4 ms·
Analog will likely come back but for other reasons: Neural networks don't require precise calculations and Hintons forward forward networks put into hardware wo
by singularity2001 4y ago
Analog will likely come back but for other reasons: Neural networks don't require precise calculations and Hintons forward forward networks put into hardware would be several orders of magnitudes more efficient, even without photons. "AI inferencing is heavily dependent on multiply/accumulate operations, which are highly efficient in analog."
If you know of any startup working on this let me know because I'd love to join the revolution.
- frafra 4y agoAlmost one year ago, Veritasium interviewed https://mythic.ai/ https://mythic.ai/ in one of his videos on YouTube: https://youtu.be/GVsUOuSjvcg?t=898 https://youtu.be/GVsUOuSjvcg?t=898.
- Drunk_Engineer 4y agoThey ran out of money last year.
- denton-scratch 4y agoNeural nets, AFAIAA, usually rely on the digital simulation of analog processes. NN weights, inputs and outputs are all continuous values. I suspect an integrated all-analog NN chip will be along soon, and might run more efficiently and faster than the digital simulation. I don't see why it shouldn't use photonics.
- brianderson 4y agoCoincidentally, I recently posted a collection of neuromorphic job openings to LinkedIn: https://www.linkedin.com/posts/brian-anderson-6739ba25_neuromorphic-machinelearning-jobs-activity-7025813029011300352-sClP https://www.linkedin.com/posts/brian-anderson-6739ba25_neuro... My team at Intel Labs is hiring, as are a number of well-funded startups.