The US government held the superintelligence makers responsible for any future incident, placing moral obligations on them until formal laws are framed, with the judicial system ready to act if required. However no time frame has been set for legal bindings stating it has started. At the same time, users should be assured that the technology is safe to use. The agreement was signed by tech giants, including OpenAI President Greg Brockman, Elon Musk, Anthropic CEO Dario Amodei, Mark Zuckerberg, NVIDIA CEO Jensen Huang, and Google CEO Sunder Pichai.
The agreement details tech giants’ responsibility not to produce dangerous superintelligence and requires companies to take steps to assure and ensure people worldwide of the moral obligations of superintelligence companies.
This deal requires tech leaders four layers of controls and audits ensuring companies must develop refined, robust internal processes so the superintelligence they build adheres to high-expectation standards. Any anomaly must be handled/resolved and reported on a timely basis. These manual controls would become laws and regulations over time. The four layers are:
— Internal controls monitor the models during training and after deployment for biochemical risks and do not perform cyber threats of any kind.
— Each company must appoint teams to do the same.
— Each company must partner with external auditors to achieve transparency and productivity without harming technical systems, biochemical exploitation, or other related issues.
— Appoint a board of directors to oversee all aspects of the release process, before, during, and after. To ensure all issues with teams and auditors are addressed.
— They also intend to meet regularly to update on the status and share best practices with each other, so that the safety of super intelligence systems is taken care of with mutual benefits.
This can provide some breathing room. But the question is the timeline? How much time would be allocated to each of the four processes? How much time before the model’s release should be spent on processes and bureaucracy? Would too much bureaucracy slow it down? What is the speed we need?
The umbrella is small; we need to enlarge it to include Chinese companies’ actions as well. What about new startups with unintended intentions thatmay go rogue? What about EU AI labs? Would other countries’ AI labs also partner with the USA if their URLs are accessible?
Thank you for reading.
Subscribe for updates.
