Long Form Essays, EU Regulations and Llama 2 | On my Feeds #4
And other worthy resources that I read this week.
Welcome to this week of On My Feeds! In this section, I keep you up to date on things happening around the AI, Tech and Science that you need to know.
It’s been a while since I have done it. So, let’s jump right in without any wait!
Highlights
Long-form Content
Lately, I have been writing long-form content on Creative Block (CB) that spans around 200 words. This is different from other posts you see on CB. I wanted to try my hands on detailed coverage and analysis - something which I usually don’t do on CB.
Though writing and editing long-form essays takes longer time, I enjoy it. This is because - firstly, it allows me to cover the topics in much detail, and secondly, long-form essays make full use of the research.
Of course, it takes a higher attention span (especially a lot more than average social media users) to read long-form content. However, the insights that the long-form offers tend to counterweight this flaw.
Also, as opposed to short-form writing - where you (typically) try NOT to write more than 800 or 1000 words - long-form essays don’t have this constraint.
You keep on writing till you cover the central point as in whole. Thus, Idea rule over the word count in long-form essays.
Moreover, Long Form essays allow me to venture far and beyond, covering all the subtopics. This adds value to the post in my opinon.
Here are some of the long-form posts that I posted this month -
I.
An essay on the EU’s AI act - a brief overview (ironically around 2000 words) and discussion on possible loopholes in the AI act.
II.
A small yet big update to the CB - change of the logo (with minor changes in formatting).
Plus, the story behind the previous logos of CB.
III.
My personal favourite post of this month. The research behind this post was the most fun to do. The essay goes over, asking the question - “Can AI be considered as an Intelligent Species?
If yes, did it overcome the Great Filter? or we are yet to see one?”
EU Commission to make it easier to sue over AI products (Liv McMahon, BBC)
“The European Commission has proposed new rules to help people harmed by products using artificial intelligence (AI) and digital devices like drones.”
“The AI Liability Directive, published by the European Commission on Wednesday, will introduce a "presumption of causality" for those claiming injuries by AI-enabled products.
This means victims will not have to untangle complicated AI systems to prove their case, so long as a causal link to a product's AI performance and the associated harm can be shown.”“The proposal will ensure that when AI systems are defective and cause physical damage or data loss, it's possible to seek compensation from the AI-system provider or from any manufacturer that integrates an AI system into another product.”
Introducing Llama 2 (Meta)
“Llama 2 is available for free for research and commercial use.”
“This release includes model weights and starting code for pretrained and fine-tuned Llama language models — ranging from 7B to 70B parameters.”
“The latest powerful model has been tipped to go up against OpenAI’s GPT-4. In a Facebook post announcing the launch, Zuckerberg said that Meta has a long history of open-sourcing its infrastructure and AI work – from PyTorch, the leading machine learning framework to Segment Anything, ImageBind, and Dino. He said that open-sourcing has helped Meta build better products and drive progress across the industry.”
European Data Protection Board Launches ChatGPT Task Force (Jody Serrano, Gizmodo)
“The EDPB announced its new initiative in a brief, two-sentence statement in a press release on Thursday, the same day the Spanish Data Protection Agency, known as the AEPD in Spanish, stated it was launching a preliminary investigation into OpenAI over possible privacy violations by ChatGPT.”
“The Spanish regulator joined Italy, which became the first country in the world to ban ChatGPT, albeit temporarily, in scrutinizing the chatbot.”
“The EDPB members discussed the recent enforcement action undertaken by the Italian data protection authority against Open AI about the Chat GPT service,” the board said in its press release. “The EDPB decided to launch a dedicated task force to foster cooperation and to exchange information on possible enforcement actions conducted by data protection authorities.”
Evaluating the ability of ChatGPT and other large language models to detect fake news (Ingrid Fadelli, Techxplore.com)
"We evaluated the performance of these large language models using a test suite of 100 fact-checked news items from independent fact-checking agencies,"
"We presented each of these news items to the models under controlled conditions and then classified their responses into one of three categories: True, False, and Partially True/False. The effectiveness of the models was measured based on how accurately they classified these items compared to the verified facts provided by the independent agencies."
"We found that OpenAI's GPT-4.0 outperformed the others, hinting at the advancements in newer LLMs. However, all models lagged behind human fact-checkers, emphasizing the irreplaceable value of human cognition.”
The EU Urges the US to Join the Fight to Regulate AI (Paresh Dave, WIRED)
“Didier Reynders told WIRED why the US needs to finally step up, where a probe into ChatGPT is headed”
“Reynders has heard a series of proposals for how the US could follow suit, including from Meta CEO Mark Zuckerberg and other tech executives, Facebook whistleblowers, and members of Congress and federal officials. But he says there has been no ‘real follow up.’”
“Now Reynders fears history is repeating with AI regulation, leaving this powerful category of technology unchecked”
“‘If you have a common approach in the US and EU, we have the capacity to put in place an international standard,’ Reynders says.”
(Link)
(Link)
(Link)
(Link)
(Link)
Got interesting tweet that I can add here? Tweet it at me - @its_aditya_an1l
That’s it for this On My Feeds Post! Hope you enjoyed it.
Have your own suggestions? Send it in the comments, or drop me a mail
See you in next post!
