• AI Periscope
  • Posts
  • New chip runs on light waves, not electricity

New chip runs on light waves, not electricity

And: Quantum-enhanced GenAI for cancer drug discovery | Groq creates fastest generative AI in the world

Chip research

Summary - Engineers at University of Pennsylvania have developed a revolutionary silicon-photonic (SiPh) chip that runs on light waves to perform complex mathematical computations. This innovation promises to drastically increase computational speeds for heavy AI loads, and significantly reduce energy consumption. We covered the rising energy consumption needed by AI processing here.

Key points:

  • Speed and Efficiency: The chip uses light waves for computations, “allowing the chip to perform mathematical calculations at the speed of light.” Many computations can occur simultaneously.

  • Innovative Design: The silicon design used in the chips is 150 nanometers, about 1/300th the thickness of human hair.

  • Commercial Viability: The chip's design is compatible with current commercial production processes.

  • Enhanced Privacy: Simultaneous processing eliminates the need for working memory, thus no storage of sensitive data.

  • Collaborative Effort: The research was a collaboration between the engineering groups led by U. Penn professors Nader Engheta and Firooz Aflatouni.

  • Funding Support: The project was supported in part by a research grant from U.S. Air Force Office of Scientific Research and the Office of Naval Research.

POV - This could be a potential paradigm shift in computing. As mentioned, with so much concern lately of the high energy consumption that AI is projected to need for LLM model training, and the purported current commercial viability of the SiPh chip due to the fact that production processes already exist for this design, could this be another huge leap forward in AI?

Quantum computing & cancer research

Summary - Zapata AI has successfully used quantum-enhanced generative AI to identify new cancer drug candidates. This marks the first occasion where quantum generative AI models outperformed traditional computer hardware to find viable cancer drug candidates. The research focused on creating inhibitors for the historically challenging KRAS protein, with the quantum AI producing molecules that showed superior binding characteristics compared to those from classical AI models.

Key points:

  • This research was the combined efforts of Zapata AI, Insilico Medicine, University of Toronto and St. Jude Children’s Research Hospital.

  • Generated a million drug candidates, which were then screened by algorithms and humans.

  • The result produced 15 molecules, which were then chemically tested, yielding 2 viable drug candidates.

POV - First, chips running on light waves and now quantum computing AI for drug discovery. Pretty amazing world of science and engineering we live in. Will we see the fruits of these research efforts any time soon?

Groq

Summary - Groq, founded in 2016 in California, has developed a groundbreaking AI model that is setting new standards for speed and efficiency in the world of AI. Unlike traditional AI models that rely on graphics processing units (GPUs), Groq's model utilizes its own custom-designed chip called the Language Processing Unit (LPU).

Key points:

  • LPUs are optimized for large language model (LLM) processing.

  • Capable of processing a blazing 500 tokens per second.

  • Groq has been focused on speed to achieve real-time interactions.

  • Generally runs 10x faster than equivalent systems running on GPUs.

  • Similar name to Elon Musk’s Grok, but no relation.

  • Groq sent Elon a cease-and-desist letter after Grok was announced.

POV - The legal scuffle of Groq vs. Grok is an interesting layer of drama that we’ll continue to watch. Groq does seem pretty fast. Check it out at Groq.com.

Don’t forget to check out other AI headlines in The Ocean.

The Ocean is designed to help you stay informed with the wide array of AI headlines, giving you just the snippet of news that you need if you are short on time, but also gives you one-click ease if you want to dive deeper.