Patrick J.

Building production geospatial AI for insurance. Zurich & Lausanne. ETH Zürich CS alum. Views my own.

[Patrick J. @xrb]

Prithvi running in orbit is a quiet but important GeoAI milestone. Less latency, fewer downlinks, more on-board filtering for floods, fires and other events.
Foundation models for Earth observation are moving closer to the sensor. #GeoAI

[NASA Marshall @NASA_Marshall]

NASA's Prithvi has become the first geospatial AI foundation model in orbit! 🛰️

This model, which was developed by NASA’s IMPACT AI team at #NASAMarshall, could improve satellite observations of floods, wildfires, and more.

Read the full story: https://t.co/YMXYrj9uM0

[Patrick J. @xrb]

Even now, I'm a strong proponent of augmentation over automation in coding and AI in general. Iron Man suit, not Iron Man robot. Fully with @karpathy on this one. #vibecoding #AI

[Rohit @rohit4verse]

Andrej Karpathy at YC AI Startup School:

"build Iron Man suits, not Iron Man robots."

i've been watching builders ship and builders stall. the shippers are still coding. they wear the suit. they direct a team of agents and the keys stay in their hands.

the stalled ones became the robot. they only prompt now. the muscle that catches when the model is wrong has gone quiet.

there is no neutral way to use AI. you either get sharper or you get hollower. most are getting hollower and won't notice till the chat won't open.

how to stay in the first group, inside:

[Patrick J. @xrb]

@thatguybg Up for it!

[Patrick J. @xrb]

GPT-5.5 and Claude Opus 4.7 are both pointing in the same direction: less prompt magic, more outcome delegation. For teams, the question is shifting from “which model is best?” to “which tasks deserve the expensive model?” #openai #anthropic #gpt55

[TechCrunch @TechCrunch]

OpenAI releases GPT-5.5 Instant, a new default model for ChatGPT https://t.co/p9B3o5XlMt

[Patrick J. @xrb]

What happened to #social media?

↗ Promoted

[Patrick J. @xrb]

I am seeing lots of such messages on Reddit. Drawback of current #AI development is that humans (e.g. teachers) are not well informed about how to detect AI-generated content. They are using latest shit tool as if it was reliable. #ChatGPT

[Patrick J. @xrb]

This is what happens when a zen master challenges a large language model on #consciousness. #llm #gpt

https://t.co/NFn1vLVa3y

[Patrick J. @xrb]

Current status. #chatgpt https://t.co/ILDyCV0ro4

[Tansu Yegen @TansuYegen]

1988: Math teachers protest against calculator use.
2023: Teachers protest against ChatGPT use.
via @HeyImCastro

[Patrick J. @xrb]

Cute example with #GPT4

https://t.co/haE5ouN5gI

[Patrick J. @xrb]

To my appreciation, this is already past human level. Even a good computer scientist would not be able to write down a solution to a random question without looking it up.

The backend used was #GPT4.

[Patrick J. @xrb]

Pre-#ChatGPT looking for a solution to a programming error: google the error message + a few keywords. Browse the results from Stackoverflow until there is a post that has a relevant solution. Try it out and test.

[Patrick J. @xrb]

No lookup to #Stackoverflow or any documentation. These are all internal knowledge stored in the model weights.

[Patrick J. @xrb]

Current state of #AI development.

[Future of Life Institute @FLI_org]

📢 We're calling on AI labs to temporarily pause training powerful models!

Join FLI's call alongside Yoshua Bengio, @stevewoz, @harari_yuval, @elonmusk, @GaryMarcus & over a 1000 others who've signed:
https://t.co/3rJBjDXapc

A short 🧵on why we're calling for this - (1/8)

[Patrick J. @xrb]

Just the multimodality with #ChatGPT model performance would be a game changer.

[Patrick J. @xrb]

Just read (quickly) GPT-4 paper. Impressive staff. Solid improvement over GPT 3.5 and ChatGPT. Lost of comparisons on benchmarks presented, including a whole collection of exams designed for humans. https://t.co/smuXSLzwHo #gpt4

[Patrick J. @xrb]

Is it AGI? Probably not (fully). But who cares. Though top-10% on the bar exam without learning is massive (as a single example).

Will it be a game changer? Surely! As has been #ChatGPT 3.5 months ago (my goodness, how fast).