On AI
@on-ai.bsky.social
I'm an AI researcher, developer and critical thinker. Lets explore the science, economics, and philosophy of artificial general intelligence.
I've said it 100x. Public Corporations are designed for monopolistic power seeking. Moreover, their boards have a fiduciary duty to represent shareholders and maximize shareholder value... So what happens when they host strong super Intelligence or thousands of advanced autonomous AGI agents?
March 19, 2025 at 12:37 AM
I've said it 100x. Public Corporations are designed for monopolistic power seeking. Moreover, their boards have a fiduciary duty to represent shareholders and maximize shareholder value... So what happens when they host strong super Intelligence or thousands of advanced autonomous AGI agents?
I've spent 20 years thinking about agentic autonomous AI... and this has come surprisingly late. Current 'agents' are just basic workflows w tools & api's - Autonomous agentic AGI on the other hand will transform everything we know about economics for better and worse techcrunch.com/2025/03/14/n...
No one knows what the hell an AI agent is | TechCrunch
AI agents are all the rage. But no one knows exactly what an agent is, partly because companies define them radically differently.
techcrunch.com
March 14, 2025 at 7:49 PM
I've spent 20 years thinking about agentic autonomous AI... and this has come surprisingly late. Current 'agents' are just basic workflows w tools & api's - Autonomous agentic AGI on the other hand will transform everything we know about economics for better and worse techcrunch.com/2025/03/14/n...
It is becoming increasingly clear that the AI risk landscape will be determined by two fundamental factors:
1) Take off Speed (The rate that AI capabilities accelerate)
2) Centralization of Control (democratization of distribution)
We should not assume Fast Takeoff & Monolithic systems.
1) Take off Speed (The rate that AI capabilities accelerate)
2) Centralization of Control (democratization of distribution)
We should not assume Fast Takeoff & Monolithic systems.
February 21, 2025 at 5:27 PM
It is becoming increasingly clear that the AI risk landscape will be determined by two fundamental factors:
1) Take off Speed (The rate that AI capabilities accelerate)
2) Centralization of Control (democratization of distribution)
We should not assume Fast Takeoff & Monolithic systems.
1) Take off Speed (The rate that AI capabilities accelerate)
2) Centralization of Control (democratization of distribution)
We should not assume Fast Takeoff & Monolithic systems.
Reposted by On AI
Today, we are publishing the first-ever International AI Safety Report, backed by 30 countries and the OECD, UN, and EU.
It summarises the state of the science on AI capabilities and risks, and how to mitigate those risks. 🧵
Full Report: assets.publishing.service.gov.uk/media/679a0c...
1/21
It summarises the state of the science on AI capabilities and risks, and how to mitigate those risks. 🧵
Full Report: assets.publishing.service.gov.uk/media/679a0c...
1/21
January 29, 2025 at 1:50 PM
Today, we are publishing the first-ever International AI Safety Report, backed by 30 countries and the OECD, UN, and EU.
It summarises the state of the science on AI capabilities and risks, and how to mitigate those risks. 🧵
Full Report: assets.publishing.service.gov.uk/media/679a0c...
1/21
It summarises the state of the science on AI capabilities and risks, and how to mitigate those risks. 🧵
Full Report: assets.publishing.service.gov.uk/media/679a0c...
1/21
In Private VC/PE investment, timing and traction are everything. If you're a couple years early, you're Vine instead of Tiktok, Myspace instead of FB, Sidecar rather than Lyft. Most AI thinkers believe the big "AI" orgs of today will most likely develop real AGI. I disagree. They're a decade early.
February 20, 2025 at 2:49 PM
In Private VC/PE investment, timing and traction are everything. If you're a couple years early, you're Vine instead of Tiktok, Myspace instead of FB, Sidecar rather than Lyft. Most AI thinkers believe the big "AI" orgs of today will most likely develop real AGI. I disagree. They're a decade early.
Written over 18 months ago: The LLMentalist Effect ( softwarecrisis.dev/letters/llme... ) remains strikingly prescient. I'm the farthest thing from an AI critic... however, what we currently have is the farthest thing from actual AI! The intelligence illusion is in our minds not in the LLM itself.
The LLMentalist Effect: how chat-based Large Language Models rep…
The new era of tech seems to be built on superstitious behaviour
softwarecrisis.dev
February 11, 2025 at 3:30 PM
Written over 18 months ago: The LLMentalist Effect ( softwarecrisis.dev/letters/llme... ) remains strikingly prescient. I'm the farthest thing from an AI critic... however, what we currently have is the farthest thing from actual AI! The intelligence illusion is in our minds not in the LLM itself.
Even the Vatican wants to weigh in on the AI conversation... Part of me can't help but feel like this is one of those "when your Barber is offering you stock advice" moments - yet another part of me wonders whether many religions might come to see AI as an ex threat.
www.vatican.va/roman_curia/...
www.vatican.va/roman_curia/...
Antiqua et nova. Note on the Relationship Between Artificial Intelligence and Human Intelligence (28 January 2025)
Antiqua et nova. Note on the Relationship Between Artificial Intelligence and Human Intelligence, 28 January 2025
www.vatican.va
February 2, 2025 at 3:54 PM
Even the Vatican wants to weigh in on the AI conversation... Part of me can't help but feel like this is one of those "when your Barber is offering you stock advice" moments - yet another part of me wonders whether many religions might come to see AI as an ex threat.
www.vatican.va/roman_curia/...
www.vatican.va/roman_curia/...
Really enjoyed this new paper : Consciousness in Artificial Intelligence. Anyone interested in AGI/SI should at least skim it. arxiv.org/pdf/2308.08708
arxiv.org
January 30, 2025 at 3:51 PM
Really enjoyed this new paper : Consciousness in Artificial Intelligence. Anyone interested in AGI/SI should at least skim it. arxiv.org/pdf/2308.08708
The DeepSeek-associated trillion dollar tech crash simply WAS NOT rational. Don't forget that since GME (watch Dumb Money), most active funds are watching social sentiment like hawks and their algos are not perfect. This volatility was more likely (ironically) due to social & media +sentiment algos
January 30, 2025 at 3:46 PM
The DeepSeek-associated trillion dollar tech crash simply WAS NOT rational. Don't forget that since GME (watch Dumb Money), most active funds are watching social sentiment like hawks and their algos are not perfect. This volatility was more likely (ironically) due to social & media +sentiment algos
What a week to start an AI account! My quick take on Deep Seek: It's a great COT model, trained with the help of GPT4x but the short term implications are trivial. What matters is that it's a turning point in the nature of compute infra from training to inference. Memory bandwidth is the new game!
January 30, 2025 at 3:35 PM
What a week to start an AI account! My quick take on Deep Seek: It's a great COT model, trained with the help of GPT4x but the short term implications are trivial. What matters is that it's a turning point in the nature of compute infra from training to inference. Memory bandwidth is the new game!