Latest / Today’s AI News / Sam Altman’s Home Attacked, AI Uncovers Hidden Ozempic Side Effects, Anti-AI Rage Goes Mainstream

Transcript

Sam Altman’s Home Attacked, AI Uncovers Hidden Ozempic Side Effects, Anti-AI Rage Goes Mainstream · Today’s AI News

• 0:00
  Building advanced AI right now is like wiring a house for
• 0:03
  electricity while the walls are still on fire.
• 0:06
  We are bringing this incredible life changing power into our
• 0:08
  living rooms, but the sparks are clearly starting to catch and
• 0:12
  some people are aggressively reaching for the fire
• 0:14
  extinguisher. Oh, absolutely.
• 0:16
  Welcome back to The Deep Dive, covering today's AI news,
• 0:19
  brought to you by 9X Productions.
• 0:21
  Today, we're looking at the friction created by that exact
• 0:24
  paradox. We're going to unpack the
• 0:26
  intense physical backlash hitting AI leaders in the real
• 0:30
  world right now. Yeah, before pivoting to the
• 0:33
  flip side of that same technological coin, because the
• 0:36
  contrast is just staggering. It really is.
• 0:38
  At the very same moment you have individuals taking extreme
• 0:41
  measures to protest the proliferation of this tech,
• 0:44
  these exact same large language models are quietly pulling off
• 0:48
  medical miracles that human doctors and even massive
• 0:51
  pharmaceutical trials have entirely missed.
• 0:54
  It completely flips the narrative.
• 0:56
  So we are going to cover those major developments, breakdown
• 0:59
  the mechanisms behind today's massive industry shifts, and
• 1:02
  wrap it all up with a community AI workflow that genuinely might
• 1:06
  change how you approach your own healthcare.
• 1:08
  Sounds good. Let's start with the sparks
• 1:10
  Catching Fire. Because the cultural temperature
• 1:11
  around AI just went from a low simmer to a rolling boil around
• 1:15
  3.45 in the morning. Recently, a 20 year old suspect
• 1:19
  named Daniel Moreno Gama threw a Molotov cocktail at the front
• 1:23
  gate of open AICEO Sam Altman's home.
• 1:26
  Right. And to set the baseline here, no
• 1:28
  one was injured in this incident.
• 1:30
  The San Francisco Police Department apprehended the
• 1:32
  suspect about an hour later, curiously enough, over near the
• 1:36
  open AI headquarters. But to understand the gravity of
• 1:39
  this, we really need to look at the ideological framework
• 1:41
  driving the action. Right, his background.
• 1:43
  Exactly. Moreno Gamma was highly active
• 1:46
  on a Discord server for an organization called Pause AI,
• 1:50
  and he operated under the alias Butlerian Jihadist.
• 1:54
  Wow. OK, for the sci-fi fans
• 1:57
  listening, you probably recognize that immediately.
• 1:59
  The Butlerian Jihad is a cornerstone of the Dune
• 2:02
  universe. Yeah, it was this massive,
• 2:05
  violent human crusade against computers, thinking machines,
• 2:09
  and conscious robots. It essentially outlawed the
• 2:13
  creation of machines in the image of a human mind.
• 2:16
  So choosing that alias tells you exactly the philosophical
• 2:19
  headspace he was in. I mean, that ideology was
• 2:22
  clearly a guiding principle for him.
• 2:24
  He had published several essays arguing that artificial
• 2:27
  intelligence is on a guaranteed trajectory toward human
• 2:29
  extinction. Right On that Pause AI server,
• 2:32
  he posted 34 messages. And while the Pazza AI
• 2:35
  organization itself swiftly and entirely condemned the attack,
• 2:38
  the digital footprint is pretty revealing.
• 2:40
  Oh, for sure. One of his messages had
• 2:42
  previously been flagged by the community moderators because it
• 2:44
  appeared to be inciting direct, real world action.
• 2:48
  But looking past the attack itself, Sam Altman's response is
• 2:51
  what really elevates this from a local Crime Story to a global
• 2:56
  tech moment. Right.
• 2:56
  It was unusual. Very usually when a CEO faces a
• 3:00
  security threat, you get a highly sanitized PR statement.
• 3:03
  Just a blanket condemnation of violence may be a thank you to
• 3:07
  law enforcement, and that's it. Altman didn't do that.
• 3:09
  No, he didn't. He published A deeply personal
• 3:12
  essay. He admitted to past leadership
• 3:14
  mistakes. He outright stated that the
• 3:17
  public's anxiety regarding AI is, quote, justified.
• 3:21
  He even went so far as to compare the current corporate
• 3:24
  power struggle over artificial general intelligence to a ring
• 3:27
  of power from Lord of the Rings. Which is wild, and the timing of
• 3:30
  that essay provides crucial context.
• 3:33
  Just days before this incident, Open AI released a sobering 13
• 3:37
  page policy document. Yeah, I read that.
• 3:40
  The core thesis wasn't promotional, it was a stark
• 3:42
  warning. They outlined that AI has the
• 3:45
  potential to reshape society at a velocity that nobody, not the
• 3:48
  federal government, not the private sector, and certainly
• 3:51
  not the general public is currently prepared for.
• 3:53
  Right. So the anxiety we're seeing
• 3:54
  manifest in the physical world isn't just external paranoia,
• 3:58
  it's being validated by the very people building the systems.
• 4:01
  OK, wait, hold on. This is the paradox that I keep
• 4:03
  getting stuck on. Sam Altman has essentially
• 4:06
  positioned himself as the lightning rod for global tech
• 4:10
  anxiety. Right, but how can you seriously
• 4:13
  call for societal de escalation while simultaneously flooring
• 4:18
  the gas pedal on the exact technology you openly admit is
• 4:21
  going to upend human civilization?
• 4:24
  Aren't those two actions fundamentally incompatible?
• 4:28
  It's a tough spot. It feels a bit like pouring
• 4:30
  gasoline on a bonfire and then asking the flames to exercise a
• 4:33
  little restraint. That fundamental tension is
• 4:36
  exactly what is fracturing the cultural discourse right now.
• 4:38
  If we look at the broader polling data, we have to
• 4:41
  recognize that this is no longer a fringe sci-fi thought
• 4:44
  experiment. No, definitely not.
• 4:45
  Societal anger and fear regarding AI are becoming
• 4:48
  decidedly mainstream. Current surveys indicate that
• 4:51
  four out of five Americans harbor significant worries about
• 4:55
  the trajectory of artificial intelligence.
• 4:57
  Four out of five. Wow.
• 4:58
  And the volatility is sustained. Just days after the Molotov
• 5:02
  incident, there was reportedly a second attack involving gunshots
• 5:06
  fired outside Altman's residence.
• 5:08
  The ideological debate over existential risk is just
• 5:11
  spilling over into physical reality.
• 5:13
  It is genuinely unnerving to watch the radical debates turn
• 5:16
  into police. Reports The escalation is real,
• 5:19
  but to grasp the full scope of what's happening right now, we
• 5:23
  have to examine the whole picture.
• 5:25
  The profound paradox of this moment is that while the fear of
• 5:28
  a machine driven apocalypse is driving people to extreme
• 5:31
  dangerous actions, the exact same underlying architecture,
• 5:35
  these massive data synthesizers are simultaneously saving lives.
• 5:40
  They're catching critical real world details that are most
• 5:43
  rigorous human institutions are missing entirely.
• 5:46
  That is the perfect pivot. I want to look at a massive new
• 5:49
  study that just came out of the University of Pennsylvania
• 5:51
  because it completely flips the scientific method on its head.
• 5:54
  It really does. Researchers used a technique
• 5:56
  called computational social listening to investigate the
• 5:59
  wildly popular weight loss and diabetes drugs Ozempic and
• 6:03
  Monjaro, and they discovered widespread side effects that the
• 6:06
  official FDA approved clinical trials completely.
• 6:09
  Missed. What's fascinating here is the
• 6:10
  methodology. The researchers didn't run
• 6:12
  another traditional double-blind clinical trial with a small
• 6:16
  group of highly monitored patients.
• 6:18
  Instead, they went directly to the messy, unfiltered reality of
• 6:23
  patient experiences on the Internet.
• 6:25
  They took over 400,000 Reddit posts generated by roughly
• 6:29
  67,000 different users over a span of five years.
• 6:33
  Then they fed that massive data set directly into large language
• 6:37
  models. Specifically GPT and Gemini,
• 6:40
  right? Exactly.
• 6:41
  But wait, if you've ever spent 5 minutes on Reddit, you know it
• 6:43
  is not a pristine database of medical terminology.
• 6:46
  Oh, definitely. Not it is full of slang,
• 6:48
  hyperbole and sarcasm. How is an AI making scientific
• 6:52
  sense out of someone posting? I literally cannot peel myself
• 6:55
  off the mattress today. That is where the specific
• 6:57
  semantic power of an LLM comes into play.
• 7:00
  It isn't just running a keyword search for the word tired.
• 7:04
  The AI acts as a sophisticated translation engine.
• 7:08
  It was tasked with mapping that informal Internet chatter, the
• 7:11
  slang, the venting, the casual complaints, directly into
• 7:15
  standardized clinical medical terms.
• 7:17
  It understands the context of a sentence well enough to
• 7:20
  categorize peeling myself off the mattress as clinical
• 7:23
  fatigue. And the findings from that
• 7:25
  translation are incredible. Nearly half of the entire sample
• 7:28
  size reported experiencing at least one side effect that is
• 7:31
  nowhere to be found on the current drug labels.
• 7:33
  Yeah, almost half. We are talking about severe
• 7:36
  chills, menstrual irregularities and hot flashes, but the one
• 7:40
  that really stood out was fatigue.
• 7:42
  Fatigue was the second most common complaint among all
• 7:45
  67,000 users, yet if you look at the official clinical trial
• 7:49
  reporting that got the drug approved, fatigue barely even
• 7:52
  registered. But let me push back on this a
• 7:54
  bit. If the AI is just listening to
• 7:56
  the neighborhood Grapevine of Reddit, as the researchers
• 7:59
  called it, how do we know that AI isn't just validating mass
• 8:02
  Internet paranoia? That's a fair question.
• 8:04
  You know how social media works. One person says the drug gave
• 8:08
  them a headache, the algorithm boosts it and suddenly 1000
• 8:11
  people convince themselves they have a migraine.
• 8:15
  Is this actual science, or is it just a very sophisticated rumor
• 8:18
  aggregator? That distinction is exactly what
• 8:21
  makes this study so important. The LLM isn't replacing peer
• 8:24
  reviewed science, it's acting as a directional beacon.
• 8:28
  It's like panning for gold in a muddy river.
• 8:31
  I like that analogy. Humans just see the mud, you
• 8:33
  know, the typos, the daily complaining, the viral panic.
• 8:37
  But the LLM is a microscopic sieve that only catches the
• 8:41
  specific molecular weight of real clinical symptoms.
• 8:45
  When you have 10s of thousands of diverse geographically
• 8:49
  separated people flagging the exact same physiological
• 8:51
  symptoms independently over a five year period, the
• 8:55
  statistical weight rules out a localized Internet rumor.
• 8:58
  That makes a lot of sense, but why did the traditional clinical
• 9:00
  trials miss something as massive as chronic fatigue in the first
• 9:03
  place? It comes down to the structure
• 9:05
  of how we approve drugs. Clinical trials are looking for
• 9:09
  what the industry calls primary endpoints.
• 9:12
  Basically, the FDA requires them to check for severe life or
• 9:16
  death red flags. Like does the drug cause liver
• 9:19
  failure? Exactly.
• 9:20
  Does it trigger heart attacks? The trials are relatively short
• 9:24
  and the patient pool is highly controlled.
• 9:26
  Because they're hyper focused on those primary dangerous
• 9:30
  endpoints, they frequently miss subjective quality of life
• 9:34
  experiences like daily fatigue. Which is huge for the patient.
• 9:38
  Right. And this matters immensely right
• 9:40
  now because AI is accelerating the drug discovery pipeline.
• 9:44
  Novel treatments are hitting the market faster than ever, which
• 9:46
  means post market surveillance, catching these side effects in
• 9:49
  the real world has to become faster too.
• 9:51
  That ability to synthesize massive, messy data and pull
• 9:56
  actionable truth out of it isn't just revolutionizing medicine,
• 9:59
  it's the exact same underlying logic driving the massive
• 10:02
  software updates that dropped across the tech sector this
• 10:05
  morning IT. Really is everywhere.
• 10:06
  Let's look at the rapid fire updates and unpack the
• 10:08
  mechanisms behind what's moving the needle today.
• 10:11
  The sheer volume of major releases today is notable
• 10:14
  specifically because they are moving away from simple chat
• 10:17
  bots and toward autonomous systems.
• 10:19
  Let's start with cursor 3. For the developers listening,
• 10:22
  you know that the real bottleneck in coding isn't
• 10:24
  typing the syntax, it's context switching.
• 10:28
  Cursor 3 just dropped a new Agent First interface designed
• 10:31
  specifically for parallel coding agents.
• 10:34
  Right, and to explain the mechanism there, we have to move
• 10:36
  past the idea of AI as an advanced autocorrect.
• 10:40
  Cursor 3 isn't just fixing a broken line of code while you
• 10:43
  type, it deploys autonomous workers in the background.
• 10:47
  So while you are manually adjusting the user interface on
• 10:49
  your screen, one AI agent is autonomously rewriting your back
• 10:53
  end database logic and a second agent is simultaneously updating
• 10:57
  your API documentation. It is true parallel processing
• 11:01
  for human workflow. It's basically giving every
• 11:04
  developer a team of junior engineers, and that same leap
• 11:07
  and complexity is happening over at Meta with the release of Muse
• 11:10
  Spark. Yeah, their new model.
• 11:12
  This is their new multimodal reasoning AI.
• 11:15
  By multimodal it means it doesn't just read text, it
• 11:19
  natively understands audio, video and code simultaneously,
• 11:22
  and it also features a multi agent mode.
• 11:25
  Meanwhile, Hagen just released Avatar V for generating studio
• 11:29
  quality videos from text prompts.
• 11:31
  We also saw the release of Hermes Agent, which introduces
• 11:34
  persistent memory across platforms.
• 11:36
  Which is a game changer. It is.
• 11:38
  That means the AI remembers a conversation you had on your
• 11:41
  laptop on Tuesday and applies that context to a totally
• 11:45
  different task you ask it to do on your phone on Friday.
• 11:47
  While the public is getting their hands on those tools, the
• 11:50
  industry shake UPS happening behind closed doors are just as
• 11:53
  aggressive. Let's talk about the model wars,
• 11:54
  right? A major leak just confirmed that
• 11:57
  Alibaba is the developer behind a mysterious AI model named
• 12:00
  Happy Horse. This unreleased video model
• 12:03
  suddenly appeared on the global leaderboards and took the
• 12:05
  absolute top spot, officially knocking bike dances sedence 2
• 12:09
  point O down to 2nd place the. Geopolitical implications of
• 12:13
  Chinese tech giants battling for the top of the global video
• 12:16
  generation leaderboard is a whole separate deep dive.
• 12:19
  Oh for sure. But to build those models you
• 12:20
  need people, and the talent wars are arguably more ruthless than
• 12:25
  the model wars right now. Meta is aggressively poaching
• 12:29
  top tier executives who recently departed open AI Stargate
• 12:33
  project. Yeah, I saw.
• 12:34
  That they just secured Peter Hoshel, Shomesh Imani and Anu
• 12:37
  Saharan to build out their newly formed Metacompute group.
• 12:41
  Why the intense focus on the compute group specifically?
• 12:45
  Because the bottleneck in AI development is no longer just
• 12:47
  writing clever software algorithms.
• 12:49
  The true bottleneck is physical infrastructure, networking 10s
• 12:53
  of thousands of GPU's together without melting the data center.
• 12:57
  Meta is recognizing that the future of AI requires hardware
• 13:00
  and physics expertise just as much as software engineering.
• 13:03
  OK, but beyond the corporate chess moves and the hardware
• 13:06
  bottlenecks, we have to talk about the sheer duality of how
• 13:10
  this technology is being perceived at the institutional
• 13:12
  level, because two meetings happened recently that perfectly
• 13:15
  capture the absolute whiplash of the current AI zeitgeist.
• 13:19
  Yeah, the contrast between these two summits is a perfect
• 13:21
  reflection of our broader societal confusion.
• 13:24
  On one hand, Anthropic recently hosted a highly unusual summit
• 13:28
  at their headquarters. The entire focus was on the
• 13:31
  moral development of their primary AI model, Claude.
• 13:35
  Moral Developer. Yes, they brought in Christian
• 13:38
  theologians and leaders to debate incredibly profound
• 13:41
  philosophical questions. They spent time discussing
• 13:44
  whether an advanced artificial intelligence can experience
• 13:47
  grief or if an AI could theoretically be considered a
• 13:49
  child of God. Which is incredibly heavy
• 13:53
  existential territory for a tech company to wade into.
• 13:55
  It is. But then you look at Washington,
• 13:57
  DC. That very same week, U.S.
• 13:59
  Treasury Secretary Besant and Federal Reserve Chair Powell
• 14:03
  summoned Wall Street CEO's for an urgent high level briefing.
• 14:07
  The topic of that meeting? The severe systemic cyber risks
• 14:10
  and potential financial destabilization posed by
• 14:13
  Anthropics new mythos model. It just perfectly mirrors the
• 14:16
  tension we saw earlier with the Molotov attack.
• 14:19
  You have the exact same company holding a summit treating their
• 14:23
  AI like a developing spiritual entity while the federal
• 14:26
  government is treating it like a weapon of mass financial
• 14:29
  destruction. We literally don't know whether
• 14:31
  to baptize these things or ban them.
• 14:34
  It highlights the macro level paralysis the CEOs, religious
• 14:38
  scholars and government officials are locked in these
• 14:40
  massive debates about the existential nature of the
• 14:43
  technology. But if we bring it back down to
• 14:46
  earth, everyday people are not waiting for Washington or
• 14:49
  Silicon Valley to resolve those philosophical questions.
• 14:52
  They're not. They're already using these
• 14:53
  reasoning engines to bypass broken bureaucratic systems in
• 14:58
  their own daily lives. Which brings us to an incredible
• 15:01
  community AI workflow. This is all about taking your
• 15:05
  health back into your own hands. This story comes from a listener
• 15:08
  named LW up in Ontario, Canada. If you've ever felt completely
• 15:13
  abandoned by the modern healthcare system, you are going
• 15:15
  to relate to this. It's profound demonstration of
• 15:18
  the synthesis capabilities we discussed earlier with the Penn
• 15:20
  study. So Elle suffered from severe
• 15:23
  chronic lower back pain for two solid years, and she did
• 15:28
  everything you're supposed to do.
• 15:29
  She went to the general practitioners and they sent her
• 15:32
  to the specialist, but the specialist had no answers for
• 15:35
  her specific daily agony. She got the X-rays, she got the
• 15:39
  Mris, and the imaging showed osteoarthritis and degenerative
• 15:43
  discs, which sounds like a definitive answer, but as anyone
• 15:47
  who deals with chronic pain knows, those are incredibly
• 15:50
  common findings in adults. Yeah, very common.
• 15:52
  They often don't actually explain the root mechanical
• 15:55
  cause of the sharp pain you feel when you bend over to tie your
• 15:58
  shoes. The systemic issue Elle ran into
• 16:01
  is how specialized medicine has become.
• 16:04
  A radiologist reads the MRI and sees disc degeneration, but they
• 16:08
  don't watch how she walks. The physiotherapist watches her
• 16:11
  walk, but they don't have the holistic medical history or the
• 16:14
  detailed pharmacological background.
• 16:16
  The medical professionals are siloed in their specific fields,
• 16:20
  and patients with complex, overlapping mechanical issues
• 16:23
  often fall right through the cracks between those silos.
• 16:25
  Exactly what happened to her. Her physiotherapy failed.
• 16:28
  The medical injections failed and she was left with generic
• 16:32
  pain management advice but zero actionable solutions for
• 16:36
  recovery. So Elle decided to breakdown
• 16:38
  those silos herself. She gathered everything.
• 16:41
  She took her daily symptom logs, all of the dense radiologist
• 16:44
  reports from her imaging, the list of failed treatments, and
• 16:47
  her entire detailed medical history.
• 16:50
  And she fed all of that unstructured data directly into
• 16:53
  Gemini. But she didn't just type in Why
• 16:56
  does my back hurt? Together, she and the AI went
• 16:59
  back and forth in a deep diagnostic dialogue.
• 17:03
  They analyzed her specific movement patterns against her
• 17:05
  historical data, and they pinpointed the exact hidden
• 17:09
  mechanical triggers of her pain that the Silence specialists had
• 17:12
  missed. The mechanism of how she used
• 17:14
  the tool is the key take away here.
• 17:16
  If you use Google for this, you get millions of generic articles
• 17:19
  until you just stretch more. Right, WebMD just says you're
• 17:21
  dying. Exactly.
• 17:23
  Gemini, however, acted as an ongoing, highly personalized
• 17:26
  medical troubleshooter. Because of its massive context
• 17:29
  window, it was able to hold the entire holistic picture of her
• 17:32
  health in its memory. Simultaneously, it synthesized
• 17:36
  years of disparate medical data and found a unified mechanical
• 17:39
  pattern. Then it generated a highly
• 17:41
  targeted 4 month recovery, an exercise plan complete with
• 17:45
  weekly milestones specifically adapted to her daily progress.
• 17:49
  And the result speaks for itself.
• 17:51
  At 3 1/2 months into this AI generated hyper personalized
• 17:55
  physical therapy plan, L is nearly completely pain free.
• 17:59
  That's amazing. She literally told us that it
• 18:02
  gave her her life back. This is the ultimate aha moment
• 18:05
  of the week. All the headlines, the policy
• 18:07
  docs and the protests are focused on apocalyptic risks or
• 18:10
  billionaire corporate power struggles.
• 18:13
  But out here in the real world, AI is stepping in as an
• 18:15
  incredibly capable, tireless advocate, giving people their
• 18:19
  mobility and their lives back when traditional, highly funded
• 18:22
  systems fail them. It perfectly grounds the paradox
• 18:24
  we started with. Today, we're navigating a
• 18:26
  technology with unprecedented capabilities.
• 18:29
  But rather than just marveling at the technology, we should
• 18:32
  consider the imminent implications.
• 18:34
  We've seen AI find missing drug side effects from 400,000 casual
• 18:38
  Reddit posts by recognizing invisible statistical patterns,
• 18:42
  and we've seen it pinpoint the highly specific root cause of
• 18:46
  one woman's chronic back pain by synthesizing her siloed medical
• 18:50
  history. So the question of ponder is
• 18:53
  this. What happens in the very near
• 18:55
  future when these exact same large language models are
• 18:59
  directly plugged into our smartwatches, our health rings,
• 19:02
  and our biosensors 247? Oh wow.
• 19:05
  Will AI become an ever present digital immune system,
• 19:08
  continuously analyzing our biometric data in the
• 19:11
  background, capable of diagnosing us with complex
• 19:14
  illnesses days or even weeks before we feel a single physical
• 19:18
  symptom ourselves? That is a wild and honestly very
• 19:21
  hopeful thought to leave on. The fire might be catching in
• 19:24
  the walls, but the light it's casting on the things we
• 19:26
  couldn't see before is pretty incredible.
• 19:28
  Make sure you subscribe to stay updated on all these fast moving
• 19:31
  developments. If you enjoyed the show, please
• 19:33
  invite your friends and take a second to rate US five stars.
• 19:35
  We'll see you right back here for tomorrow's deep dive.