Product1 publisher3 min readPublished
Musk already had $38 million in OpenAI when he asked governors to regulate AI
Five executives who profit from AI spent the past few days asking for a slowdown. The Verge's timeline puts the same request in Musk's mouth in 2014 and Microsoft's in 2019, with few binding results either time.
The Product Desk · Product desk
What happened
- Altman, Amodei, Hassabis, Nadella and Musk all said publicly in the past few days that it is time to make everyone slow AI down before we lose control, according to The Verge.
- Musk's warnings escalated through 2014, from a real-life Terminator in June to "potentially more dangerous than nukes" in August and "summoning the demon" in October, then the Hawking open letter in January 2015.
- Microsoft president Brad Smith told a Brookings Institution audience that it was important for governments in 2019 to start adopting laws to regulate facial recognition technology.
- Microsoft researcher Eric Horvitz, now the company's chief scientific officer, had gathered leading AI researchers in 2009. The topic was policies that might constrain the behaviour of autonomous systems.
Compiled by The Product DeskSomething wrong?How this is made
Why it matters
- decision A team that slows its own roadmap this quarter on the strength of these statements is acting on a stated direction. The Verge's account does not tie any of the five asks to a rule or a date.
- precedent Horvitz asked in 2009, Smith asked in 2019, and The Verge counts few meaningful results from either. The reasonable planning assumption is another round of statements before an obligation lands.
- contradiction The people warning about loss of control are the ones funding and selling the systems, and The Verge treats that pairing as grounds for skepticism. A buyer should give the warning less weight as evidence.
- constraint The only enforceable AI terms a buyer has this quarter are still the vendor's own documents. The record gives procurement no clause to write.
Somebody forwards the week's quotes into the Monday planning review and asks whether the AI roadmap needs to change. What is in the quotes, according to The Verge, is agreement among five executives who stand to make money from AI that it is time to make everyone slow down before we lose control [1]. The Verge's account describes that direction without attributing a specific legislative proposal to any of the five [18].
The paper trail behind the request is long. Turing told a 1951 lecture audience that undying machines "would not take long to outstrip our feeble powers" and would eventually "take control" [4]. Bill Joy, then Sun Microsystems' cofounder and chief scientist, wrote in 2000 that self-replicating robots could be more dangerous than nuclear weapons. His reason, in part, was that they were being developed "within the now-unchallenged system of global capitalism" [5]. Sixty-eight years after Turing's lecture came the year Microsoft president Brad Smith asked governments to act on [15].
Musk's version is the one with dates on the money. In July 2017 he told the National Governors Association that "AI is a rare case where we need to be proactive about regulation instead of reactive. Because I think by the time we are reactive in AI regulation, it's too late." By then he had already invested $38 million in OpenAI and held stakes in at least two other AI companies, and had been warning publicly about AI's dangers for three years [8][7]. In March 2018 he said: "AI is far more dangerous than nukes. So why do we have no regulatory oversight?" [9] His dated public warnings run about 45 months, from June 2014 to that March 2018 remark [17].
Two of the five names in the current round belong to organisations already in the record [16]. Microsoft is there through Horvitz in 2009 and through Smith, who opened the door to voluntary regulation while the company was under fire for its facial recognition work with ICE [12]. Mark Zuckerberg's offer after Cambridge Analytica came with a condition attached: regulation, if it is the "right" regulation [13].
The Verge's own reading of the pattern is an industry that has been claiming to want legal guardrails for years, "with few meaningful results" [2]. It also notes that profit and alarm arriving together is itself a reason for skepticism [3].
So two tests before a plan moves. First, whether the statement names an obligation someone can be fined for missing. Smith's Brookings speech named a technology and a start year [11]. This week's statements name a direction [1]. Second, whether the same vendor changed a document your customers sign in the same week, meaning the pricing page, the retention default, the deprecation window.
Two controls are worth building either way: a model inventory, and a human approval step on anything that writes to a customer record. Both help if an obligation lands later, and both cut the Friday support load if none ever does. The cost is sprint time spent on controls no regulator has yet asked for.
What to watch
- A named bill or draft rule with a date attached to any of the five asks. That would turn a stated direction into a compliance input.
- Any of the five changing a customer-facing document in the weeks after the statements: retention defaults, deprecation windows, or model access tiers.
- Whether Microsoft again names a specific technology and a start year, as Smith did for facial recognition in his Brookings speech.