Project update + 3 news stories

Today’s project update is a bit technical, so I put it after the news.

No significant news for 3 days.

Today ChatGPT read 1184 top news stories. After removing previously covered events, there are 3 articles with a significance score over 6.8.

[7.2] March 2024 global temperatures hit record high — The Guardian

March set a new global heat record, being 0.1C above the previous March record from 2016, and 1.68C above pre-industrial levels, reported by Copernicus Climate Change Service. This marks the 10th consecutive record-breaking month, temporarily surpassing the 1.5C limit of the Paris climate agreement. Scientists are surprised, attributing factors like El Niño and reduced cooling pollutants, but note these don't fully explain the increase.

[6.9] Google launches global Find My Device network for Android users — TechRadar

Google has launched its Find My Device network globally, starting in the US and Canada. Users can now track offline devices leveraging over a billion Android devices. Google Pixel 8 and Pixel 8 Pro have a unique feature to locate even when shut down or with a dead battery. The network supports Bluetooth tags from Chipolo and Pebblebee, with more brands to be added.

[6.8] Non-invasive CT scans and AI improve heart surgery feasibility — HospiMedica

For the first time, researchers tried a new way to plan and do heart bypass surgery using non-invasive scans and AI to check blood flow. The FAST TRACK CABG study showed this method is safe and works well. It used GE Healthcare’s scans and HeartFlow’s AI blood flow analysis on 114 patients, finding it as good as traditional methods but less invasive. This could change how heart surgery is planned, making it safer and simpler by avoiding direct artery catheterization.

Highly covered news with significance over 6.5:

[6.6] Oral spray vaccine prevents recurring UTIs for up to 9 years
(The Telegraph + 5)

Since the beginning of the project, I’ve used a classic GPT-4 model to score the news, with its knowledge cutoff in September 2021.

This weekend, I updated it to GPT-4-Turbo with a knowledge cutoff in December 2023 — an advancement in historical knowledge of over 2 years.

As a result, scores got noticeably lower.

I haven’t conducted any formal tests, but my guess is that with the old model, ChatGPT was like a person from 2021 time-travelled into 2024 — too many stories were shocking to it. There are two major wars happening? We've mostly beaten Covid? AI is crazy smart now?

Now it’s much calmer. Today we have only 15 stories with rating over 6, while usually there’s about 50.

Let’s see what the new, informed ChatGPT has to show us.

Thanks for reading and see you soon,

Vadim