The $1 Trillion Question, Mortality in the Age of Generative Ghosts, Your Mind and AI, How to Design AI Tutors for Learning, The Imagining Summit Preview: Adam Cutler, and Helen's Book of the Week.
Magritte's "Son of Man" metaphor explores digital obscurity of reality. AI model collapse mirrors human isolation effects. Minds struggle without diverse input. Balancing AI benefits and genuine human connection crucial. Develop metacognitive skills to navigate this new landscape.
Generative AI promises personalized learning at scale but risks creating dependency. Complementary cognitive artifacts enhance skills; competitive ones replace them. Effective AI tutors balance engagement and autonomy, expanding human cognition without diminishing critical abilities.
After four years, we are relaunching Artificiality with a new site, new focus, and new business model. Woohoo!
It's time to celebrate!
After four years, we are relaunching Artificiality with a new site, new focus, and new business model. Woohoo!
If you read nothing else this week: make sure to read the section below about how you, our current subscribers, can access our new site—including your complimentary subscription for the next year.
Artificiality wouldn't be what it has become without you, our subscribers. Thank you for your attention, care, and feedback. We are thrilled to have had you on our journey so far and are thrilled to keep you with us.
Our new site: We have migrated from Substack to Ghost to have more flexibility in expanding our work. Our URL remains the same—www.artificiality.world—but you'll see we have updated everything else. We think we've found most of the bugs but please let us know if you find anything we've missed.
At Artificiality, we are obsessed with the complex phenomenons emerging from the expanding relationships we humans have with machines. Our current obsessions include:
Our new business model: We are shifting to subscriptions to fund the expansion of our work. A contributing factor is using a paywall reduces the possibility that generative AI companies will continue to hoover up our content in their training nets. We will produce our work for three audiences:
All of our subscribers prior to December 1, 2023 have been automatically provided a complimentary subscription tier for one year. Thank you for being with us this far—we hope that you'll continue to be part of the Artificiality community.
Thank you again for being part of the Artificiality community. Please feel free to reach out any time. Feedback is a gift and we'll be grateful for any you provide.
Cheers!
New research reveals that large language models can generate superior prompts for themselves through automated techniques, reducing reliance on specialized fine-tuning.
This new paper out of Microsoft is less about medicine as it is about prompting. This matters because the effectiveness of automated prompting looks to be somewhat general across domains. Which means that more people have more options to cost effectively work with foundation models.
Helen Edwards
An introduction to our obsession with the parallels between human and machine intelligence.
Key Points
Helen Edwards
Maybe it's because of this: Security researchers systematically tested over 200 unique, user-designed GPT models and revealed most possess a critical security weakness—leaving them prone to "prompt injection" attacks. By crafting malicious text prompts, attackers could exploit these customized ChatGPTs to extract private customization data and training documents without permission. Sensitive information like custom goals encoded into the AI and private files used to specialize the models could be stolen through simple text queries.
Researchers demonstrated two major risks from such attacks:
Perhaps most alarmingly, attempted defense mechanisms were also shown to fall short against experienced attackers. Simply relying on defensive prompts or disabling certain functions proved inadequate to fully protect customized GPTs.
Are you tired of all the breathless tech-bro AI talk, the hype about the next human thing AI will replace, the latest gossip and speculation about some weirdly named model? This article in the New Yorker is wonderful and makes me want to go back in time and watch toddlerdom all over again.
Angie Wang doesn't deny the impact and importance of AI but reframes it in the context of what humans are as biological, sensemaking, growing, learning creatures whose intelligence derives from being in the world, data that comes via the physical world of our senses, outputs that matter to those around us.
"What we say is not weighted by probability alone but is given its heft and color by our wish to share our lives. Mere mimicry is not enough."
"You're human too. You understand love. With that in mind, surely you understand how even the most useless tiny people on this planet are more real than any probabilistic word predictor, and how you must be, too. Human obsolescence is not here and can never be."
The Artificiality Weekend Briefing: About AI, Not Written by AI
