Graphics Processing Unit (GPU)

GPU not CPU?

You may remember GPUs as the graphics cards that drive computer monitors. This is the same type of processor that drives the AI found in LLMs and are the brains behind the EDR, XDR and Advanced Email Threat Protection that Bryley represents.

Why so? Your computer’s Central Processing Unit (CPUs) is still essential as a decision-maker – CPUs are designed to handle the complex tasks.

But many years ago, it was discovered that GPUs – not built to handle complex tasks, but multiple simultaneous simple tasks – could be tapped to do more than illumine an array of LEDs.

One of the companies that was at the forefront of this discovery is Natick’s MathWorks. MathWorks MatLab software – largely intended to create scientists’ and engineers’ computer simulations – must execute many calculations at once.

Mike Croucher gives a MatLab parallel-computing (which is many computations at once) history, but it also sheds light on the incremental steps that have led to today’s GPU computing … [10 min. read] linkedin.com

Leslie Daigle (that’s not her, above) is after a freely operating internet – a condition that’s more cyber-secure than alternatives – and stopping specific bad actors through cooperation.

One of the good guys

From the top of Vermont, Leslie Daigle watches over the internet.

Daigle is the Chief Technical Officer at the Global Cyber Alliance. She spoke on the Modern Cyber podcast. Daigle’s been working in the field for twenty years.

The strategy she employs involves setting traps for malicious actors and tracing activity back to originating servers so that the hosting companies, cloud service providers and internet service providers can crack down on specific violators. The Global Cyber Alliance’s aim is to preserve the internet’s freedom while stopping specific actors.

The alternative would be to set-up a broader blocking scheme, which though it appears counter-intuitive, can hinder the cooperation needed to fight the attackers.

It’s harder to track threats that cross fragmented borders. Shared intelligence and mitigation efforts suffer. For example, Extended Detection and Response (XDR) depends on live attack information on the internet to shield your organization … [30 min. watch] youtube.com

Holyoke’s MGHPCC is a series of supercomputers used by a consortium of universities, and is comparable in scale to artificial intelligence data centers.

How big is big?

The machines currently occupy about 1.4 acres – that’s like one-and-a-half football fields full of servers.

The facility was built on the site of an old paper mill, and powered by the Connecticut River. It was designed (a dozen years ago) to use 15 megawatts of power – that’s like all the power needed by the residents of the City of Holyoke. When a planned computing renovation occurs, the amount of power usage will be 67 megawatts – enough to power Hartford.

Artificial Intelligence data centers (like the ones used for OpenAI’s ChatGPT or Google’s Gemini) are proprietary and unless you live near one there’s not a lot of readily available, reliable information about how vast they are.

But this massive data center gives a picture. Massachusetts Green High-Performance Computing Center (MGHPCC) is a collaboration of Harvard, MIT, Boston University, Northeastern, UMass, the state and private industry … [6 min. read] masslive.com

Verify

Everybody lies
–Gregory House, MD (well, that doctor bit is a lie)

There are myriad reasons people deceive (they are actors, they want to sell, they’re embarrassed, to steal, etc.). And one of the ways we can see if something’s on the up-and-up is by going through the records.

Stalwart snopes.com talks up stalwart archive.org as a pretty good source of learning if something claimed on the internet really happened. Snopes shows that archived web pages are much harder to manipulate than images or text out of context. Sometimes it pays to do some homework … [6 min. read] snopes.com

Google Search has been an unchanging fact of business outreach a long time now. Are there signs of movement?

Disturbances in the wash

Alphabet reported its earnings (which grew) and showed Google Search and YouTube slowing a bit. Investors, and maybe the rest of business owners, are watching with interest to see how disruptive AI is to search. Google claims now that AI is enhancing its searches on behalf of businesses, especially as concerns shopping (like searching “cheapest laptop to run Adobe Suite,” for example).

A year ago Gartner predicted that search traffic would drop by 25% by 2026 — not far off, so people are looking for indications whether this looks like it will be accurate.

Earlier this year, it was found that chatbot results reflected business’ rankings in Google – the message seeming to be that working on how you rank on Google will have a positive effect on the answers chatbots give. Here’s coverage of the Alphabet report: [4 min. read] bostonhearld.com

^{Note: The section directly above is Bryley’s curated list of external stories. Bryley does not take credit for the content of these stories, nor does it endorse or imply an affiliation with the authors or publications in which they appear.}