Mastodawn

brunosan Dec 9, 2022

At the same time, this is a clear business case for "air-gapped" or "on-prem" inference services that train on the NDA and (most useful) organic hive mind of your user base using the service. It's a feature, not a bug, but one easily mishandled.

brunosan Dec 9, 2022

To get good results on #gptchat, you usually give context, volunteering personal, IP and NDA stuff. I don't think we realize that our conversations are NOT private. From the ToS, it "will be used to improve the service". GPT trains on your gossips/secrets. https://help.openai.com/en/articles/5722486-how-your-data-is-used-to-improve-model-performance

How your data is used to improve model performance | OpenAI Help Center

Learn more about OpenAI's data usage policies for our API, ChatGPT and DALL-E

brunosan Dec 8, 2022

RT @[email protected]

The photo of the U.S. Air Force's new stealth aircraft is on the left, taken at night, with stars in the background.
We can use them to find the exact location of the jet

(thread) https://twitter.com/usairforce/status/1598858254615912448

🐦🔗: https://twitter.com/johnmcelhone8/status/1600683623250030593

U.S. Air Force on Twitter

“Unveiled today, the B-21 Raider will be a dual-capable, penetrating-strike stealth bomber capable of delivering both conventional and nuclear munitions. The B-21 will form the backbone of the future Air Force bomber force consisting of B-21s and B-52s.(U.S. Air Force photo)”

Twitter

brunosan Dec 8, 2022

RT @[email protected]

🎙️ «@[email protected] aims at developing an instrument of anticipation and action towards future challenges that humanity will be facing by bringing the multilateral policy-making and diplomacy ecosystem closer to the global scientific and #SciDip community» @[email protected]

#WSF_ZA2022

🐦🔗: https://twitter.com/SciTechDiploHub/status/1600918425127497728

SciTech DiploHub on Twitter

“🎙️ «@GESDAglobal aims at developing an instrument of anticipation and action towards future challenges that humanity will be facing by bringing the multilateral policy-making and diplomacy ecosystem closer to the global scientific and #SciDip community» @margagual #WSF_ZA2022”

Twitter

brunosan Dec 7, 2022

RT @[email protected]

In this must-listen podcast episode of @[email protected], @[email protected] sat down with @[email protected] to talk about all things #machinelearning tech for Earth observation & Radiant's role (+ plans) within the ML4EO ecosystem.
🔵Watch the full interview here ⤵️ https://youtu.be/dy3lpWBjt1c

🐦🔗: https://twitter.com/OurRadiantEarth/status/1600505421143367685

Open Machine Learning for Satellite Imagery: Hamed Alemohammad - MBM#34

YouTube

brunosan Dec 6, 2022

Me as investor to a female startup CEO: "Tell me your problems and I'll tell you what I see, what I did in a similar situation, why it might work and why not. But in any case make your own decisions and I'll support you".
Other investors: https://t.co/MwORIlU1T4

Bruno Sánchez-Andrade Nuño on Twitter

“Me as investor to a female startup CEO: "Tell me your problems and I'll tell you what I see, what I did in a similar situation, why it might work and why not. But in any case make your own decisions and I'll support you". Other investors:”

Twitter

brunosan Dec 5, 2022

RT @[email protected]

This is amazing. This email is from a marketing person offering sources to discuss how European space startups are surpassing the United States.

🐦🔗: https://twitter.com/SciGuySpace/status/1599837998337638402

Eric Berger on Twitter

“This is amazing. This email is from a marketing person offering sources to discuss how European space startups are surpassing the United States.”

Twitter

brunosan Dec 5, 2022

🚨When using #GPT like in #gptchat, remember that the @[email protected] authors explain that it makes fake facts 2/3 of the times when tested. Fact checking is not part of the loss function. https://openai.com/blog/webgpt/

WebGPT: Improving the Factual Accuracy of Language Models through Web Browsing

We've fine-tuned GPT-3 to more accurately answer open-ended questions using a text-based web browser. Our prototype copies how humans research answers to questions online—it submits search queries, follows links, and scrolls up and down web pages. It is trained to cite its sources, which makes it easier to give

OpenAI

Show thread

brunosan Dec 5, 2022

but anyway it doesn't matter, in a couple of years we'll all just voice the intent to Whisper and then GPT will write the code in whatever language 🤷‍♂️

brunosan Dec 5, 2022

I am professional #DataScientist and I've never used R. #confession Yes I should, but also yes it's fine if you don't. IMO in the "real world"™️ knowing python is most useful most of the time, then Excel (yes, Excel well), and far lower SQL, shell tools, then R. 🤷‍♂️

Impact Science	https://impactscience.dev/
Personal	https://brunosan.eu