Профиль

Been here for a week (Kerry) and the weather has been grand but standard Irish summer grand rather than exceptional

[but with a 99 at least]

Have you come across @vtobin.bsky.social The Element of Surprise?

On t'other, it suggests that a lot of what we flatter ourselves as reflecting individual level human thinking is produced by cultural tropes/heuristics and collective structures for coordinating search across large scale information spaces that can now be compressed and maybe radically improved.

On the one hand, it may not ever be great at truly open ended forms of discovery, which aren't readily testable and aren't obviously amenable to some kind of reinforcement learning type feedback. So "geniuses in a lab" may never turn out to be generalizable. arxiv.org/abs/2607.27191

Can AI agents conduct open-ended AI research? Early evidence from two case studiesForecasts of explosive AI progress hinge on AI agents automating AI research. But evidence on whether agents can carry out open-ended AI research is thin. Current evaluations either test agents on nar...arxiv.org

My own loose bet, which I need to write up one of these days, is that this is Herbert Simon's Revenge. AI deludes us into thinking that this is an equivalent to human type individual intelligence. Instead, it's a massive scaling up of heuristics and collective information processing.

This, via @melaniemitchell.bsky.social is super helpful background on what is going on in background. "“Number one: It works. It improves things ...Number two: the chain of thought “isn’t necessarily faithful to what’s going on ... number three: A lot of that text isn’t even useful."

Is AI Reasoning Right for the Wrong Reasons? | Quanta MagazineThe idea that artificial intelligence can “reason” is more intuitive than ever. But intuitions can be wrong, and the science is far from settled.www.quantamagazine.org

thanks - I had missed this!

isn't it good?

I'd be over the moon too!

Chris Hayes was good on this in a conversation with Ezra Klein before Mamdani was elected - www.nytimes.com/2025/06/28/o... - the whole discussion is good on the experimentation/attention dynamic.

if my 20 year old and his friends are anything to go by, MBV are absolutely still a thing.

Building process knowledge. Regional clusters. Feedback loops from practice. Experimenting with different kinds of award processes. Obviously I would not trust the current lot as far as they could be thrown on implementation of any of this. But as starter ideas for a different administration

I think @mariafarrell.eurosky.social has met her through the Blackwater Valley Anglo-Irish Network?

Obviously you can't and shouldn't separate the report from the administration that produced it, but some of the arguments and ideas could usefully be taken up by a different administration towards different purposes.

In Praise of Blindspots - LPE Project share.google/Ai8XJbl9ADRi...

In Praise of BlindspotsEconomic models produce blindspots, compressing qualitative differences into quantitative measures. Yet, this vice is also the source of their power.share.google

[That should have been an exclamation rather than a question mark ]

Thanks so much?

Also much more selfishly - on my way to Santander right now. What was your rec again for place to wander around?

On the basis of one meeting he seemed like a genuinely lovely bloke who was sincerely interested in people who weren't obviously ever going to be useful to him.