Latest / Elon Musk Podcast / Musk vs Altman OpenAI Commercial Pivot

Transcript

Musk vs Altman OpenAI Commercial Pivot · Elon Musk Podcast

• 0:00
  The courtroom battle between Elon Musk and Open AI has really
• 0:03
  turned from a standard contract dispute into, well, a personal
• 0:08
  referendum on whether open AICEO Sam Altman can actually be
• 0:12
  trusted. Right.
• 0:13
  And you really have to look at the financial stakes involved
• 0:16
  here. I mean Musk put roughly
• 0:18
  $44,000,000 into Open AI early on and that money was
• 0:23
  specifically to fund a non profit open source research lab,
• 0:26
  you know 1 focused entirely on safety.
• 0:29
  But now Open AI Guy is operating as a caft profit company.
• 0:32
  Exactly. They're taking over $13 billion
• 0:34
  from Microsoft and keeping the architecture for GPT 4 as this
• 0:37
  closely guarded trade secret. Which really brings us to the
• 0:40
  core question of this trial. I mean, can a founder promise
• 0:43
  one path to early backers, take their money and legally pivot to
• 0:47
  a completely different corporate structure when billions of
• 0:50
  dollars are on the line? Our mission here is to piece
• 0:53
  together the reality behind that open source ideal.
• 0:55
  Yeah. We want to figure out how a
• 0:56
  nonprofit research lab morphed into a tech.
• 1:00
  Well, Altman actually took the stand to defend this exact
• 1:03
  transformation, and his whole argument is grounded in the
• 1:07
  physical reality of building artificial intelligence.
• 1:10
  He basically testified that building cutting edge AI
• 1:14
  requires an immense amount of computational power.
• 1:17
  Yeah, it's not cheap. No, not at all.
• 1:19
  So According to him, the shift away from a pure nonprofit
• 1:22
  structure wasn't a betrayal of the original mission.
• 1:25
  It was a matter of survival. I mean, I struggle a bit with
• 1:30
  the purity of that survival narrative.
• 1:32
  How so? Obviously servers are expensive,
• 1:35
  but there is a huge leap between needing money to pay the
• 1:38
  electricity bill and creating a complex corporate shell company
• 1:43
  to take $13 billion from Microsoft.
• 1:45
  Well, hiding the core technology, right?
• 1:48
  Exactly. You do have to look at the
• 1:49
  mechanics of what it actually takes to train these models
• 1:52
  though. We aren't talking about running
• 1:54
  a traditional software startup where you can just rent some
• 1:57
  cloud space on Amazon and build an app.
• 1:59
  Yeah, it's not a couple of coders in a garage anymore.
• 2:02
  No, the capital requirements for training large language models
• 2:05
  are much closer to heavy manufacturing.
• 2:07
  Wow. Yeah, they need hundreds of
• 2:09
  millions and eventually billions of dollars to purchase highly
• 2:12
  specialized silicon chips. And to secure huge data centers.
• 2:17
  Right, and literally power the cooling systems so the servers
• 2:22
  don't just melt. I mean, you simply cannot fund
• 2:24
  that level of physical infrastructure with standard
• 2:28
  terrible donations. That makes sense on a technical
• 2:30
  level, but Musk's legal team is really attacking the timeline of
• 2:34
  that realization. They brought the receipts.
• 2:36
  They did. They brought internal emails
• 2:38
  into the courtroom, and those messages paint a picture that
• 2:41
  directly contradicts the whole reluctant survival story.
• 2:45
  Right, because in one e-mail, Altman is openly discussing
• 2:48
  paths to AGI that make money. And in another thread, there are
• 2:52
  conversations about needing to compete directly with Google
• 2:55
  rather than just, you know, publishing academic research
• 2:58
  papers. Which for anyone listening AGI
• 3:00
  is Artificial General Intelligence.
• 3:02
  It's the theoretical point where a machine can do any cognitive
• 3:06
  task a human can. The presence of those emails is
• 3:09
  legally damaging. It shifts the entire argument
• 3:12
  away from technical necessity and really focuses the trial on
• 3:16
  original intent. Because if Altman and the
• 3:19
  leadership team were discussing how to commercialize the
• 3:22
  technology, if they were talking about beating corporate rivals
• 3:26
  early in the company's lifespan, it means they were operating
• 3:29
  with a commercial mindset from the start.
• 3:32
  Yeah, and in contract law, the concept of good faith is the
• 3:35
  operating principle for disputes like this.
• 3:38
  And good faith isn't just about being polite.
• 3:40
  No, it means you are acting transparently with your
• 3:43
  partners. You aren't secretly working
• 3:45
  against the spirit of the agreement you signed.
• 3:47
  Right When you solicit funds for a specific stated charitable
• 3:51
  mission, your donors assume you are fully committed to that
• 3:54
  path. Yeah.
• 3:55
  So if you have a hidden parallel track focused on
• 3:57
  commercialization, the argument that you were forced into a
• 4:00
  capped profit model loses its legal protection.
• 4:03
  We should probably clarify how that capped profit structure
• 4:06
  actually works, because it's central to the dispute.
• 4:09
  Yeah, it's a bit strange. Open AI essentially built a
• 4:11
  Public Library, realized the books were too expensive to
• 4:14
  print, and decided to put a tool booth at the front door.
• 4:17
  While letting Microsoft open the VIP lines in the back.
• 4:21
  Pretty much the capped profit model is a hybrid setup.
• 4:25
  Investors can put money in and make return, but only up to a
• 4:28
  strictly defined limit. And anything generated beyond
• 4:31
  that cap is supposed to flow back to the nonprofit entity.
• 4:35
  Right, except when the cap is set at 100 times the initial
• 4:38
  investment, it functions exactly like a traditional corporate
• 4:42
  entity. For all practical purposes,
• 4:43
  yeah. And if Altman's internal
• 4:45
  transparency is the weapon Musk's team is using to prove
• 4:49
  bad faith, that naturally opens up a broader evaluation of how
• 4:53
  Altman communicates with the people around him.
• 4:55
  Which became a major focal point in the testimony.
• 4:57
  Right with Helen Toner. Yeah, former board member Helen
• 5:00
  Toner took the stand, and she discussed Altman providing
• 5:03
  incomplete information to the directors.
• 5:05
  And this wasn't just about the core operations of ChatGPT.
• 5:08
  No. The testimony included mentions
• 5:10
  of his personal side ventures, specifically his investments in
• 5:14
  AI chip manufacturing and biometric identity verification
• 5:17
  projects. I mean, a board of directors
• 5:19
  relies entirely on the information the CEO chooses to
• 5:22
  share. Right.
• 5:23
  They have a fiduciary duty to oversee the organization, but if
• 5:26
  the chief executive filters that information or omits details
• 5:30
  about personal ventures that directly intersect with the
• 5:33
  company's supply chain. The board is effectively
• 5:36
  paralyzed. Yeah, they cannot govern what
• 5:38
  they cannot see. So Must's team gragged Altman's
• 5:42
  entire history of being fired and quickly reinstated as CEO
• 5:46
  into the court record. To establish a pattern of
• 5:48
  behavior. Exactly because in founder
• 5:51
  disputes, a judge isn't just reading the literal language of
• 5:54
  the incorporation documents, they have wide discretion to
• 5:57
  interpret the credibility of the parties involved.
• 5:59
  They are evaluating the character of the executive
• 6:01
  making the decisions. Trust literally becomes a
• 6:04
  tangible legal metric. We have to connect this legal
• 6:07
  strategy to the real world consequences for Open AI.
• 6:12
  The company is reportedly out of the market, seeking new funding
• 6:15
  at evaluation north of $100 billion.
• 6:18
  And institutional investors writing checks of that
• 6:20
  magnitude, they demand predictability, right?
• 6:23
  They want absolute assurance that the company can operate as
• 6:27
  a normal, functional business. So a judgement finding that
• 6:30
  Altman violated his founding commitments would be
• 6:33
  catastrophic for a capital raise.
• 6:35
  Absolutely. If the court decides there was a
• 6:37
  fundamental breach of trust, the remedies aren't limited to a
• 6:40
  slap on the wrist. The judge could force sweeping
• 6:44
  governance changes, restructure the board or levy severe
• 6:48
  financial damages. And capital flees from legal
• 6:50
  ambiguity. If there is a risk that a court
• 6:53
  might fundamentally alter how open AI is allowed to operate,
• 6:56
  Microsoft and other potential backers might just pause their
• 6:59
  investments. But, you know, if we are
• 7:01
  evaluating Altman's motives and character, we probably have to
• 7:04
  look closely at the person funding the lawsuit against him.
• 7:07
  Yeah, the internal documents presented by Open A IS defense
• 7:10
  team show that Elon Musk was not exactly a Saint of open source
• 7:14
  purity during those early years years.
• 7:16
  The defense strategy here is incredibly effective.
• 7:18
  They introduced Musk's own proposals from before he left
• 7:22
  the organization. Because Musk actually pushed for
• 7:24
  a for profit structure himself. Right when he realized the scale
• 7:28
  of computing power required to build AGI, he proposed merging
• 7:32
  Open AI with Tesla. Or creating a new for profit AGI
• 7:36
  effort housed entirely within the automaker.
• 7:38
  The mechanics of that proposed merger are pretty revealing.
• 7:42
  Tesla has vast computational resources and an ocean of real
• 7:46
  world video data gathered from their vehicles on the road.
• 7:49
  Right, so merging the AI lab into a publicly traded car
• 7:52
  company is a fiercely commercial move.
• 7:55
  It would have instantly killed the open source nonprofit ideal.
• 7:57
  Completely open AI would have simply become a research
• 8:00
  division of Tesla. And when those plans to merge
• 8:02
  fell through and the board rejected his proposal, Musk left
• 8:06
  the organization. Fast forward a bit and he
• 8:08
  launches XAI. Which is operating as a direct
• 8:12
  aggressive competitor to open AI in the consumer and enterprise
• 8:16
  markets. Which completely reframes the
• 8:18
  narrative of this lawsuit. Open AI's attorneys are arguing
• 8:22
  that this legal action is born out of competitive jealousy.
• 8:25
  Not some principled stand to protect humanity from corporate
• 8:28
  AI. Right.
• 8:29
  They are framing the entire trial as a tactical discovery
• 8:32
  expedition. Musk is using the legal system
• 8:35
  to force his biggest rival to open up their books, reveal
• 8:38
  their internal emails. And slow down their momentum
• 8:41
  while X AI tries to catch up in the market.
• 8:43
  It severely damages Musk's moral high ground in the courtroom.
• 8:46
  It really does. What initially looked like a
• 8:48
  clear cut story of a billionaire trying to hold a greedy
• 8:51
  corporation accountable turns into a messy, complicated battle
• 8:55
  between 2 rival executives. With parallel commercial
• 8:58
  interests. Yeah, neither side emerges
• 9:00
  looking like a selfless protector of academic research.
• 9:03
  If you step back from the specific dynamic between Musk
• 9:06
  and Altman, you realize they're just the loudest voices in a
• 9:09
  room where everyone is making the exact same compromises.
• 9:12
  The tension between stated ideals and the race for market
• 9:16
  dominance is just the operating reality for the entire
• 9:19
  artificial intelligence sector right now.
• 9:21
  Yeah, commercial incentives consistently win over safety
• 9:25
  commitments across all the major labs.
• 9:27
  And Musk's attorneys actually use these industry wide patterns
• 9:31
  to argue that Altman's trajectory was entirely
• 9:34
  predictable. Right, they point to the fact
• 9:36
  that Open AI recently disbanded its Super Alignment team.
• 9:40
  That was an internal group specifically tasked with
• 9:42
  researching long term safety and ensuring that future models do
• 9:46
  not act against human interests. And they dissolved it just to
• 9:49
  focus resources on product development.
• 9:51
  It's not just an open AI issue though.
• 9:54
  Google rushed out their Bard model to compete directly with
• 9:57
  ChatGPT. Despite well documented internal
• 10:00
  concerns from their own safety researchers who felt the product
• 10:03
  was not ready for public release.
• 10:04
  The pressure to capture market share overrode the internal
• 10:07
  guardrails. Meta is doing the same thing
• 10:09
  from a different angle. They pushed forward and open
• 10:12
  source their llama models. When we talk about open sourcing
• 10:15
  a model, we're talking about releasing the weights.
• 10:17
  Right, the weights are the actual mathematical parameters
• 10:20
  the software learned during training.
• 10:22
  It's the core brain of the system.
• 10:24
  And releasing those weights means anyone can download the
• 10:27
  model, run it locally and just strip away all the safety
• 10:31
  filters. Which Meta did, over the
• 10:33
  explicit objections of researchers who worried about
• 10:36
  the risks of handing powerful AI architecture directly to the
• 10:39
  public without any oversight. Every major lab claims to
• 10:43
  prioritize responsible development.
• 10:45
  They all have internal ethics boards, safety guidelines and
• 10:49
  public relations teams dedicated to talking about alignment.
• 10:52
  But this trial exposes an uncomfortable truth for anyone
• 10:56
  interacting with these tools. When there is pressure to ship a
• 10:59
  product and a chance to capture billions in revenue, the money
• 11:02
  always dictates the timeline. Look at Enthropic.
• 11:05
  Their entire corporate existence is a reaction to this exact
• 11:08
  dynamic. Yeah, the company was founded by
• 11:10
  former Open AI researchers who left the organization.
• 11:13
  Because they were uncomfortable with the rapid push toward
• 11:16
  commercialization and the partnership with Microsoft.
• 11:19
  Right. They've formed Anthropic as a
• 11:21
  public benefit corporation to legally lock in their commitment
• 11:24
  to safety. And now they are out in the
• 11:27
  market using safety as a competitive differentiator.
• 11:31
  They are essentially selling trust as a product feature.
• 11:34
  DeepMind is facing a similar internal struggle.
• 11:37
  They operate within Google and were originally founded as a
• 11:40
  pure research lab focused on solving intelligence.
• 11:44
  Now they're dealing with intense commercial pressure to integrate
• 11:48
  their AI research into every single Google product.
• 11:52
  Just to justify the huge computing costs they incur.
• 11:55
  Yeah, the structural forces of the market make it nearly
• 11:57
  impossible to operate a pure research lab once the technology
• 12:00
  proves it can generate real revenue.
• 12:02
  So regardless of who wins the specific legal arguments in this
• 12:05
  courtroom, the trial is forcing the industry to confront a much
• 12:09
  larger issue. The outcome here sets a
• 12:11
  precedent for how future startups will incorporate
• 12:14
  themselves. It dictates what kind of
• 12:16
  promises a founder can legally make to early backers.
• 12:19
  And whether an initial mission statement holds any weight once
• 12:21
  market dominance enters the picture.
• 12:24
  Even after the verdict is handed down and the lawyers go home,
• 12:27
  the central question of whether we can trust the people building
• 12:30
  humanity's most powerful technology remains entirely
• 12:34
  open. If you're not subscribed yet,
• 12:37
  take a second and hit follow on whatever app you're using.
• 12:40
  It helps us keep making this. We appreciate you being here.
• 12:43
  Also, check out our YouTube channel for more business and
• 12:45
  tech updates. There's a link in the
• 12:47
  description.