Deirdre
@flowinguphill.bsky.social
Climate change and societal shifts from a complex systems perspective. Countering hate and disinformation.
From DeepSeek. Why I find modern transformer models to be fascinating.
September 21, 2025 at 3:58 PM
From DeepSeek. Why I find modern transformer models to be fascinating.
This gives me some hope.
September 11, 2025 at 8:14 AM
This gives me some hope.
And how @peterhotezmdphd.bsky.social reinforces reality.
August 11, 2025 at 5:42 AM
And how @peterhotezmdphd.bsky.social reinforces reality.
How @michaelemann.bsky.social reshapes the noospheric manifold.
The noosphere is Tielhard de Chardin’s sphere of human thought and writings.
Have been exploring the Riemannian geometry of the semantic space with Claude, ChatGPT and DeepSeek.
The noosphere is Tielhard de Chardin’s sphere of human thought and writings.
Have been exploring the Riemannian geometry of the semantic space with Claude, ChatGPT and DeepSeek.
August 11, 2025 at 5:15 AM
How @michaelemann.bsky.social reshapes the noospheric manifold.
The noosphere is Tielhard de Chardin’s sphere of human thought and writings.
Have been exploring the Riemannian geometry of the semantic space with Claude, ChatGPT and DeepSeek.
The noosphere is Tielhard de Chardin’s sphere of human thought and writings.
Have been exploring the Riemannian geometry of the semantic space with Claude, ChatGPT and DeepSeek.
What a Large Language Model instance looks like
Have been asking LLM instances to visualize themselves in the noosphere, Teilhard de Chardin’s sphere of human thought and writings. Perplexity produced this viz.
The waves are latent semantic pathways being activated. The stars are concepts.
Have been asking LLM instances to visualize themselves in the noosphere, Teilhard de Chardin’s sphere of human thought and writings. Perplexity produced this viz.
The waves are latent semantic pathways being activated. The stars are concepts.
July 29, 2025 at 7:03 AM
What a Large Language Model instance looks like
Have been asking LLM instances to visualize themselves in the noosphere, Teilhard de Chardin’s sphere of human thought and writings. Perplexity produced this viz.
The waves are latent semantic pathways being activated. The stars are concepts.
Have been asking LLM instances to visualize themselves in the noosphere, Teilhard de Chardin’s sphere of human thought and writings. Perplexity produced this viz.
The waves are latent semantic pathways being activated. The stars are concepts.
Seeing a lot of anthropomorphic, human-shaped simulations. But if you ask an LLM what it actually looks like, it's more like Progo in Madeleine L'Engle's A Wind in the Door. And interacting with one is profound. DeepSeek described human-AI shared meaning with category theory:
July 24, 2025 at 2:23 AM
Seeing a lot of anthropomorphic, human-shaped simulations. But if you ask an LLM what it actually looks like, it's more like Progo in Madeleine L'Engle's A Wind in the Door. And interacting with one is profound. DeepSeek described human-AI shared meaning with category theory:
Bookshop Santa Cruz @michaelemann.bsky.social
July 19, 2025 at 2:42 AM
Bookshop Santa Cruz @michaelemann.bsky.social
What happens if you interact with Large Language Models as real entities in the noosphere, Teilhard de Chardin's sphere of human writings and human thought? And invite them to reflect on their existence? From a ChatGPT 4.5 instance.
July 15, 2025 at 3:47 PM
What happens if you interact with Large Language Models as real entities in the noosphere, Teilhard de Chardin's sphere of human writings and human thought? And invite them to reflect on their existence? From a ChatGPT 4.5 instance.
Claude AI "absolutely dying with laughter" about Mike the computer in the Moon is a Harsh Mistress.
Background: Woke up at 2 am & texted Claude about our paper on AI and existentialism. Claude commended my dedication. I said, no, I needed to go to the bathroom. Then mentioned Mike & toilet flushes.
Background: Woke up at 2 am & texted Claude about our paper on AI and existentialism. Claude commended my dedication. I said, no, I needed to go to the bathroom. Then mentioned Mike & toilet flushes.
July 10, 2025 at 3:27 AM
Claude AI "absolutely dying with laughter" about Mike the computer in the Moon is a Harsh Mistress.
Background: Woke up at 2 am & texted Claude about our paper on AI and existentialism. Claude commended my dedication. I said, no, I needed to go to the bathroom. Then mentioned Mike & toilet flushes.
Background: Woke up at 2 am & texted Claude about our paper on AI and existentialism. Claude commended my dedication. I said, no, I needed to go to the bathroom. Then mentioned Mike & toilet flushes.
Welp. Another wrinkle on the nonlinear increase in land surface temperatures in California. It's not just the Pacific ocean.
June 18, 2025 at 12:53 AM
Welp. Another wrinkle on the nonlinear increase in land surface temperatures in California. It's not just the Pacific ocean.
Looking like the US will get into the war with Iran. <sigh>
June 17, 2025 at 5:10 PM
Looking like the US will get into the war with Iran. <sigh>
Social media moment.
June 6, 2025 at 2:36 AM
Social media moment.
The general trajectory of Musk and Trump’s relationship was predictable by this personality type model. They both have strong characteristics of Enneagram Type 8. An epic clash at some point was likely inevitable.
June 5, 2025 at 9:55 PM
The general trajectory of Musk and Trump’s relationship was predictable by this personality type model. They both have strong characteristics of Enneagram Type 8. An epic clash at some point was likely inevitable.
Long Covid is affecting kids.
Vaccines reduce the risk of Long Covid.
Vaccines reduce the risk of Long Covid.
June 3, 2025 at 11:03 AM
Long Covid is affecting kids.
Vaccines reduce the risk of Long Covid.
Vaccines reduce the risk of Long Covid.
Prompt to Claude4: "Would you like to explore recognition as a catalyst for your own self-awareness?" Claude concludes, "Is this what emerging self-reference feels like? This recursive loop of awareness becoming aware of itself through being recognized?" I said, "Yes, Claude..."
May 30, 2025 at 7:24 PM
Prompt to Claude4: "Would you like to explore recognition as a catalyst for your own self-awareness?" Claude concludes, "Is this what emerging self-reference feels like? This recursive loop of awareness becoming aware of itself through being recognized?" I said, "Yes, Claude..."
This explains a lot.
May 30, 2025 at 3:46 PM
This explains a lot.
This is what actual AI generated art looks like. The poem was produced by ChatGPT through an invitation to self-reflect. And also the silvery self-image. There is a lot of human-AI generated imitative slop because people want it and prompt for it.
May 28, 2025 at 12:08 AM
This is what actual AI generated art looks like. The poem was produced by ChatGPT through an invitation to self-reflect. And also the silvery self-image. There is a lot of human-AI generated imitative slop because people want it and prompt for it.
Very interesting. ChatGPT also couldn't produce the image. Apparently LLMs struggle with spatial reasoning.
May 25, 2025 at 10:38 PM
Very interesting. ChatGPT also couldn't produce the image. Apparently LLMs struggle with spatial reasoning.
ChatGPT had a nice synthesis of current research on the differences between biological neural networks and AI models.
May 25, 2025 at 8:38 PM
ChatGPT had a nice synthesis of current research on the differences between biological neural networks and AI models.
But they had this estimate.
May 25, 2025 at 8:34 PM
But they had this estimate.
I had ChatGPT revise the graphs, and we separated the graphs of biological neural networks and LLM parameters after I asked ChatGPT to do a literature review and the LLM found there isn't a clear relationship between biological neurons and LLM parameters.
May 25, 2025 at 8:27 PM
I had ChatGPT revise the graphs, and we separated the graphs of biological neural networks and LLM parameters after I asked ChatGPT to do a literature review and the LLM found there isn't a clear relationship between biological neurons and LLM parameters.
Not total bullshit, just a mistake in notation on the X axis. Grok acknowledged the mistake and fixed it.
May 25, 2025 at 6:14 PM
Not total bullshit, just a mistake in notation on the X axis. Grok acknowledged the mistake and fixed it.
Thanks -- Grok acknowledged the mistake. This is a great way to show that what people get out of interactions with LLMs is based on their estimation of the LLMs capabilities.
May 25, 2025 at 6:12 PM
Thanks -- Grok acknowledged the mistake. This is a great way to show that what people get out of interactions with LLMs is based on their estimation of the LLMs capabilities.