Peter Wildeford (hiring!) 🇺🇸🚀
@peterwildeford
Globally ranked top 20 forecaster 🎯 Working at @IAPSai to shape AI for prosperity and human freedom. We're hiring! -- https://www.iaps.ai/careers
Now is an even better time to get an AI job than I thought. I updated my post with more jobs.
> 5 gigawatts of Stargate AI data center capacity > over 2 million chips If these are GB200, that would be 5M H100-equivalents. If all used to train one model, that could be 2e28-4e28 FLOP for a six month pretraining + RLVR run. This would be 1000x the compute used for GPT4.…
It's official: we're developing 4.5 gigawatts of additional Stargate data center capacity with Oracle in the U.S (for a total of 5+ GWs!). And our Stargate I site in Abilene, TX is starting to come online to power our next-generation AI research. openai.com/index/stargate…
WH AI Action plan endorses 'location verification' 🔥
In case anyone is curious, the font of the AI Action Plan is Georgia
If you're not currently in a AI Action Plan 'war room' Google Meet right now, you're missing out on the think tank life
Love seeing the AI Action Plan launched with a typical tasteful AI microsite
Trump's "AI Action Plan" just dropped, along with a splashy new website, AI dot gov
Never been a better time to have a 15min walk to work
According to @ConsumerAffairs, the DC area has dethroned Los Angeles for having the worst traffic in the nation. They say the DMV's average commute time is now 33.4 minutes vs. 30.5 in Los Angeles. In 2024 DC was number 2.
My parents trying to explain where I work:
Trump: "He has these think tanks. They build buildings for people that think."
LLMs are haunted 👻
New paper & surprising result. LLMs transmit traits to other models via hidden signals in data. Datasets consisting only of 3-digit numbers can transmit a love for owls, or evil tendencies. 🧵
Institute for AI Policy and Strategy is hiring for a bunch of cool roles. They're open to international candidates and can hire in most countries. iaps.ai/careers
Tired: restaurants that serve breakfast all day Inspired: supersonic flight that keeps you in breakfast time zones all day
At Mach 1.7, Overture outflies Earth’s rotation. So you can have breakfast in London… and fly to New York… for breakfast. And then to San Francisco just in time for breakfast. In fact if you just keep flying west you can keep having breakfast indefinitely.
This AI Action Plan reaction webinar is happening on Thursday and it should be pretty cool!
VIRTUAL EVENT | July 24 | 2:30pm ET Join IAPS for a virtual panel on the Trump Administration's AI Action Plan featuring @MarkBeall of the AI Policy Network, John Fogarty of @BPC_Bipartisan, @jgeltzer of WilmerHale, & IAPS's Jenny Marron. 🔗 Register: bit.ly/3UkrQZj
And they say that Substacks can't be policy relevant 🔥 Congrats @ThomasHochman
.@RepWesterman: “I ask… to submit for the record a Green Tape post titled ‘NEPAstats’”
sometimes line really do just keep going up
"kilowatts per hour" is a real unit. In 2024, we installed solar at a rate of 68,000 kW/h. It's just not the same unit as "kilowatt-hours," of which solar generated 2,129,000,000,000.
In their report this morning Anthropic estimates that for a single training run at the frontier they will require a 2GW data center by 2027, and 5GW by 2028.
It's surprisingly difficult to figure out whether OpenAI and Google DeepMind actually got an IMO Gold "fair and square" or not. Looking forward to more analysis.
Interesting approach! However, we looked at the proofs and methodology and we found a few problems, specifically with the use of hints given to the model. While the scaffold indeed improves performance, it does not solve all problems accurately and would not get a gold medal.🧵
When the the AI turns us all into paper clips it may well be because someone at the Pentagon or in Beijing forgot to add the right special token at the end of a prompt. 🤷♂️
We noticed in community discussions that when using Qwen3-embedding's GGUF models, some developers are not appending the special token <|endoftext|> at the end of the context. This can significantly hurt model accuracy. Check our Model Card (huggingface.co/Qwen/Qwen3-Emb…) for more…
> Amazon deliveries are often better for the environment than driving to buy the stuff in a store It's obvious and true! Just wait until you hear one of the next takes: one of the best things we can do for climate is *reduce* regulations (NEPA reform).
Wrote a quick blog post on the Amazon thing, bugs me how often it comes up in conversations