Lightmatter is building AI photonic chips
- potentially the next transistor - with performance 5x faster than an NVIDIA A100 unit on a large transformer model like BERT, while using about 15 percent of the energy.
Also, US companies investing heavily in robotics, sensors startups Aeva and Oculii release new data and raise $55m respectively, while the US Senate will deliberate the U.S. legislative proposal to allocate about $110 billion for basic and advanced technology research and science in the face of rising competitive pressure from China on May 12.
AND IEEE just shared some great AI stories:
In May’s cover story, Microsoft’s AI team argues that if AI systems could better understand human emotions, they might serve us better. And if AI systems are powered by their own approximations of emotions, things might get really interesting.
Michael I. Jordan explains why today’s artificial-intelligence systems aren’t actually intelligent.
AI pioneer says machine learning may work on test sets, but that’s a long way from real world use.
Computers designed to make simple mistakes appear more human, he suggested.
Jeff Hawkins, the Palm Pilot’s creator, thinks AI needs much more neuroscience in the mix.
The Turing Award winner wants AI systems that can reason, plan, and imagine.