Assembly theory explains and quantifies selection and evolution
"This approach enables us to incorporate novelty generation and selection into the physics of complex objects. It explains how these objects can be characterized through a forward dynamical process considering their assembly... It discloses a new aspect of physics emerging at the chemical scale, whereby history and causal contingency influence what exists..."
from Nature
https://www.nature.com/articles/s41586-023-06600-9
"Any large language model will be chock full of forbidden knowledge: how to make bombs, commit genocide, rob a bank, create a revolutionary vanguard and overthrow a government, etc. All that information has been fed into these AI chatbots during their lengthy training. As that process focuses on quantity over any particular quality of information, it can be assumed that pretty much everything we know - good & very, very bad - sits inside the largest of these models." https://windowscopilotstrategies.substack.com/p/two-weapons-grade
Very happy to announce the inaugural edition of the Windows Copilot Newsletter - a curation of some of the most important happenings in the rapidly exapanding field of AI chatbots.
This week, Microsoft releases Windows Copilot - and Bing Chat is being used to deliver malware-laden ads. Plus much, much more.
https://windowscopilotnewsletter.substack.com/p/september-29th-2023-windows-copilot
In a seismic announcement that brings AI chatbots to _three billion_ people, Meta is officially entering the AI chatbot wars, starting with its own assistant and a slew of AI characters it’s releasing in WhatsApp, Instagram, and Messenger.
http://windowscopilot.news/2023/09/28/meta-is-putting-ai-chatbots-everywhere/
Among the very pointed & immediately actionable suggestions @7a4ff4a2 made at yesterday's USyd conference on Generative AI was to seek _complete_ transparency on pay and conditions for the millions of workers in the Global South who are tasked with "Reinforcement Learning from Human Feedback" - i.e., dealing with the very worst outputs of Generative AI systems during their training to ensure they're not completely toxic upon release. First transparency, then equity.
https://www.sydney.edu.au/engage/events-sponsorships/sydney-ideas/2021/kate-crawford-atlas-of-ai.html
"Some of the things that Copilot can help you do are:
- Control settings on your PC, such as changing the volume, theme, wallpaper, bluetooth, etc. by using natural language commands.
- Launch apps on your PC by simply asking Copilot to open them for you.
- Enhance your creativity with AI-powered features in apps like Paint, Photos, Clipchamp and more..."
WHAT COULD POSSIBLY GO WRONG
🙄
@957492b3@7906d04d Tuning is moving even faster than the LLMs are these days. LoRA/QLoRA means it will happen on-device; but that means the device has to be collecting the data to use for fine-tuning. Which is a whole other way of thinking about what the device is doing.
"...adjusting for age, we experienced what seemed to be normal death rates in 2000 as essentially unthinkable [during COVID-19] in 2020. The result: a 20-year setback that felt more like a once-in-a-century, or even unprecedented, social rupture... It took an astonishing reversal to reveal astonishing gains we’d come almost entirely to take for granted."
https://www.nytimes.com/2023/09/13/opinion/covid-19-pandemic-deaths.html
"Personal large language models – running privately, on device, all the time – will soon be core features of smartphone operating systems. They'll suck in all your browsing data, activity and medical data, even financial data – all the data that today we hand off to the cloud to be used against us – and they will continuously improve themselves to represent more accurately our states of mind, body, and finances."
Latest in @0d6c6ac4https://www.theregister.com/2023/09/13/personal_ai_smartphone_future/
Earth beyond six of nine planetary boundaries
"Currently, anthropogenic perturbations of the global environment are primarily addressed as if they were separate issues, e.g., climate change, biodiversity loss, or pollution. This approach, however, ignores these perturbations’ nonlinear interactions and resulting aggregate effects on the overall state of Earth system..."
https://www.science.org/doi/10.1126/sciadv.adh2458
"...nothing that I have read, verified or done gives me any compelling reason to believe that LLMs do reasoning/planning as it is normally understood. What they do, armed with their web-scale training, is a form of universal approximate retrieval which, as we have argued, can sometimes be mistaken for reasoning capabilities."
https://cacm.acm.org/blogs/blog-cacm/276268-can-llms-really-reason-and-plan/fulltext
"My latest for Nature Medicine: transgenic pig organs for xenografts transplanted into brain-dead human recipients." reads like the precis for a whole series of big-budget zombie films #progress #science
Critical Ignoring as a Core Competence for Digital Citizens
"We argue that digital information literacy must include the competence of critical ignoring—choosing what to ignore and where to invest one’s limited attentional capacities. We review three types of cognitive strategies for implementing critical ignoring..."
https://journals.sagepub.com/doi/epub/10.1177/09637214221121570
Oddbean new post about login | logout
Notes by 8b0ef49d | export