Macchine desideranti
open.substack.com/pub/salvofed...

L'intramoenia all'inglese: guida per la comunità italiana in UK
open.substack.com/pub/salvofed...

Gemini Deep Research has gotten very good since it was upgraded to 2.5 Pro

Claude & ChatGPT tend to go a good job acting like analysts and building an argument, but Gemini is probably the best right now at synthesizing a coherent overview of a complex topic. (Error rates also have become quite low)

An enthralling conversation with @adamjkucharski.bsky.social on uncertainty in science and his new book PROOF, with some emphasis on things we learned from Covid, and much more.
video, audio, transcript; all open-access
erictopol.substack.com/p/adam-kucha...

le AI come l'elettricità o il web: "tecnologia con tempi di diffusione lenta", un punto di vista diverso da quello che @emollick.bsky.social racconta nel suo libro. Nel testo qualche esempio interessante, per esempio il fallimento nella realtà assistenziale USA dell'Epic Sepsis Predictive Model

The whole thing about buying Greenland is because Republicans *believe* in climate change.

I mean like really believe in it.
They want it. They love the idea of it.

They're excited about the world burning and the north melting.

They're practically counting on it.

After using them both, I think that Gemini 2.5 & o3 are in a similar sort of range (with the important caveat that more testing is needed for agentic capabilities)

Each has its own quirks & you will likely prefer one to another, but there is a gap between them & other models

In 2024, Paris banned cars from its city center to make room for cyclists and pedestrians with remarkable pollution reduction as a result. Earlier this year, Parisians voted to create even more car-free zones in the city.

Visual: The Washington Post

I tried the experiment of asking AI for a website of the Catalog of Ships from the Iliad two years ago & it sort of worked, with a lot of little hiccups

Now, I just gave Manus a screenshot of the tweet about it and said "do this project" and it did the research and I got a working website, one shot

Un Tutor AI per chi vive (e cura) la complessità
open.substack.com/pub/salvofed...

Ho realizzato un Tutor AI dedicato all'addestramento del paziente, dei suoi familiari e dei medici di varie sub-specialità che lo seguono su una malattia rara e a tre sue comorbidità. È finalizzato a spiegare la malattia e le motivazione che hanno indotto alla scelta di aggiungere un farmaco 1/n

I've shared this tip before but it comes up a lot:

One of the safer options to help your kids learn with AI is for YOU to use it. Take a picture of a problem and ask the AI to explain it to you "in a way that I as a parent can explain to my _ grade student." You do the coaching, don't give answers.

None of this is the fault of trainers - I have taught my students how to do Chain-of-thought, etc. But we need to start to think about how to teach people to use AI in a world that is changing quite rapidly. Focusing on exploration and use, rather than a set of defined rules.

My suspicion right now, without a lot of concrete evidence other than a bit of experience, is that o3 (powering OpenAI's version) results in smarter analysis, but that the much broader search and summation ability of Gemini results in a better wide overview.

Use case will matter in your choice.

The updated Google Deep Research is really good based on my initial tries

It is much more obviously smart & agentic in its research progression, while still casting the widest net of any of the Deep Research tools. It is becoming clear that more powerful models leads to better agents for research.

Really in-depth paper on AI hallucinations in medicine, with lots of discussion and analysis about addressing them & what is appropriate for medicine

But I found this bit on how much more
accurate the latest models have gotten to be interesting (though more work is needed) arxiv.org/pdf/2503.057...

Acute hepatitis of unknown aetiology in children: evidence for and against causal relationships with SARS-CoV-2, HAdv and AAV2 bmjpaedsopen.bmj.com/content/8/1/... @dgurdasani1.bsky.social

Nei giorni scorsi, un account WhatsApp sconosciuto mi ha contattato "ringraziandomi" per la partecipazione all'elezione del presidente dell'Ordine dei Medici di Palermo e della sua lista (rappresentante per i pediatri: il Dr. Sannasardo). 1/10