Latest / Elon Musk Podcast / Elon threatened OpenAI

Transcript

Elon threatened OpenAI · Elon Musk Podcast

• 0:00
  Elon Musk is demanding up to $150 billion from Sam Altman and
• 0:05
  Open AI, claiming they essentially stole a charity to
• 0:08
  build a massive commercial empire.
• 0:11
  I mean, that financial demand is just, well, it's staggering,
• 0:15
  especially when you factor in the whole accusation of stealing
• 0:18
  a charity. And if you look at the
• 0:19
  environment surrounding them right now, Microsoft holds this
• 0:24
  enormous financial stake. They provide all the underlying
• 0:27
  computing infrastructure while open AI is simultaneously
• 0:30
  shifting toward a public offering.
• 0:32
  Yeah. Which is a huge shift.
• 0:33
  Exactly. And it creates this intense
• 0:35
  friction between the company's altruistic origins and, you
• 0:39
  know, its current corporate structure, which really pushes
• 0:42
  this entire discussion directly toward the newly unsealed
• 0:46
  evidence and all these highly contested boardroom battles.
• 0:48
  So I guess the big question we have to figure out is how
• 0:50
  exactly does a nonprofit research lab transform into a
• 0:53
  commercial entity that requires $50 billion in computing power
• 0:57
  just to keep the lights on? Yeah, that is the core of it.
• 1:01
  And to really understand the mechanics of that
• 1:03
  transformation, we have to look closely at the unsealed private
• 1:08
  diary of Open AI President Greg Brockman.
• 1:11
  The diary entries, yeah. Right.
• 1:13
  So Brockman wrote in his journal that committing to the nonprofit
• 1:16
  structure would be, quote, a lie.
• 1:19
  Wow. Yeah, and he wrote that
• 1:21
  converting to a for profit model without Musk would be morally
• 1:24
  bankrupt. That's a pretty strong phrase.
• 1:26
  It really is. And at the exact same time, he
• 1:29
  also privately questioned what specific financial path would
• 1:34
  actually take him to $1 billion. Which makes sense why Musk's
• 1:38
  legal team frames this entire situation as a calculated bait
• 1:41
  and switch, right? Exactly.
• 1:42
  They're essentially alleging that Musk funded the project
• 1:44
  under the strict pretense of a public charity.
• 1:47
  While the founders were privately engineering a
• 1:50
  commercial flip the entire time. So they're saying the non profit
• 1:53
  status wasn't just a convenient tax designation.
• 1:56
  Right, they argue it was the actual purchase price for Musk's
• 1:59
  money and his public reputation and, you know, his ability to
• 2:04
  recruit top tier engineering talent away from huge monopolies
• 2:08
  like Google. Wait, back up.
• 2:11
  Are we sure this wasn't just him venting his frustrations?
• 2:14
  How do you mean? Well, I mean, it's entirely
• 2:18
  possible for a startup executive operating under extreme pressure
• 2:21
  to just, you know, write down private doubts in a journal.
• 2:25
  It doesn't necessarily mean it's some Grand Master plan for
• 2:28
  corporate sabotage. Sure, founders have moments of
• 2:31
  doubt constantly, right? But the issue here is the
• 2:33
  timing. These diary entries were written
• 2:35
  concurrently, with written assurances being sent directly
• 2:38
  to donors. Oh, I see.
• 2:40
  Yeah, a federal judge actually ruled that this specific overlap
• 2:43
  in timing is sufficient evidence of potential fraud to proceed to
• 2:46
  trial because, you know, you have to prove intent in a fraud
• 2:50
  case. Right, intent is everything.
• 2:51
  If someone changes their mind five years later, well, that's
• 2:53
  just business. But if they're writing private
• 2:55
  notes, calling a charity a lie at the exact moment they are
• 2:58
  cashing the donation checks? That goes directly to Intent.
• 3:01
  Exactly. It's it's like donating a
• 3:04
  fortune to build a Public Library, and the architects
• 3:07
  secretly plan to turn it into a private bookstore.
• 3:10
  But they're still using your bricks.
• 3:12
  That's a great way to put it. You would not hand over your
• 3:14
  money if you knew the final blueprint was a commercial
• 3:17
  bookstore. No, of course not.
• 3:19
  And that consequence completely changes the entire legal dynamic
• 3:23
  of this dispute. It limits the defense's ability
• 3:26
  to dismiss the lawsuit as just a simple rivalry between two
• 3:30
  billionaires. It opens up the possibility that
• 3:33
  early informal promises, you know, things that over dinner or
• 3:36
  in casual emails, actually carries strictly enforceable
• 3:39
  legal weight under charitable trust laws.
• 3:41
  Which is huge. Yeah, and the damages model
• 3:44
  presented by Musk's financial expert, see Paul Wazan, relies
• 3:47
  on this exact premise. Wazan calculates the financial
• 3:51
  damages not just on the initial seed money Musk provided, but on
• 3:55
  the enormous valuation the company holds today.
• 3:58
  Oh wow. Because, they argue, his early
• 4:00
  capital was the engine for that growth.
• 4:02
  Exactly. The plaintiff is arguing that
• 4:04
  without Musk's early capital and recruitment efforts, the company
• 4:08
  never gets off the ground. So if that foundation was built
• 4:11
  on a deceptive premise, the legal argument is that the
• 4:15
  resulting wealth should revert entirely back to the original
• 4:19
  charitable trust. Building on that complete
• 4:22
  breakdown of foundational trust, Brockman's testimony regarding a
• 4:27
  highly tense physical confrontation adds another
• 4:30
  critical layer to this. Oh, the confrontation, right?
• 4:32
  Yeah, During a major dispute over the future direction of the
• 4:35
  company, Musk demanded absolute control, majority equity and the
• 4:39
  chief executive role. That's quite a list of demands.
• 4:42
  It is, And when the other founders refused his demands,
• 4:45
  Musk became furious. Brockman actually testified
• 4:48
  Underoath that the tension in the room was so severe he
• 4:51
  thought Musk was actually going to physically hit him.
• 4:54
  Jeez. And following that
• 4:55
  confrontation, Musk allegedly stormed out of the building.
• 4:58
  He threatened to completely cut off his funding and he stated he
• 5:01
  would just go build his own artificial intelligence at Tesla
• 5:04
  without any of the safety constraints they were debating.
• 5:06
  Which is wild. Yeah, according to Brockman's
• 5:08
  testimony, Musk told the remaining employees that if the
• 5:12
  sheep are dictating safety and the wolves are not, then there's
• 5:15
  no purpose to the organization at all.
• 5:17
  Look, if you look at this from the outside, demanding absolute
• 5:20
  corporate authority and threatening to build an unsafe
• 5:23
  alternative at a car company, well, that directly contradicts
• 5:27
  the altruistic narrative. Do you think so?
• 5:28
  Yeah. I mean, it shows he just wanted
• 5:30
  total control the technology for himself.
• 5:33
  If his primary concern was truly the open and safe development of
• 5:36
  artificial intelligence for the benefit of humanity, throwing a
• 5:40
  tantrum and threatening to unleash an unrestrained version
• 5:43
  of the technology completely undermines his image as a
• 5:47
  selfless defender of charity. I see your point, but you know
• 5:51
  Musk's massive ego does not legally erase the diary entries
• 5:55
  or the fact that his initial funding was actively solicited
• 5:58
  under a nonprofit banner. That's true.
• 6:00
  The defense can certainly highlight his demands for total
• 6:02
  dominance to make him look bad to a jury, but the court legal
• 6:06
  question still remains whether the funds were accepted under
• 6:09
  false pretenses at the moment they were deposited right.
• 6:12
  His subsequent actions, however aggressive or egotistical they
• 6:15
  might be, do not retroactively validate a potentially deceptive
• 6:20
  fundraising strategy executed by the Founders.
• 6:22
  So this really alters the trials narrative.
• 6:25
  It opens up a strong defense strategy for Open AI to portray
• 6:28
  Musk as an opportunist seeking a hostile corporate takeover
• 6:32
  rather than, you know, a philanthropist protecting a
• 6:36
  sacred mission. Exactly.
• 6:37
  By highlighting his aggressive demands for equity and control,
• 6:41
  the defense limits his ability to claim he was purely an
• 6:44
  innocent victim of corporate greed.
• 6:46
  It forces the jury to weigh those notoriously messy
• 6:49
  realities of startup founder dynamics against the stripped,
• 6:53
  rigid definitions of charitable trust laws.
• 6:56
  Right. And if we look at the internal
• 6:57
  culture following Mesk's departure, we have the recorded
• 7:00
  testimony of former technology chief Mira Murati.
• 7:03
  Oh, this part is fascinating. It really is, Murati testified
• 7:06
  Underoath that Sam Altman actively misled her about the
• 7:09
  safety protocols for a new model.
• 7:10
  They just let. Her how?
• 7:12
  She stated Altman falsely claimed the legal department had
• 7:15
  determined a Safety Board review was completely unnecessary for
• 7:18
  this specific release. We.
• 7:20
  Really. Yeah, and she specifically
• 7:22
  consulted Jason Quan, the company's general counsel, and
• 7:25
  discovered a direct, alarming misalignment between what Altman
• 7:29
  told her and the legal team's actual.
• 7:31
  Position and Murati extended this into a much broader
• 7:35
  systemic critique of the organization, didn't she?
• 7:37
  She did. Yeah, she testified that Altman
• 7:40
  created complete chaos, that he actively pitted executives
• 7:44
  against one another and said completely opposite things to
• 7:47
  different people. She warned that this behavior
• 7:50
  left the entire company at a catastrophic risk of falling
• 7:52
  apart. I mean, you have a situation
• 7:54
  where the chief technology officer, right, the person
• 7:56
  actually responsible for the deployment of the code, feels
• 7:59
  she cannot trust the basic information coming from the
• 8:01
  chief executive regarding safety parameters.
• 8:04
  Right. For frontier technology, no
• 8:05
  less. Exactly.
• 8:07
  If you're using these tools everyday for your work or
• 8:09
  research, this directly impacts whether the product you rely on
• 8:12
  prioritizes strict safety rails or just raw, unvetted processing
• 8:16
  power. Yeah.
• 8:18
  And this touches on the fundamental tension within their
• 8:20
  highly unusual hybrid structure. You have a nonprofit board of
• 8:24
  directors that is legally mandated to oversee a capped
• 8:28
  profit subsidiary. The entire reason that structure
• 8:32
  exists is so the nonprofit board can pull the emergency brake if
• 8:36
  the technology gets too dangerous.
• 8:37
  Makes sense. But when the executive managing
• 8:39
  the for profit arm is allegedly bypassing those exact safety
• 8:43
  reviews and giving conflicting directives to the legal and
• 8:46
  technical teams, the governance model designed to protect the
• 8:49
  public interest begins to fracture entirely.
• 8:51
  It is. It's kind of like a ship captain
• 8:54
  telling the engineers the hole is fine and telling the
• 8:57
  passengers the engineers signed off on it, all while the ship is
• 9:01
  actively taking on water. That's a perfect comparison.
• 9:04
  If the communication channels are deliberately fragmented by
• 9:07
  the person in charge, the people responsible for preventing a
• 9:10
  collision cannot do their jobs effectively.
• 9:13
  The safety mechanisms only work if the information flowing
• 9:16
  between the departments is accurate.
• 9:18
  And that severely limits Open a Eyes public image as a
• 9:22
  transparent safety first organization.
• 9:25
  It really changes the public perception of the leadership
• 9:27
  team. Definitely.
• 9:28
  It heavily supports Musk's central claim that the nonprofit
• 9:32
  board's oversight was effectively bypassed by
• 9:34
  commercial ambition. When former board members and
• 9:37
  executives testify that major, incredibly powerful products
• 9:41
  were launched without adequate board communication, it opens up
• 9:45
  a crucial line of inquiry. About whether the charitable
• 9:47
  mission was entirely subordinated to product velocity
• 9:50
  and market dominance exactly. Well, we also have to look at
• 9:53
  the structural financial reality driving all of this internal
• 9:57
  chaos. Greg Brockman testified that
• 9:59
  Open AI expects to spend $50 billion on computing power in a
• 10:04
  single year. 50 billion. It's just hard to even
• 10:06
  comprehend. It is a staggering leap from
• 10:09
  their initial $30 million budget when they started.
• 10:12
  They're anticipating A cumulative compute spend of
• 10:15
  roughly $600 billion by the end of the current decade.
• 10:19
  And to actually afford this level of spending, the company
• 10:22
  must rent massive server capacity from the very investors
• 10:26
  funding them, specifically giants like Microsoft and
• 10:30
  Amazon. These major investments are
• 10:32
  often structured with strict conditions that require open AI
• 10:36
  to lease compute capacity directly back from the companies
• 10:39
  writing the checks. So the money just goes in a.
• 10:41
  Circle. Basically it creates a closed
• 10:43
  vendor financing loop where the investor capital never really
• 10:46
  leaves the ecosystem. The money is funneled directly
• 10:49
  back into the investors own cloud computing and hardware
• 10:51
  divisions, which artificially inflates the valuation while
• 10:55
  trapping the startup in a dependency cycle.
• 10:57
  Hold on. $50 billion just to keep the machines running.
• 11:01
  Yeah, the transition from a small research laboratory
• 11:04
  focused on publishing academic papers to a global platform
• 11:07
  serving hundreds of millions of active users every single day
• 11:11
  requires an unprecedented level of physical infrastructure.
• 11:14
  Because of the inference costs. Exactly.
• 11:16
  The inference costs alone dictate this.
• 11:19
  Inference is the processing power required to answer user
• 11:22
  queries in real time. In traditional software, you
• 11:26
  write the code once and distributing it to 1,000,000
• 11:29
  people costs practically nothing.
• 11:31
  It just scales. But in this field, every single
• 11:34
  time you type a prompt, a massive cluster of specialized
• 11:38
  chips in a data center has to physically spin up.
• 11:42
  It has to consume vast amounts of electricity and calculate the
• 11:45
  probabilities of the next word. It demands the continuous
• 11:49
  operation of specialized hardware at an unbelievable
• 11:52
  magnitude, consuming gigawatts of power.
• 11:54
  Wait, how does a company burning 50 billion on servers ever
• 11:58
  actually become profitable? You're talking about an
• 11:59
  operational expense that dwarfs the entire gross domestic
• 12:02
  product of some countries. If the cost of generating a
• 12:05
  response requires maintaining data centers that consume
• 12:08
  gigawatts of electricity, the unit economics seem completely
• 12:11
  inverted. They do.
• 12:12
  The more successful the product becomes, and the more users they
• 12:15
  get, the more money they actively lose.
• 12:16
  Just answering the. And this staggering cost limits
• 12:21
  open AI's ability to ever function as a pure charity
• 12:24
  again. It changes them from a Ninggle
• 12:26
  software startup into one of the absolute largest infrastructure
• 12:30
  buyers on the planet. Which changes everything.
• 12:32
  It does. It forces a relentless, never
• 12:34
  ending cycle of capital raising. A traditional nonprofit simply
• 12:38
  cannot sustain an operational burn rate of $50 billion.
• 12:42
  You cannot run a bake sale to buy a supercomputer.
• 12:44
  Definitely. Not the sheer necessity of
• 12:47
  securing that level of constant funding dictates a corporate
• 12:50
  structure capable of offering significant financial returns to
• 12:54
  institutional investors. The technology itself forces the
• 12:58
  corporate structure to change. So if a federal jury ultimately
• 13:01
  decides that early, informal promises to build technology for
• 13:05
  humanity are legally binding contracts, every tech company
• 13:08
  that ever used altruism as a fundraising pitch might suddenly
• 13:12
  find themselves in court. This trial exposes the raw
• 13:14
  friction between altruistic intentions and the ruthless
• 13:17
  capital intensive economics of building frontier technology.
• 13:21
  If you're not subscribed yet, take a second and hit follow on
• 13:24
  whatever app you're using. It helps us keep making this.
• 13:26
  We appreciate you being here. Also, check out our YouTube
• 13:29
  channel for more business and tech updates.
• 13:31
  There's a link in the description.