Calculation of EDGE for tomorrow

One of the biggest fundamental innovations in the last ten years is the emergence of Edge computing as a concept.

In reality, however, we need to narrow the focus of this in about a year and a half, which is the time limit in which the LLMs went from a centralized row to a decentralized ending model.

It took decades for the personal computer to go from a frame to something you can keep in your lap, and then, something you can hold in hand.

It took about a year and a half.

Interested parties began to create large monolithic models that operate in large data center operations.

But we quickly reached the point where you can escalate to a smaller calculation size, and we actually put powerful lubs on the smartphones.

This in turn unlocks all sorts of options that we are just starting to address.

Panel he speaks Edge

I saw a panel in Davos in January, where some experienced business people and researchers went through some of the EDGE computing promises and discussed the technology we have now, which is really quite impressive. Daniel Newman of Moderate Future Group.

Asa Tamson, VP in Ericsson, prompted us not to take our new profits as good and to keep a look in critical processes.

“It is actually strengthening most of our daily lives,” she told network connected technologies. “And when it doesn’t work, it’s a big annoyance.”

Kinuko Masaki, the General Director of Voicebrain, spoke about verifying the processes used for public safety work.

Gill Pratt of Toyota Research spoke about using a vehicle network in the vehicle to create better safety results. He stressed that a car is parked 95% of the time, which offers opportunities for further innovation.

“We need to make sure your vehicle will be safe, even if it is in a form of automated driving,” he said. “This can be aid to driving, where he is trying to prevent a mistake that the human being makes from causing a collision, and is guarding against doing so.”

Lifting the liquid patterns

Sometimes smaller models can better treat specific tasks.

Alexander Amini I Liquid that spoke to us about these new types of networks that are not based on transformer.

He mentioned Yann Lecun’s contribution and referred to a paradigm where LLM can help get network problems.

“Support is problematic,” he said.

(Responsibility: I have acted as a consultant for Liquid he, the company that works in these liquid networks, and I have good colleagues in the Mit Cliail lab who is also doing liquid network research.

Later, Newman asked the panel about large corporate groups, and how Edge computer can help bad battle.

IoT, he suggested, is back, under the flag of “Edge he”.

Cyril of Rockwell Automation spoke about a mistake people make with these new systems: they think it’s just about data. Closed loop actions, noted it, can occur on the edge, with no internet connection, which is a player of the game.

He, he said, can be used either for or against people, so it is important to make sure it is used properly.

These thoughts focused around Edge as a concept. I thought it was sharp to talk about calculating LLM Edge Modern as the child of “Iot” – in some ways, the idea of ​​executing data close to a final point on a massive interconnected network is – just the beginning to be important.