Martin Ruskov @mapto@social.cleverlibre.org

RT @ProfNoahGian@twitter.com

Next point I LOVE: "Drug companies cannot sell people new medicines without first subjecting their products to rigorous safety checks. Biotech labs cannot release new viruses into the public sphere in order to impress shareholders with their wizardry. Likewise, A.I. systems ..."

🐦🔗: https://twitter.com/ProfNoahGian/status/1639698816482852865

RT @ProfNoahGian@twitter.com

When I see that X% percent of "AI experts" believe there's a Y% chance AI will kill us all, sorry but my reaction is yeah that's what they want us to think so we are in awe of their godlike power and trust them to save us. It's not science.

🐦🔗: https://twitter.com/ProfNoahGian/status/1639698814503202818

RT @ShannonVallor@twitter.com

The most depressing thing about GPT-4 has nothing at all to do with the tech. It’s realising how many humans already believe they are merely a machine for generating valuable word tokens

🐦🔗: https://twitter.com/ShannonVallor/status/1639631757002547202

@bogo не знам, сигурно се разминаваме в представите си коя лудост минава отвъд

България официално е полудяла. Видях само едно заглавие, което още на първо четене не ми изглежда абсолютно безумно. Ще успеете ли да го откриете?

RT @ZelenskyyUa@twitter.com

Zaporizhzhia. Right now, residential areas where ordinary people and children live are being fired at.
This must not become "just another day" in 🇺🇦 or anywhere else in the world. The world needs greater unity and determination to defeat Russian terror faster and protect lives.

🐦🔗: https://twitter.com/ZelenskyyUa/status/1638498087051116545

RT @emilymbender@twitter.com

Apropos #openai refusing to disclose any information about the training data for #GPT4 and #Google being similarly cagey about #Bard...

From the Stochastic Parrots paper, written in late 2020 and published in March 2021:

w/@timnitGebru @mmitchell_ai@twitter.com @mcmillan_majora@twitter.com

🐦🔗: https://twitter.com/emilymbender/status/1638569048274513922

My parents are born in the 1950s, I'm born in the 1980s, my son is born in the 2020s. This graphics is about me. And I want my son, his cousins, their friends, other kids of their age to be able to live in natural conditions. Not chained to an air conditioning, forced into a car

RT @AlexRuane@twitter.com

Very proud to have led team developing #IPCC SYR Figure SPM.1, released today, which shows how current and future generations have different experiences with climate change. 1/4

🐦🔗: https://twitter.com/AlexRuane/status/1637810764411961344

RT @AlexRuane@twitter.com

Very proud to have led team developing #IPCC SYR Figure SPM.1, released today, which shows how current and future generations have different experiences with climate change. 1/4

🐦🔗: https://twitter.com/AlexRuane/status/1637810764411961344

RT @mapto@twitter.com

@SigalSamuel@twitter.com Well, it's not really about slowing down AI, is it? I think it is much more about quality of delivery. It is about studying the implications before delivering the product. But companies are notoriously bad at this. Other social organizations exist for this purpose

🐦🔗: https://twitter.com/mapto/status/1638042483253813253

@boris_steipe it's clear we'll not agree, but I struggle to comprehend what though process would lead to the apparent confidence of saying "ChatGPT embodies Paul Grice's "cooperative principle". Is there some sort of a widely agreed verification process that it has passed, or is this based on some personal observations?

@boris_steipe sorry, let me try to explain better. This is all about intent. Is the objective that we use technology to explain things to us, or do we want to be able to explain things to technology in a way that it can reproduce basic questions whose responses have no added value, and arguably are reduced to fact checking? I'm interested in the former, the latter is what I see in your guiding simplifications. Maybe you meant it differently?

Remember from school that facts are the mostly basic type of knowledge. The more valuable ones are functional and others that are widely called conditional, but as I hinted in my article, are very diverse.

@boris_steipe you've simplified the syntax to strip away the semantics and you got the correct answer. Yes, I confirm this is the case.

@boris_steipe here's another example: ChatGPT got me confused about my family. It's in the counting and the greats that https://en.wikipedia.org/wiki/Rote_learning fails. See below:

mapto:
Polihron is the grandfather of Rusko. Rusko has a grandson who is also called Rusko. This child is my grandfather. What is my relationship to Polihron?

ChatGPT:
If Rusko (the grandson) is your grandfather, then Polihron must be your great-great-grandfather.

Here is the reasoning:

Polihron is the grandfather of Rusko (the first generation).
Rusko (the first generation) has a son who has a son who is also named Rusko (the third generation, your grandfather).
Therefore, Polihron is the great-grandfather of Rusko (the third generation and your grandfather).
Since Polihron is the great-grandfather of your grandfather, that makes Polihron your great-great-grandfather.

mapto:
And what is my son Lazar to Polihron?

ChatGPT:
If Polihron is your great-great-grandfather, then your son Lazar would be Polihron's great-great-great-grandson.

Here is the reasoning:

Polihron is your great-great-grandfather (the first generation).
You are one generation away from Polihron (the second generation).
Your son Lazar is two generations away from you (the third generation).
Therefore, Lazar is three generations away from Polihron, making him Polihron's great-great-great-grandson.

@simon what I find interesting is to understand whether algorithms are bad at URLs (to put the problem short) or ourselves better at catching them there. Could we try to generalise the problem? A URL is a reference and it's used when one refers to a related piece of text. Turns out algorithms are able to identify when a reference is due, but extremely bad at finding the correct reference. Would such generalisation make sense to you?

RT @jamesbridle@twitter.com

"It has read most of the internet, and it knows what human language is supposed to sound like, but it has no relation to reality whatsoever." I'm in the @guardian@twitter.com today, writing about image generators, ChatGPT, and the stupidity of AI: https://www.theguardian.com/technology/2023/mar/16/the-stupidity-of-ai-artificial-intelligence-dall-e-chatgpt

🐦🔗: https://twitter.com/jamesbridle/status/1636270610643836928

@simon we don't even need to consider adversarial attacks. Hallucinations are a big enough problem already. The other day someone "read" my article using Bing Chat. It took us several exchanges to figure out that the criticism she was sharing with me wasn't about the actual text, but about something Sydney (or whatever hallucinated name you prefer) decided to insert to its interpretation

@philipncohen since you've worked in the material and came up with the questions, what are the chances it engages in cramming (i.e. https://en.wikipedia.org/wiki/Rote_learning)? If so, he would know the answers because he's previously repeated them many times, hopefully with some variations