What is Really in the White House AI Agreement
It is a reasonable question – what is the real impact?
We did see today that OpenAI delayed the rollout of Astra .1, so maybe these guys are getting sensitive to the bad PR and concerned that they might actually get sued or investigated by the DoJ right in front of their IPO. If so, that is a small win and we will take it.
Gary Marcus, a well known writer on matters AI and industry pundit summarized the agreement in VERY UNFLATTERING terms:
- They agree not to be regulated
- We agree NOT to give the public any voice in the matter, even though they are the ones who will be impacted
- You should trust us
Those are Gary’s words and he is pretty knowledgeable in this area, so I would not consider him some random guy with an opinion.
They also agree to implement robust internal controls (I assume they thought they already had that), empower internal teams to ensure controls and monitoring are working, partner with an external vendor to carry out assessment of controls (note that many of these breaches were actually the fault of independent vendors) and designate an independent committee of the board to oversee this.
Given these companies are private, it is very unclear what independent actually means.
Credit: Gary Marcus
The prez also let the veep talk a bit. He said they would prefer working with industry than regulating it. That is probably a good thought given that Congress is tech-clueless and any regulation would either be clueless or onerous or both. BUT, my confidence in self regulation is low. Still, it is good that they are at least discussing the subject.
He also signed an EO renaming AI to SI (not to be confused with Sports Illustrated for you sports nerds). It only affects what federal government agencies do and only when convenient, so I don’t think it will catch on elsewhere, but who knows.
Credit: NextGov/FCW
by 