00:00Amid growing concerns over the risks posed by advanced AI,
00:05tech leaders have used the United Nations General Assembly
00:07to call for stronger safety standards and regulation.
00:11Among them is OpenAI CEO Sam Altman,
00:14who called for greater caution in AI development,
00:17warning against rushing ahead when safety cannot be guaranteed.
00:22The industry must not accept too much technological risk
00:25just because the benefits are too great
00:27and that they feel too important to slow down.
00:30Beating companies in a competitive race is not a reason to make rash decisions,
00:33nor do we believe we are locked in a race where we are unable to do that.
00:36We have unilaterally slowed down in the past.
00:38We will do so in the future.
00:39We do not want to build systems.
00:41We think it's important that everyone avoids building systems
00:43where we cannot make the alignment, monitorability, and safety guarantees that we must.
00:49Antropic CEO Dario Amode echoed Altman's call for greater safeguards,
00:54outlining several steps towards global AI regulation.
00:57First, we should begin with narrow agreements that every member can support,
01:03such as a ban on using AI to make biological weapons
01:07or permitting your AI technologies to be used to make biological weapons.
01:13Second, we should build evaluation and verification systems
01:17that keep pace with AI development
01:20so that states can have visibility into frontier model capability
01:24and can verify each other's commitments.
01:27Third, we should establish common global standards
01:30for testing AI models for loss of control risks and misuse risks,
01:36and a notification system for AI incidents
01:39that are significant to global security.
01:42third, we should plan on finding multiple
01:42third, we'll one, two, two, one, two, three, two, one, four, one, three.
01:42Next we should do insight is a sponsors as a Lightning V loss,
01:42and as a黃 was one, three to the last NES data system
01:44that are effective.
01:45So we should probably have information as a way
Comments