“Boston Consulting Group believes that AI and regular data center demand will grow to 7% of total electricity demand by 2030. To put this in context, this is the equivalent of the electricity used for lighting in every home business and factory across the United States. It’s a huge amount of energy. Most traditional data centers that were built 10 years ago were 10 megawatts or less. Today, it’s not uncommon to see 100-megawatt data centers. And with our clients, we’re talking about data centers that approach 1,000 megawatts. And they require 24/7 power. This is something that doesn’t get talked about enough in my opinion.”
That is from Constellation Energy’s CEO Joseph Dominguez (Source: The Transcript) – who of course is talking his own book but still.
Others confirm this, like the IEA – which thinks AI energy demand will double by 2026 – “that’s equivalent to adding a new heavily industrialized country like Germany to the planet”.
There are other huge environmental impacts – e.g. water.
Healthcare is one area where the application of AI, in its LLM and other forms, could be enormous.
This nice article from AlphaSense Expert Insights explores the topic, mirroring the huge rise in expert calls in the sector mentioning the term.
It is not all areas that can be bent to the will of ML. As this piece argues, academic literature and the correspondent knowledge graph is both difficult and not that useful to program.
If you want to read some of these transcripts, you can grab a two-week free trial.
The latest 2023 report is worth a flick (all 160 slides).
This graph, for example, shows the largest Nvidia H100 chip clusters – interesting to see TSLA there, who also run the 4th largest A100 cluster in the world.
Or see Slide 76 which suggests that Nvidia’s advantage (the use of its chips in academic papers) continues to increase.
Interview between Goldman Chair/CEO David Solomon and former CEO/Chair of Google on the future of Generative AI is worth a read.
“In general, the disruption occurs first in the industries that have the most amount of money and the least amount of regulation.”
Pairs nicely with this analysis of the latest batch of Y-combinator companies that are using AI/ML startups (139 in total!) and what areas they are working on.
It is a bit technical but left me with a feeling that though LLMs are a big breakthrough, they have big limitations.
Models beyond the autoregressive LLM that start to mimic some of the planning and reasoning required to rival human intelligence are a lot more complicated with not-so-neat solutions.
In the spirit of Feynman this superb blog post, by none other than Stephen Wolfram, gives a lucid explanation of what is going on under the hood of the latest tech phenomenon.
The short answer is “it’s maths”.
“But in the end, the remarkable thing is that all these operations—individually as simple as they are—can somehow together manage to do such a good “human-like” job of generating text. It has to be emphasized again that (at least so far as we know) there’s no “ultimate theoretical reason” why anything like this should work. And in fact, as we’ll discuss, I think we have to view this as a—potentially surprising—scientific discovery: that somehow in a neural net like ChatGPT’s it’s possible to capture the essence of what human brains manage to do in generating language.”
This dataset tracks the flow of talent in AI around the world.
The chart shows the top 25 institutions for AI research.
“The United States has a large lead over all other countries in top-tier AI research, with nearly 60% of top-tier researchers working for American universities and companies. The US lead is built on attracting international talent, with more than two-thirds of the top-tier AI researchers working in the United States having received undergraduate degrees in other countries.”
