Have we finally figured out how to make efficient AI?
AI Today25 Syys 2025

Have we finally figured out how to make efficient AI?

A fantastic research paper published in this month's Nature Computational Science suggests a solution may be in our midst for the incredible inefficiency in generative AI.

Large Language Models' (LLMs) transformer architecture requires the next token (generally part of a word) be predicted based on all the output tokens before it.

Power demands for this process are huge. Shuffling data between memory and processors isn't an easy pipeline, and when you need it to work quickly, those energy demands quickly stack up.

And in an AI arms race, where everyone wants bigger and better models, requiring increasingly powerful compute is required to stretch their limits, the dependence on energy to power, and cool, those processing units grows exponentially.

But what if there was a different, and better, way, to make AI work? That's the driving force behind work of Nathan Leroux and his team proposing a totally different paradigm: analog in-memory computing.

And that's exactly what we're discussing today.

Zip yourself in that flame retardant suit: things are about to get hot in here...

Ping me at dave@wordandmouth.com to get on the show or talk about AI in your world.

Jaksot(90)

DEEP DIVE: Remote work

DEEP DIVE: Remote work

Spotify chiefs say 'work isn't where you are. It's what you do.' While Dell and Amazon say RTO - or GTFO. The COVID-19 pandemic forced companies and employees to work remotely. In the wake of this dis...

15 Loka 202414min

Dario Amodei, Anthropic CEO: Machines of Loving Grace

Dario Amodei, Anthropic CEO: Machines of Loving Grace

On today's AI Today we review an important discussion piece by Dario on the future of this world-changing technology. What's the big idea? The one key message for business leaders from Dario Amodei's ...

15 Loka 202415min

DEEP DIVE: Connecting the bots, with OpenAI's Swarm

DEEP DIVE: Connecting the bots, with OpenAI's Swarm

On the latest episode of AI Today we're connecting the bots with OpenAI's Swarm, a breakthrough framework creating teams of AI agents to complete complex business tasks. Discover how this innovative t...

14 Loka 20249min

Unlocking the Power of Agentic Reasoning: Revolutionising Business with AI

Unlocking the Power of Agentic Reasoning: Revolutionising Business with AI

Imagine AI that can set its own goals and achieve them. That's the power of agentic reasoning, and it's already being used to develop AI lawyers, software engineers, and medical scribes. Discover how ...

14 Loka 202422min

State of AI Report 2024 - the deep-dive!

State of AI Report 2024 - the deep-dive!

This is a big, sprawling listen. So buckle in and get ready for a half-hour journey to the very heart of AI. We're focusing on the State of AI Report 2024 - hundreds of pages diving deep into where we...

11 Loka 202430min

AI Breakthroughs: From Nobel Prizes to Legal Revolution

AI Breakthroughs: From Nobel Prizes to Legal Revolution

AI is reshaping our world at an incredible pace. Tune in to AI Today for a concise overview of the latest breakthroughs, including AI's role in the Nobel Prize in Chemistry and its transformative impa...

10 Loka 202416min

Cheaper, better, faster: data analysis with Anthropic's new Message Batches API

Cheaper, better, faster: data analysis with Anthropic's new Message Batches API

Businesses working with my favourite hosted LLM solution can now save 50% when sending over huge volumes of data for AI-nalysis. This and so much more on today's thrilling instalment of AI Today...

9 Loka 20249min

Google DeepMind's Concordia: AI simulating human behaviour

Google DeepMind's Concordia: AI simulating human behaviour

We regularly cover research papers that hallucinate a different future. Today we're covering code that's already in the world. Sure - there's a research paper ( https://arxiv.org/pdf/2312.03664 ). But...

4 Loka 202410min