Latest / Elon Musk Podcast / Why Boring AI Might Beat Brilliant AI in the Fight for Corporate America

Transcript

Why Boring AI Might Beat Brilliant AI in the Fight for Corporate America · Elon Musk Podcast

• 0:00
  Something interesting is happening in corporate
• 0:02
  boardrooms right now, and it's not showing up in the consumer
• 0:05
  tech headlines. The battle for AI dominance
• 0:08
  isn't being fought on Twitter or in App Store rankings anymore.
• 0:11
  It's being fought in procurement meetings, security audits, and
• 0:15
  enterprise contracts for 10s of millions of dollars in
• 0:19
  Anthropic, the company behind Claude, making serious inroads
• 0:23
  against open AI in places that matter most.
• 0:26
  The big companies, The Fortune 500 companies.
• 0:29
  I want to talk about why the shift is happening, what it
• 0:31
  tells us about why AI is actually headed, and why the
• 0:35
  enterprise market might determine who wins the entire AI
• 0:38
  race. And let's get into some context
• 0:42
  here. For the past few years, when
• 0:44
  people talked about AI competition, they mostly meant
• 0:47
  consumer products chat, TPT versus Claude versus Gemini.
• 0:51
  Which one writes better emails? Which one helps you code faster?
• 0:55
  Which one your teenager uses for homework now?
• 0:58
  Stuff matters, but it's not where the real money is.
• 1:01
  Enterprise AI is a completely different game.
• 1:04
  We're talking about companies deploying AI across thousands of
• 1:07
  employees, integrating it into core business operations,
• 1:10
  handling sensitive customer data, and needing to explain
• 1:14
  their AI decisions to regulators.
• 1:16
  The requirements are stricter, the contracts are longer, and
• 1:19
  the switching costs are enormous.
• 1:23
  Open AI got there first. They launched ChatGPT Enterprise
• 1:27
  in 2023 and built an impressive roster of corporate clients.
• 1:31
  They had the brand recognition, the first mover advantage, and
• 1:34
  frankly, the hype machine working in their favor.
• 1:37
  And for a while, AI and Open AI were basically the same things
• 1:42
  in a lot of boardrooms. And here's what's interesting.
• 1:46
  Anthropic has been quietly eating into that lead, and
• 1:50
  they're doing it by being boring.
• 1:53
  And I mean that as a compliment. Think about what a chief
• 1:56
  information officer actually cares about when deploying AI
• 1:59
  across an organization. They're not impressed by viral
• 2:03
  demos or creative writing samples.
• 2:06
  They want to know, will this thing make stuff up?
• 2:10
  Will it hallucinate? Can we audit what it does?
• 2:13
  Let handle our proprietary data safely?
• 2:16
  What happens when something goes awry?
• 2:20
  Anthropic Bill Claude around those concerns from the
• 2:23
  beginning. Their constitutional AI
• 2:25
  approach. The emphasis on honesty and
• 2:28
  harmless outputs, the attention to reliability over raw
• 2:32
  capability. These aren't exciting features.
• 2:35
  A demo at a tech conference. They're exactly what risk averse
• 2:39
  enterprises want to hear. Now I want to hear from you.
• 2:43
  I've been digging through the analytics of the show and I
• 2:46
  noticed 39% of you are following the channel.
• 2:49
  And for you, thank you so much. The other 61% of you, you've
• 2:54
  been listening, you've been watching, but you haven't hit
• 2:56
  the follow or subscribe button yet.
• 2:58
  And I'm an independent journalist covering Elon Musk,
• 3:01
  Space Flight tech, AI for the last seven years.
• 3:05
  And I'll continue for the next 10.
• 3:07
  And all I ask from you is one second of your time and hit the
• 3:10
  follow or subscribe button on the platform you're watching or
• 3:14
  listening on right now. I'm extremely grateful for you,
• 3:17
  blessed to have you in this community.
• 3:19
  So thank you for doing that. Now let's get back to Claude.
• 3:24
  There's another factor here that doesn't get enough attention.
• 3:27
  It's called the trust deficit. Open AI has had a rough couple
• 3:31
  of years on the governance front.
• 3:33
  Sam Altman firing and unfiring in 2023 spooked a lot of
• 3:38
  enterprise customers. Then you had ongoing questions
• 3:41
  about their transition from nonprofit to for profit
• 3:44
  structure, leadership changes, and debates about their safety
• 3:48
  commitments. None of that necessarily effects
• 3:50
  how well GPT 4 or GPT 5 works. But enterprises care about
• 3:55
  stability. They're signing multi year multi
• 3:59
  hundreds of $1,000,000 contracts.
• 4:02
  They need to believe their AI vendor will still exist and
• 4:04
  still operate the same way in around 3 years from now.
• 4:08
  Entropic by contrast has maintained a pretty consistent
• 4:11
  story. Same leadership they or safety
• 4:16
  focused, same research direction.
• 4:19
  For a procurement team trying to justify a major vendor decision
• 4:23
  and to spend millions of dollars, that consistency really
• 4:26
  does matter. No, I don't want to overstate
• 4:29
  this. Open AI is still the market
• 4:32
  leader in enterprise AI by most measures.
• 4:35
  They have more customers, more revenue, and more name
• 4:37
  recognition. But market share shifting and
• 4:40
  the direction of that shift tells you something important
• 4:43
  about what corporations actually value when the hype settles
• 4:45
  down. Here's the key part.
• 4:48
  Enterprise adoption is a leading indicator for the whole AI
• 4:52
  industry. Consumer products come and go.
• 4:55
  Remember when everyone was excited about character based AI
• 4:58
  chat bots or AI generated art tools?
• 5:01
  You know some of them are still around, but the hype cycles move
• 5:04
  pretty fast. Enterprise deployment stay a lot
• 5:07
  longer. They're stickier.
• 5:09
  And when a bank or a healthcare system or an automaker builds AI
• 5:13
  into their core workflows, they're making a bet that lasts
• 5:16
  years, not just months of hype. So the companies that win
• 5:21
  enterprise contracts today are positioning themselves to be
• 5:24
  dominant for the next decade. They're getting access to
• 5:27
  proprietary data that makes their models better.
• 5:30
  They're building relationships with decision makers who will
• 5:32
  influence the next round of purchases.
• 5:35
  They're generating the kind of steady revenue the funds,
• 5:38
  continued research, and continued growth.
• 5:41
  And This is why Anthropic has been so focused on enterprise.
• 5:45
  Even though the consumer product arguably gets less attention
• 5:48
  than ChatGPT, they understand that winning corporate
• 5:51
  boardrooms might matter more than winning a Twitter argument.
• 5:56
  And there's also a regulatory angle worth mentioning.
• 5:59
  Governments around the world are getting more serious about AI
• 6:02
  oversight. The E US AI Act is now in
• 6:05
  effect. the US has executive orders and proposed legislation.
• 6:10
  Other countries are developing their own frameworks around AI.
• 6:14
  Enterprises operating in regulatory industries need to
• 6:17
  demonstrate that their AI systems meet certain standards.
• 6:21
  They need audit trails. They need explainability.
• 6:25
  They need to show that they've done due diligence on safety and
• 6:28
  bias in a company that build its entire AI system around those
• 6:31
  principles from day one has a huge advantage over a company
• 6:35
  that's retrofitting safety features onto a system designed
• 6:39
  primarily for capability. It's not a knock on open A IS
• 6:43
  engineering. It's recognition that design
• 6:46
  choices made early tend to persist and you got to start
• 6:51
  from the right perspective. So what does this all mean going
• 6:54
  forward? I think we're entering a phase
• 6:58
  where AI competition looks less than like a Sprint and more like
• 7:03
  a long term marathon. No more flashy demos that stay
• 7:10
  around. The demos will come and go, but
• 7:13
  the real action in deployment, in integration, in the boring
• 7:16
  work of making AI reliable enough that a risk averse
• 7:20
  corporation will actually bet on it and spend those hundreds of
• 7:24
  millions of dollars. Entropic isn't going to out
• 7:27
  spend open AI on compute or out market them on brand awareness,
• 7:31
  but they might out trust them. And in enterprise that's worth
• 7:36
  more than just hype. In the next few years will tell
• 7:39
  us whether this strategy works. I'm a huge proponent of Claude.
• 7:43
  Claude code has saved the day so many times for me.
• 7:46
  I think it's amazing. And enterprise contracts take
• 7:49
  time to materialize. And financial results.
• 7:52
  It takes time to get that money. But if you want to understand
• 7:55
  who's actually winning, stop looking at App Store rankings
• 7:58
  and start looking at which companies are getting invited to
• 8:01
  procurement meetings at major corporations.
• 8:05
  That's where the future is being decided, one really boring
• 8:08
  contract at a time.