  when you have an idea wherever you are. It really
• 5:27
  simplifies and speeds up creating visuals, whether
• 5:29
  it's for personal stuff or even small business
• 5:31
  marketing. It does feel like another step towards
• 5:34
  making complex tech really easy to use, almost
• 5:37
  like that Star Trek replicator idea, but for
• 5:39
  images and videos on your phone. It's getting
• 5:41
  closer. OK, let's shift gears to our second story.
• 5:45
  This one's about how we watch TV. Ah, yes. The
• 5:48
  straining takeover. The report highlights a pretty
• 5:51
  major milestone from Nielsen data. It is a big
• 5:54
  one. For the very first time in May 2025, Americans
• 5:58
  actually spent more time watching streaming platforms
• 6:01
  than they did watching traditional cable and
• 6:03
  broadcast TV combined. Combined. Wow. So streaming
• 6:06
  didn't just be cable. It beat cable plus channels
• 6:09
  like NBC, CBS, ABC. That's right. Aggregated.
• 6:13
  We knew this was coming, you know, with everyone
• 6:14
  cutting the cord, but this is the official confirmation
• 6:17
  that streaming is now the dominant way people
• 6:19
  watch TV. It's quite a shift. I remember seeing
• 6:21
  reports maybe a couple of years ago where streaming
• 6:24
  passed broadcasts, but not both together. Exactly.
• 6:28
  The report mentioned July 2022 data for context.
• 6:31
  Back then, streaming had about 34 .8 % share,
• 6:35
  just edging out cable at 34 .4 % while broadcast
• 6:38
  was down at 21 .6%. So in less than three years.
• 6:42
  It's gone from just beating cable to beating
• 6:44
  both combined. That's fast. It really is. And
• 6:46
  the volume is huge. Even back in July 2022, folks
• 6:49
  were watching, what was it, nearly 191 billion
• 6:52
  minutes of streamed content every week. Incredible.
• 6:55
  Now, here's a part of the report I found really
• 6:56
  interesting. A big driver of this recent surge
• 6:59
  is older viewers. Yeah, the 65 plus demographic.
• 7:02
  They are apparently the fastest growing age group
• 7:04
  for streaming adoption right now, especially
• 7:06
  free services. Really free services. Their time
• 7:09
  spent watching YouTube, specifically on a TV
• 7:11
  screen, jumped 106 % between May 2023 and May
• 7:15
  2025. It's now apparently equal to the time kids
• 7:18
  under 11 spend watching YouTube on TV. Wow. Okay,
• 7:21
  that explains why those free ad -supported streaming
• 7:23
  services, fast channels they call them, are doing
• 7:25
  so well. Totally. Services like Tubi, Roku Channel,
• 7:29
  Pluto TV. The report said those three combined
• 7:33
  accounted for 5 .7 % of all TV time in May 2025.
• 7:38
  That's more than Disney Plus and Hulu put together.
• 7:41
  More than Disney Plus and Hulu combined. Yep.
• 7:44
  And Nielsen's Brian Fuhrer had a quote about
• 7:47
  why older viewers might like these free platforms.
• 7:50
  He suggested the ads aren't really jarring for
• 7:52
  them because, well... They grew up with ads on
• 7:54
  traditional TV. That makes sense. It's a familiar
• 7:56
  experience. Right. He even mentioned that a classic
• 7:58
  show like Gunsmoke regularly pops up on streaming
• 8:01
  tarts now because that audience is finding it
• 8:03
  on these free platforms. Meanwhile, traditional
• 8:05
  cable continues to slide down 39 percent overall
• 8:09
  in the last four years, the report said. And
• 8:11
  you have to figure the media companies themselves
• 8:13
  kind of push this along, right? By putting all
• 8:15
  their best new shows and movies onto their own
• 8:17
  streaming platforms instead of their cable channels.
• 8:19
  Good point. They sort of cannibalized their own
• 8:22
  cable business to boost streaming. Looking back
• 8:25
  again at those July 2022 numbers, Netflix and
• 8:28
  YouTube were leading then. Yeah, Netflix had
• 8:31
  8 % share, YouTube 7 .3%, then Hulu, Prime Video,
• 8:35
  mostly driven by big original hits like Stranger
• 8:38
  Things back then. But the report also notes that
• 8:40
  while streaming is dominant now, the super fast
• 8:43
  growth might be slowing down a bit. It mentions
• 8:45
  things like Netflix losing subscribers back in
• 8:48
  2022, changes at HBO Max, Disney lowering forecasts.
• 8:53
  Right. The market is definitely maturing. It's
• 8:55
  not the Wild West it was a few years ago. There's
• 8:57
  consolidation, strategy shifts. It's evolving.
• 9:01
  So connecting this one to the listener, what's
• 9:03
  the takeaway for you? Well, your own TV habits
• 9:06
  are clearly part of this huge trend. Whether
• 9:09
  you've cut the cord entirely, or you juggle multiple
• 9:11
  subscriptions, or maybe you lean on those free
• 9:13
  services, you're living this shift. And it impacts
• 9:17
  what gets made, right? More shows designed for
• 9:19
  binge watching, different kinds of stories maybe,
• 9:22
  and it definitely highlights the appeal of free
• 9:24
  options as those subscription bills stack up.
• 9:27
  Absolutely. You've got more choice than ever,
• 9:28
  which is great, but navigating it all can be
• 9:31
  a bit overwhelming sometimes. Bottom line is
• 9:33
  streaming isn't just part of the TV landscape
• 9:35
  anymore. It is the landscape. OK, on to our third
• 9:38
  story. This one takes us into the nitty gritty
• 9:41
  of AI models themselves. A new player from China.
• 9:46
  Yeah, a startup called Minimax. They've released
• 9:48
  a new large language model called Minimax M1,
• 9:51
  and it's causing a bit of a stir. Why is that?
• 9:54
  What's special about M1? A few things. First,
• 9:57
  it's designed specifically for complex productivity
• 10:00
  tasks. Second, and this is key, it's open source.
• 10:03
  Open source meaning? Meaning they've released
• 10:05
  the model's code and weights essentially under
• 10:07
  a permissive license, Apache 2 .0 in this case.
• 10:11
  So developers and companies anywhere can download
• 10:13
  it, use it, modify it. build products on top
• 10:16
  of it, even commercially, without paying MiniMax.
• 10:18
  It's out there on places like Hugging Face and
• 10:21
  GitHub. OK, that's significant. What else stands
• 10:23
  out? It's context window. The report says M1
• 10:25
  supports a 1 million token input context window.
• 10:29
  1 million tokens. Can you remind us what that
• 10:31
  means in simple terms? Sure. Tokens are like
• 10:33
  the basic units of text, the AI processes, words,
• 10:36
  parts of words, punctuation. The context window
• 10:39
  is how much information the AI can hold in its
• 10:42
  head at one time to understand the request and
• 10:44
  generate a response. So a bigger window means
• 10:46
  it can handle much longer documents or conversations.
• 10:50
  Exactly. One million tokens is massive. For comparison,
• 10:55
  OpenAI's powerful GPT -4 .0 model handles 128
• 10:58
  ,000 tokens. A million tokens. That's like feeding
• 11:02
  in a whole collection of books worth of information
• 11:04
  at once. It's huge for tasks like summarizing
• 11:07
  very long reports or maintaining context in a
• 11:09
  really extended interaction. Okay, wow. But here's
• 11:12
  the kicker, according to the report. The cost
• 11:15
  to train this supposedly powerful model. Yeah,
• 11:17
  this is the really eye -popping part. Minimax
• 11:19
  claims they trained M1 for only around $534 ,700.
• 11:24
  $534 ,000? That sounds incredibly low. How does
• 11:26
  that compare? Well, estimates for other big models
• 11:28
  are way higher. DeepSeek R1 was reportedly $5
• 11:31
  to $6 million. And models like GPT -4 are estimated
• 11:34
  to cost well over $100 million to train. So yeah,
• 11:37
  half a million is astonishingly cheap if the
• 11:39
  performance holds up. How do they manage that?
• 11:41
  They say it's down to a new, super -efficient
• 11:43
  training technique they developed called the
• 11:46
  CISPO algorithm, plus using a type of model architecture
• 11:50
  called a mixture of experts or MOE. MOE. That's
• 11:54
  where it uses smaller specialized parts instead
• 11:56
  of one giant network. Basically, yeah. It can
• 11:58
  make training and running the model much more
• 12:00
  efficient, especially for large models. They
• 12:02
  also claim M1 is efficient to run, using only
• 12:05
  about 25 % of the computing power of DeepSeq
• 12:08
  R1 for long outputs. And are they claiming it
• 12:10
  performs well? They are. They state M1 outperforms
• 12:14
  another model, DeepSeq R10528, and also all comparable
• 12:17
  Chinese close source models on complex productivity
• 12:20
  tasks. They've put out benchmark scores showing
• 12:23
  strong results in reasoning, coding, and handling
• 12:26
  that long context information. And it has other
• 12:29
  features too, like using tools. Yes, it has function
• 12:32
  calling abilities, so it can use external tools
• 12:34
  or APIs, and the package includes a chatbot API
• 12:37
  with built -in web search, image and video generation,
• 12:39
  even text -to -speech and voice cloning capabilities.
• 12:42
  Okay, so this M1 model seems like a big deal
• 12:45
  technically. But again, why should the average
• 12:48
  person you listening right now, care about some
• 12:51
  open source AI model from China? Because models
• 12:54
  like this, powerful, efficient, and open, are
• 12:57
  the engines that will power the AI tools you
• 12:59
  actually use in the future. So better AI assistance
• 13:02
  on your phone, smarter features in your productivity
• 13:05
  apps, that kind of thing. Potentially, yes. Because
• 13:08
  it's open source and claims to be cheap to run,
• 13:11
  companies might be able to integrate more advanced
• 13:13
  AI capabilities into their products more easily
• 13:16
  and affordably. Imagine an AI that could reliably
• 13:19
  read and understand that massive PDF report your
• 13:22
  boss sent you, or help you analyze a huge spreadsheet
• 13:25
  just by talking to it. That long context window
• 13:28
  makes things like that more feasible. Right,
• 13:30
  and this release just highlights how fast the
• 13:33
  AI field is moving and that innovation is coming
• 13:35
  from everywhere globally. Cheaper, more accessible,
• 13:38
  powerful AI could ultimately lead to better AI
• 13:41
  experiences for everyone. And for businesses,
• 13:43
  the report touches on lower costs, handling big
• 13:46
  data, customizing Lots of potential benefits
• 13:48
  there, too. Definitely. It raises interesting
• 13:50
  questions about competition and who leads AI
• 13:53
  development moving forward. Okay, let's quickly
• 13:55
  recap the three big stories we connect to the
• 13:57
  dots on today. First, Adobe is putting serious
• 14:00
  AI creative power, including partner models,
• 14:02
  onto your phone with the Firefly app, making
• 14:04
  content creation easier and more accessible on
• 14:07
  the go. Second, it's official. Streaming is now
• 14:10
  the number one way Americans watch TV, beating
• 14:13
  cable and broadcast combined with older viewers
• 14:15
  and free ad -supported services playing a surprisingly
• 14:18
  large role in that shift. And third, Chinese
• 14:22
  startup Minimax dropped a potentially groundbreaking
• 14:25
  open source AI model, M1, that boasts a massive
• 14:29
  context window and claims incredibly low training
• 14:31
  costs, signaling accelerating efficiency and
• 14:34
  global competition in AI development. So here's
• 14:36
  a thought to chew on as we wrap up. What happens
• 14:39
  when all these trends really start to merge?
• 14:41
  Yeah, think about it. You've got AI models getting
• 14:43
  cheaper, more powerful, more open. capable of
• 14:46
  understanding vast amounts of info. At the same
• 14:49
  time, sophisticated AI creative tools are landing
• 14:52
  directly in everyone's pocket via their phones.
• 14:54
  And all this is happening while our media consumption
• 14:57
  shifts almost entirely online, dominated by streaming.
• 15:01
  How might AI start to shape not just the content
• 15:04
  we create, but also how we experience entertainment,
• 15:07
  how we find information, maybe even how our brains
• 15:10
  process this increasingly digital AI -infused
• 15:14
  world? Definitely something to ponder as technology
• 15:16
  keeps connecting those dots for us. That's all
• 15:19
  the time we have for today. We are here to help
• 15:20
  here at Snarful, so reach out on our website
• 15:22
  if you have any more questions. That's a wrap
• 15:28
  on today's top stories. Remember, tech doesn't
• 15:30
  have to be complicated and we're here to help
• 15:33
  connect the dots. If you've got questions or
• 15:36
  want to learn more, visit us at snarful .com.
• 15:39
  Thanks for listening and we'll catch you next
• 15:41
  time.