The Promise and Peril of AI Therapy
IBM's new AI chip NorthPole uses an innovative architecture with memory embedded directly within the chip's 256 cores.
Casey Clifton
November 13, 2023
Welcome back! Today is Monday, November 13th and we're looking at Microsoft's AI therapist patent, and the battle between Google, Intel and Nvidia on AI training hardware. New to The Intelligence Age? Sign up here.
Drop us a line anytime at [email protected] with feedback.
News & Insights
Microsoft Patents AI-Powered Therapy App
AI therapy is not uncharted territory, but Microsoft's recent patent filing signals a deeper dive into this complex realm. The AI therapist, built to accumulate a user's emotional triggers and life events like memories, might not be your first port of call in a crisis, but offers a space to vent, seek advice, or even undergo an "explicit psychological test" assessed by a scoring algorithm.
Potential pitfalls? Inevitable. The delicate handling of mental health necessitates AI with a striking degree of emotional nuance. A bot's wrong word or silence could have grave consequences. And what if users form deep bonds, only for the AI to be unplugged? The distress could be real. futurism
Why does this matter?
As AI delves into emotional care, businesses should consider the broader implications. A well-executed AI therapist could pave the way for AI in other sensitive areas, such as grief counselling or trauma support. However, the risk of failure and the potential for harm in these contexts is significant, raising ethical and liability issues that must be robustly addressed.
Google, Intel, Nvidia Battle In Generative AI Training
AI giants Google, Intel, and Nvidia are in a fierce competition within the realm of generative AI training, as revealed through recent MLPerf training tests. Nvidia, with its 10,000 GPU supercomputer, emerged leading while Intel was hot on its tail and Google lagged behind.
Nvidia's triumph was largely due to a system made from its H100 GPUs, specifically the company's new 10,752-GPU AI supercomputer, Eos. Capable of an aggregate 42.6 billion billion floating point operations per second, Eos completed the GPT-3 training benchmark in just under 4 minutes. It's worth noting**,** the task would take even the gargantuan Eos eight days to fully complete its large language model job.
Intel's results showed a promising rise, using the Gaudi 2 accelerator chip and its 4th generation Xeon CPU. The company saw a 103% reduction in time-to-train for a 384-accelerator cluster. Meanwhile, Google's TPUv5e lagged, performing at a third of the speed of an Nvidia system on a per-chip basis. spectrum
Together with Kosmik
Transform Teamwork with Kosmik
Ever lost more time organizing information than actually using it? Meet Kosmik, our sponsor and your new workspace ally.
Capture and collaborate on notes, images, or websites—no more endless searching.
Operates offline, keeping your data secure and local.
An infinite canvas lets you browse the web and centralize resources with your team.
What sets Kosmik apart? It adapts to your workflow while bringing your team onto the same literal page.
Elevate your team's productivity and creativity.
Please support our sponsors!
Mastering AI
Top Tools
AI Audience Accelerator: If you’ve ever wanted to build an audience, AI Audience Accelerator will help you—in a fraction of time time. (website) (sponsored)
notreload: AI service automates web monitoring, delivers stock-moving content. (website)
Postwise: Craft engaging posts, schedule easily, grow followers with AI-powered tools. (website)
Did you know that with no code tools, you can use AI to understand and engage with customers on social media? We recommend Brandwatch's suite of specialized tools for this purpose.
Get access to our full library of AI use cases at: https://portal.theintelligenceage.xyz/diy
Practical AI
Creating Your Own ChatGPT Without Coding — A Step By Step Guide
Harnessing the power of OpenAI, customizing ChatGPT is now within your grasp, no coding required. This is a game changer - a boost in productivity, a tailored tool, and the ultimate way to personalize your AI experience.
"GPTs come to play to help you make a specialized version of ChatGPT, personalized for your needs, and very good at one specific thing, instead of being very generalist." Perhaps it's time to ponder - how could a specialized version of ChatGPT transform your day-to-day operations? towardsdatascience
Why does this matter?
Consider the implications beyond the business world. Could a custom ChatGPT help students struggling with complex subjects, serving as a bespoke tutor, breaking down complicated concepts into digestible information?
Around The World
More News
OpenAI intensifies its game in the Artificial Intelligence sector this week with Generative Pretrained Transformers (GPTs).
Fei-Fei Li revolutionizes AI by developing algorithms that perceive like the human eye.
LLaVA offers a platform to create your own Vision Chat Assistant, enhancing user interaction by integrating AI and visual understanding.
Trending Today
My new custom GPT
ConvertAnything: The Ultimate File Converter! 🔄
Convert images, audio, videos, PDFs & more with ease. Batch uploads, ZIP support, and easy download links included.
A one-stop solution for all conversion needs.
Try now! 👇
— Pietro Schirano (@skirano)
Nov 12, 2023
Research Review
IBM Used AMD's Other Not-So-Secret Weapon To Deliver Yet Another Brain-Like Chip — NorthPole Is Faster And Uses Less Energy Than Nvidia's Best AI GPU And It's Only The Beginning
IBM's NorthPole chip, a decade-long labor, is showing no signs of slowing down, hitting another high with its remarkable benchmarking scores published in Science. The 12nm chip, a product of TrueNorth architecture, outdoes conventional 12nm GPUs and 14nm CPUs in power efficiency by a staggering 25 times. This efficiency was gauged on the ResNet-50 model, factoring in the number of frames per joule of power consumed.
The secret to NorthPole's success? Embedded memory. The chip houses memory within its 256 cores, setting it apart from the typical setup where memory is a separate entity. This unique design bypasses the so-called Von Neumann bottleneck - the delay caused by data traveling to and fro between the CPU and RAM. Hence, NorthPole operates at a speed that leaves even the top-notch GPUs in the dust. techradar
Written by Isaac R. Ward, Casey Clifton, and Alex Brogan.
Send us feedback at [email protected] and help us provide the best coverage of Artificial Intelligence possible.
Interested in reaching smart business leaders like you? To become an Intelligence Age partner, apply here.