The AI race has lengthy been framed as a zero-sum recreation: both the US or China will win on the finish. However as issues pile up across the growing capabilities of AI fashions—particularly AI brokers—researchers in each international locations are attempting to staff as much as work on AI security. This week, Zoë Schiffer speaks with senior author Will Knight about what he noticed and heard on the bottom when he visited China this summer season—and why the 2 international locations may really want to start out working collectively to keep away from a serious AI disaster.
That is our second episode from our summer season break collection. We’ll be again subsequent week with our common roundtable.
Articles talked about on this episode:
You may observe Zoë Schiffer on Bluesky at @zoeschiffer and Will Knight on Bluesky at @willknight. Write to us at [email protected].
Tips on how to Hear
You may at all times hearken to this week’s podcast by means of the audio participant on this web page, however if you wish to subscribe without spending a dime to get each episode, this is how:
For those who’re on an iPhone or iPad, open the app referred to as Podcasts, or simply faucet this link. You can too obtain an app like Overcast or Pocket Casts and seek for “uncanny valley.” We’re on Spotify too.
Transcript
Be aware: That is an automatic transcript, which can include errors.
Zoë Schiffer: That is WIRED’s Uncanny Valley. I am Zoë Schiffer, contributing editor. For those who’ve been following tech information this summer season, and positively in case you’ve been listening to the present, you most likely already know that China has been within the headlines quite a bit, notably in the case of the AI race. The nation’s open fashions proceed to shut the hole with US frontier fashions at what some folks estimate is a fraction of the price. In flip, the US has maintained its tight restrictions on chips and export controls to decelerate China’s rise. We consider AI development in so some ways as zero-sum, if China wins, the US loses, and vice versa. However there is a concern that actually stretches throughout these strains, and it is about AI security.
Archival audio: AI cybersecurity dangers have been high of thoughts after a number of cases of AI brokers from each OpenAI and Anthropic breaking out of their enclosures.
Zoë Schiffer: Information of AI agents hacking platforms added urgency to this situation over the summer season. In flip, authorities officers have been pressured to concentrate and act on AI regulation.
Archival audio: President Trump has signed an government order asking tech firms to offer the federal government oversight of latest AI fashions earlier than their public launch.
Zoë Schiffer: So, might the US and China truly profit from working collectively? And what wouldn’t it take to even make that occur? Earlier this summer season, WIRED’s senior correspondent Will Knight visited China to get some solutions. Will Knight, thanks a lot for being right here.
Will Knight: Thanks for having me.
Zoë Schiffer: OK, so I need to begin along with your journey. You went to China earlier this summer season and on the time, we weren’t listening to as a lot about AI security within the US. In truth, it felt like with Trump’s second time period within the White Home, it was like a, the US must win framing, and AI security virtually began to sound like anti-growth. However I am curious what you have been listening to and seeing in China on the AI security entrance.
