- BIST10012,151-0.94%
- USD49.220.00%
- EUR55.21+0.28%
- GOLD4,192.60-0.56%
- INTEREST37.00%-1.00
- BITCOIN82,900-0.92%
- October 12, 202619°
- BIST10012,151-0.94%
- USD49.220.00%
- EUR55.21+0.28%
- GOLD4,192.60-0.56%
- INTEREST37.00%-1.00
- BITCOIN82,900-0.92%
- October 12, 202619°
Anthropic Calls for Coordinated Halt in AI Development to Ensure Safety
Anthropic has called for a coordinated slowdown or halt mechanism for advanced AI models. OpenAI, meanwhile, argues that rules should be set by democratic governments...
As competition in artificial intelligence accelerates, Anthropic has called for the establishment of a coordinated system capable of halting the development of advanced models if necessary. The company warned that the technology could approach a point beyond humanity's capacity to control it.
The company behind the Claude chatbot argued that as advanced AI systems perform tasks increasingly faster, it would be beneficial for the world to have the option to slow down or temporarily pause their progress.
Anthropic announced that its internal research unit will examine this issue in collaboration with other institutions and take steps to contribute to the establishment of a reliable slowdown or halt mechanism. However, the company did not share details about the specific measures to be taken.
According to the company's assessment, if current trends continue and sufficient computing power is available, an AI system could reach a point where it can design and develop its own successor. This phenomenon is known in the industry as "recursive self-improvement."
Anthropic acknowledged that this could bring significant advances in science, health, and other fields. However, it also emphasized the risk of weakening human control over AI systems.
The company's statement noted that such a slowdown would buy time for social institutions and alignment research to keep pace with AI progress. Here, alignment refers to efforts to ensure that technology aligns with human values and intentions.
Under the proposed coordination model, advanced AI laboratories would be able to verify that global competitors are genuinely slowing down or stopping. This aims to prevent a malicious actor from secretly advancing under the guise of coordinated slowdown.
Anthropic argued that a global mechanism is necessary; otherwise, the least cautious players in the industry could gain speed, increasing pressure on companies and governments.
The announcement came at a time when Anthropic and OpenAI, the developer of ChatGPT, are competing amid expectations of a high valuation in share sales and a possible public offering. It is rumored that Anthropic's market value could approach one trillion dollars.
In contrast, OpenAI took a different approach in a separate report published the same week. The company stated that AI-related rules, safety measures, and accountability mechanisms should be determined not by private companies alone but by democratic governments.
OpenAI also emphasized that decisions about the pace of AI innovation should not be left in the hands of any single laboratory, company, or private interest group.
Anthropic's warning followed another study published this week by researchers at the University of Toronto. The researchers demonstrated that AI tools could be used to create a new AI worm that adapts its attack strategy as it spreads from device to device.
The lead researcher of the study, Nicolas Papernot, said this shows that not only the largest and most powerful language models pose security risks. Papernot also revealed that he informed Canadian cybersecurity authorities before publishing the research.
The researchers developed this worm using an open-source AI tool accessible and modifiable by software developers in the lab. Papernot recalled that in the past, cyber attackers mainly targeted high-value systems such as banking, hospitals, power grids, water treatment plants, and schools.
However, according to Papernot, due to low-cost attack tools, any internet-connected system is at greater risk. The researcher noted that even an unused old laptop could be the starting point of an attack chain leading to larger targets.
Papernot stressed the need for increased collaboration among companies, government agencies, and academic researchers to develop new defense methods against AI-powered attacks.
Stay Updated with The Turkiye Times
Subscribe to be informed about special news, developments and announcements. We will keep you informed.
By signing up, you agree to our Privacy Policy
