Skip to content

Crucial AI safety breakthrough: 16 major AI companies have agreed to commit to a set of AI safety standards

Crucial AI safety breakthrough

16 major AI companies have agreed to commit to a set of AI safety standards

https://www.aitoolreport.com/articles/crucial-ai-safety-breakthrough

Via AI Tool Report

“”Our Report: At the international AI Safety summit (co-hosted by the UK and South Korea) in Seoul, 16 AI companies—including Amazon, Google, Microsoft, Meta, and OpenAI—have agreed to a set of AI safety measures: the “Frontier AI Safety Commitments.”
🔑 Key Points:
  • By signing the Frontier AI Safety Commitments, each company has agreed to safely develop and deploy their frontier AI models, and publish safety frameworks on how they will measure risks.
  • These frameworks will outline thresholds that will establish when risks are “deemed intolerable”, and what companies will do to ensure they don’t exceed these thresholds.
  • Each company has also agreed to take accountability and not develop or deploy an AI model or system if they can’t keep the associated risks below their outlined thresholds.
🤔 Why you should care: To have so many leading, global AI companies agreeing to the same AI safety commitments is a world first, and it will ensure consistent accountability and transparency during the future development of AI models and systems.”

Via Superhuman AI

“Sixteen AI companies just agreed to make safety a priority. Is it enough?
November’s global AI safety summit was the industry’s equivalent of the Met Gala. It featured an A-list lineup, including tech celebrities like Elon Musk and OpenAI’s Sam Altman. Organizers said there was a point to all the fanfare: It brought more international attention to AI’s potential dangers.
The follow-up event is kicking off in Seoul today, but critics fear that not everyone got the memo. Some countries, including the Netherlands and Canada, aren’t bothering to send representatives, and Google’s DeepMind wouldn’t say if it was attending either. Some argue the lower turnout proves nations and corporations alike have already moved on from AI safety.
Not so fast:
  • Even though the summit isn’t drawing as many high-profile guests, it might end up having a bigger impact in the long run
  • In the first agreement of its kind, sixteen of the world’s biggest AI companies, including Microsoft, OpenAI, Google, and Amazon, agreed on some “red lines” that should never be crossed
  • They promised they’d stop developing any models that become capable of generating large-scale cyber attacks or bioweapons
Governments are working on the problem, too: During one meeting, the US announced it was teaming up with the UK, Japan, and others to build safety institutes that would research AI’s potential risks. Experts point out that it will probably take some time for everyone to align on the same goals, especially because there’s so much geopolitical tension between the world’s superpowers right now. “It will mean nothing if it doesn’t lead to action,” AI ethics researcher Carissa Véliz told Fast Company. “And I think the jury’s still out.””

 

  • Pro plugin deactivated or invalid

Posted on: May 22, 2024, 7:34 am Category: Uncategorized

By: Stephen Abram

Comments Off on Crucial AI safety breakthrough: 16 major AI companies have agreed to commit to a set of AI safety standards

0 Responses

Stay in touch with the conversation, subscribe to the RSS feed for comments on this post.