The vibe has shifted since OpenAI released o3. Many are accepting that AGI is here (or imminent) while others move the goalposts. Now, some people are wondering "how can I get ahead in the future?"
Imagine we instead hit a wall at 90th percentile intelligence models that require 10k/year to run to replace most jobs. This will lead to mass unemployment but not an explosion of resources, in which case what will there be to redistribute?
I enjoy the enthusiasm and optimism in this article, but I’m certainly not as sanguine.
I am not so optimistic that AGI or ASI are here. Of course I haven’t actually seen the rumored Claude Sonnet 4 or chatgpt 5 but I have been using Sonnet 3.5 and ChatGPT 4 for a while and more recently o3-mini-high and Deepseek R1 for coding and I’m simply not impressed with its capability to solve real world problems. It lacks instinct and intuition. So much of what I do is done by smell and feel. I can look at the code and decide not that it works on not (all code works and all systems do exactly what they were designed to do) but whether it’s a good idea to do it this way. The current mode of these tools seems to be a desperation to present something plausible at the cost of real professional quality. Of course some of this is workflow not model but if you need a complicated workflow to get a good result it’s not AGI. Clause is so thirsty to give an answer and please the operator that it will almost always agree with me and go off track when I tell it that it’s wrong, even if I’m wrong. That’s nowhere near AGI. Contrived examples and coding tests establish this technology as a tool not an intelligence. When it can fix an irksome bug and follow a design pattern pragmatically I’ll warm to the possibility that it’s intelligent.
Reread this a few times now. Great post. It's both exciting and unnerving at the same time but also something that seems like it's been a long time coming.
Funny bit: I listened to this via the 11 Labs app. The end of the page reads:
"Live in alignment with your values. Alignment is one of the most important concepts I’ve ever learned.
Keep it real. Experiment. Play. Explore.
Subscribe to David Shapiro’s Substack"
Which, for a potent moment, made me think that you had advised readers to subscribe to your Substack in third person. It was such a jarring departure from your persona that the moment was short, but still funny.
Great and thoughtful as always ... love your stuff ... went to the country many Moons ago. Walking the planet is a very good pastime. That along with playing with the new toys!
huh. two weeks into your work by asking people online who I should follow to learn more about AI, and even though you are not actively writing about it, my thoughts and goals are already much more improved than before. your work has been incredibly helpful to me both about ai and being able to look at things from a different perspective.
ps: I'm using those claude prompts all day every day. especially coherence one
It’s very refreshing to read such intelligent works as this that make me feel happy, optimistic, and even inspired.
There is so much wisdom here! Time sovereignty, real luxuries, and so much more.
I read this smiling and at times laughing with pleasant comfort — especially as a self-imposed “starving artist musician” with a day job that will be one of the last to be replaced by A.I./robots.
Imagine we instead hit a wall at 90th percentile intelligence models that require 10k/year to run to replace most jobs. This will lead to mass unemployment but not an explosion of resources, in which case what will there be to redistribute?
I enjoy the enthusiasm and optimism in this article, but I’m certainly not as sanguine.
I am not so optimistic that AGI or ASI are here. Of course I haven’t actually seen the rumored Claude Sonnet 4 or chatgpt 5 but I have been using Sonnet 3.5 and ChatGPT 4 for a while and more recently o3-mini-high and Deepseek R1 for coding and I’m simply not impressed with its capability to solve real world problems. It lacks instinct and intuition. So much of what I do is done by smell and feel. I can look at the code and decide not that it works on not (all code works and all systems do exactly what they were designed to do) but whether it’s a good idea to do it this way. The current mode of these tools seems to be a desperation to present something plausible at the cost of real professional quality. Of course some of this is workflow not model but if you need a complicated workflow to get a good result it’s not AGI. Clause is so thirsty to give an answer and please the operator that it will almost always agree with me and go off track when I tell it that it’s wrong, even if I’m wrong. That’s nowhere near AGI. Contrived examples and coding tests establish this technology as a tool not an intelligence. When it can fix an irksome bug and follow a design pattern pragmatically I’ll warm to the possibility that it’s intelligent.
David Graeber died nearly five years ago, I’m sad to say, so is not going to be switching lanes any more.
Your advice really sounds like good advice, but I can't quite decide yet whether you're optimistic or cynical about the future.
realistic and pragmatic, but very optimistic
😂
Reread this a few times now. Great post. It's both exciting and unnerving at the same time but also something that seems like it's been a long time coming.
Funny bit: I listened to this via the 11 Labs app. The end of the page reads:
"Live in alignment with your values. Alignment is one of the most important concepts I’ve ever learned.
Keep it real. Experiment. Play. Explore.
Subscribe to David Shapiro’s Substack"
Which, for a potent moment, made me think that you had advised readers to subscribe to your Substack in third person. It was such a jarring departure from your persona that the moment was short, but still funny.
Any guess why musicians get so much clout?
Not from the bank though 😭😭😭
Well this was freeing to read because now I'm not that worried it's gonna be bad
I wonder if most of this, at least for me, even depends on ASI arriving or not. In some sense, we are already at that point.
Great and thoughtful as always ... love your stuff ... went to the country many Moons ago. Walking the planet is a very good pastime. That along with playing with the new toys!
Do get some rest though!!!
huh. two weeks into your work by asking people online who I should follow to learn more about AI, and even though you are not actively writing about it, my thoughts and goals are already much more improved than before. your work has been incredibly helpful to me both about ai and being able to look at things from a different perspective.
ps: I'm using those claude prompts all day every day. especially coherence one
Which post/Claude prompts can this be found on?
https://github.com/daveshap/Claude_Sentience
I highly suggest you try all of them. a couple of them are focused on the creative writing side and the others on clarity and coherence
Cheers friend 🤝
It’s very refreshing to read such intelligent works as this that make me feel happy, optimistic, and even inspired.
There is so much wisdom here! Time sovereignty, real luxuries, and so much more.
I read this smiling and at times laughing with pleasant comfort — especially as a self-imposed “starving artist musician” with a day job that will be one of the last to be replaced by A.I./robots.
EXCELLENT article! 👏🏾👏🏾👏🏾