Dario Amodei calls OpenAI’s messaging around military deal ‘straight up lies’ (techcrunch.com)

Article summary

Anthropic CEO Dario Amodei has criticized OpenAI's deal with the US Department of Defense, calling their messaging 'straight up lies'. Amodei claims that OpenAI's contract allows for the use of their AI systems for 'all lawful purposes', which could potentially include domestic surveillance or autonomous weaponry. Anthropic had previously refused to agree to a similar deal with the DoD due to concerns over the potential misuse of their technology. The public appears to be siding with Anthropic, with a significant increase in uninstalls of OpenAI's ChatGPT.

Main themes

• AI ethics
• Military contracts
• Surveillance concerns
• Corporate responsibility
• Government partnerships
• Autonomous weaponry

What commenters say

• Anthropic's partnership with Palantir, a surveillance company, undermines their claims of being a responsible and ethical steward of LLM technology.
• The distinction between collecting and processing surveillance data is important, and companies like Palantir should not be held solely responsible for the actions of their clients.
• The fact that Palantir's software can be used for surveillance and data integration makes them complicit in potential human rights abuses, regardless of their intentions.
• Anthropic's public statements about their values and ethics are mere theater, and their actions are ultimately driven by financial interests.
• The US government's ability to change laws and policies overnight, as well as the existence of secret FISA court decisions, makes it difficult to determine what constitutes 'lawful use' of AI technology.
• Anthropic's status as a Public Benefit Corporation may protect them from investor pressure to prioritize profits over their mission statement, but this does not necessarily mean they will always act in accordance with their stated values.
• OpenAI's contract with the DoD likely contains loopholes that would allow for the development of autonomous weaponry or other potentially harmful applications, despite their claims of having similar constraints to Anthropic's proposed terms.
• The public's perception of Anthropic as a hero and OpenAI as a villain may be misguided, and both companies may be equally culpable in the potential misuse of their technology.