Deep learning is hitting a wall | Curio
News and insights read out for you.
10,000+ audio articles. 50+ world-leading publications.

All in 1 subscription.

Nautilus logo

Deep learning is hitting a wall

33 mins | Mar 23, 2022
story image
What would it take for artificial intelligence to make real progress? One minute it's revolutionising medicine, the next, we're on the verge of developing algorithms with the capacity for logic and natural conversation, yet none of this has materialised. Essentially, artificial intelligence has promised us the moon and only occasionally delivers. That's not to say it will never happen. However, no matter what tech entrepreneurs like Elon Musk say, we're still a long way from machines that can genuinely understand human language. Turning the tide and getting to AI we can trust isn't easy. However, it's not impossible. Deep learning is fundamentally a technique for recognising patterns and works best with low stakes like photo tagging or text prompts as it's prone to error. However, when the stakes are higher such as in radiology or driverless cars, we must be more cautious. Founder and CEO of Geometric Intelligence, Gary Marcus, discusses the history of AI, the challenges in ethics and computation, and the vast responsibility resting on its shoulders.
Get unlimited access free for 7 days, then $6.67/month (billed annually)
Get started