Conversation
-
In the “Unites States” of “Super Intelligence,” “Issues Are Remediated”
source
As many people before me have pointed out, once you notice that the signature page for Trump’s “morally binding” agreement on AI safety spells “United States” “Unites States,” the entire charade becomes more ominous than it already was.
Over the course of the day yesterday, Trump first posted a seating chart showing himself surrounded by some of the most powerful (and/or perhaps fraudulent) men in the world, then had lunch with them, then had them sign a non-binding and totally inadequate document (only the signatures of the most powerful men appeared on the same page as Trump’s), then signed an Executive Order (which binds the US government) renaming AI “Super Intelligence.”
This was elder care masquerading as world-saving policy, stringing together some of Trump’s favorite activities — proximity to power, document signings, and arbitrary renamings — for an entire day. I found myself wondering (since he’s not on the seating chart) what was Stephen Miller doing while he had the richest men in history babysitting his ward?
The agreement itself is totally inadequate fluff, like something McKinsey might dream up with no mechanism to ensure it works.
In order to build a positive future for the American people and the world, we believe every company is responsible for developing its own technology safely and in a way that builds trust with customers and the public.
This starts with every company that is training and deploying frontier models having robust internal processes and controls to ensure that their technology behaves as intended and that any issues are promptly identified and resolved. Therefore, in addition to any other precautions, we believe each company should implement the following four layers of controls and audits:
- Implement robust internal controls to monitor the capabilities and alignment of its models during training and deployment around areas like cybersecurity, biosecurity, and chemical threats, and to ensure that its models do not hack or access technical systems in unintended ways.
- Empower an internal team to ensure all of the controls, monitoring, and detection are operating as intended, and that any issues are remediated.
- Partner with an independent external auditor or evaluator to carry out independent assessments of whether the controls, monitoring, and detection are operating as intended.
- Designate an independent committee of the board of directors to oversee and receive reports from the teams operating the controls and the internal and external auditors and evaluators, as well as to ensure any issues identified are remediated.
Together, these steps will give each company, its customers, and the public confidence that the technology is operating as intended.
The participating companies will meet regularly to establish standards and best practices to improve the safety of their systems.
Over time, it may make sense to codify these steps into laws or regulations. Regardless of whether this is required of companies, we believe that implementing these controls and audits is critical to ensuring a safe future for everyone, and each of our companies are committed to doing this. [my pink emphasis]
Underneath all the bullet points and active verbs, this agreement is designed to be useless, for two main reasons. First, it only requires AI to operate “as intended;” (and the parallel structure in bullet 1 would envision hacks so long as they do what they are intended to do). The self-imagined Masters of the Universe who signed it don’t have to provide assurances about their intentions, which are mostly malign.
In addition, the single thing that such documents must do is assign accountability for implementing it — naming the person who will go to jail for hacks that do not function as intended. This does not. Bullet 2 “empowers” an internal team to ensure that “any issues identified are remediated,” but it doesn’t say that that internal team (or anyone else) has to do the remediation. Bullet 4 requires a committee on the board that will “receive reports” and “ensure any issues identified are remediated.” But it doesn’t hold them accountable for remediating those issues themselves. The passive voice lurks where it is most dangerous, where someone is accountable for remediating the things that go wrong.
But even if the document weren’t so inadequate and the official renaming so pathetic, there is no one left inside the Oval Office with the self-respect or competence to spell “United States” correctly. There is no one inside the Oval Office who gave a rat’s ass to make this propaganda at least look serious.
This was not a commitment to make AI safe for humanity. Rather, it was a commitment to collectively tell the public that the Masters of the Universe can take care of this themselves, without regulation.
And even the propaganda was too slovenly to get basic details right.
The post In the “Unites States” of “Super Intelligence,” “Issues Are Remediated” appeared first on emptywheel.