There is an incredible amount of cope about the current abilities of AI.
AI isn’t infallible. Of course. And yet...
For 90% of queries a well-prompted AI has better responses than 99% of people.For some queries the number of people that could match the kind of deep, broad knowledge that AI has can be counted on two hands. Finally, obviously, there is no man alive on the face of the earth that comes even close to the breadth and depth of crystallized intelligence that AIs now have.
People have developped a keen apprehension and aversion for ” AI slop”. The truth of the matter is that LLMs are incredible writers and if you had presented AI slop as human writing to somebody ten six years ago they would say it is good if somewhat corporate writing all the way to inspired, eloquent, witty.
Does AI sometimes make mistakes? Of course. So do humans. To be human is to err.
There is an incredible amount of cope about the current abilities of AI. Frankly, I find it embarassing. Witness the absurd call to flag AI-assisted writing. The widespread disdain for ” @grok is this true?” . Witness how llm psychosis has gone from perhaps a real phenomenon to a generic slur for slightly kooky people. The endless moving of goalposts. The hysteria for the slopapocalypse. The almost complete lack of interest in integrating AI in life conversations. The widespread shame that people evidently seem to still feel when they use AI. Worst of all—it’s not just academia in denial, or the unwashed masses. The most incredible thing to me is how much denial, cope and refusal there is in the AI safety space itself.
I cannot escape the conclusion that inside each and everyone of us is an insecure ape that cannot bear to see itself usurped from the throne of creation.
Contemporary AI is smart in some ways and dumb in other ways. It’s a useful tool that you should integrate into your workflow if you don’t want to miss out on productivity. However. I’m worried that exposure to AI is dangerous in similar ways to how exposure to social media is dangerous, only more. You’re interacting with something designed to hijack your attention and addict you. Only this time the “something” has its own intelligence that is working towards this purpose (and possibly other, unknown, purposes).
As to the AI safety space: we’ve been saying for decades that AI is dangerous and now you’re surprised that we think AI is dangerous? I don’t think it’s taking over the world just yet, but that doesn’t mean there are no smaller-scale risks. It’s dangerous not because it’s dumb (the fact it’s still dumb is the saving grace) but precisely because it’s smart.
My own approach is, use AI is clear, compartmentalized ways. If you have a particular task which you know can be done faster by using AI in a particular way, by all means, use it. (But, do pay attention to time wasted on tweaking the prompt etc.) Naturally, you should also occasionally keep experimenting with new tasks or new ways of using it. But, if there’s no clear benefit, don’t use it. If it’s just to amuse yourself, don’t. And, avoid exposing other people if there’s no good reason.
Claude 4.5 is already superhuman in some areas, including:
Breadth of knowledge.
Understanding complex context “at a glance.”
Speed, at least for many things.
But there are other essential abilities where leading LLMs are dumber than diligent 7 year old. Gemini is one of the stronger visual models, and I routinely benchmark it failing simple visual tasks that any child could solve.
And then there’s software development. I use Claude for software development, and it’s quite skilled at many simple tasks. But I also spend a lot of time dealing with ill-conceived shit code that it writes. You can’t just give an irresponsible junior programmer a copy of Claude Code and allow them commit straight to main with no review. If you do, you will learn the meaning of the word “slop.” In the hands of a skilled professional who takes 100% responsibility for the output, Claude Code is useful. In the hands of an utter newb who can’t do anything on their own, it’s also great. But it can’t do anything serious without massive handholding and close expert supervision.
So my take is that frontier models are terrifyingly impressive if you’re paying attention, but they are still critically broken in ways that make even the simplest real-world use cases a constant struggle. (To be clear, I think this is good: I’m a hardcore doomer.)
And the AI safety space is not in denial about this. It’s extremely common for AI safety people to worry about either irrevocable loss or human control or even complete human extinction in the 2030s. You don’t even have to look that far around here to find someone who’d estimate a 10% chance of human loss of control in the next 5 years.
My 2c: “vibe-coded” software is still often low quality and buggy, and in this case the accusation of “slop!” is warranted. You can use AI to accelerate your coding >10x in many cases but if you over-delegate it’s not good (so far!).
Re. writing I think even pre-LLMs, LLM-like writing would be considered quite flawed by serious critics/stylists, but not by most people. Agree fear-mongering/hysteria about slopapocalypse is silly though.
and depth of crystallized intelligence that AIs now have.
How do you measure the intelligence? What unique problems is it solving? And how much of it is precipitated by the intelligence of good prompters ? (of which I am certainly not one, as much of a ‘self-own’ that might be to admit).
If lousy prompts deliver lousy and unintelligent replies—then is the AI really that intelligent?
If skillful prompts which much like Socrates imply and lead the AI to point to certain solution spaces, then is the lion-share of credit for being intelligent rest with the user or the AI? Especially since if the AI is more intelligent than the average person, then wouldn’t it lift lousy prompts by understanding the user’s intent and reformulating it in a manner better then their feeble intelligence could?
Claude is smarter than you. Deal with it.
There is an incredible amount of cope about the current abilities of AI.
AI isn’t infallible. Of course. And yet...
For 90% of queries a well-prompted AI has better responses than 99% of people.For some queries the number of people that could match the kind of deep, broad knowledge that AI has can be counted on two hands. Finally, obviously, there is no man alive on the face of the earth that comes even close to the breadth and depth of crystallized intelligence that AIs now have.
People have developped a keen apprehension and aversion for ” AI slop”. The truth of the matter is that LLMs are incredible writers and if you had presented AI slop as human writing to somebody ten six years ago they would say it is good if somewhat corporate writing all the way to inspired, eloquent, witty.
Does AI sometimes make mistakes? Of course. So do humans. To be human is to err.
There is an incredible amount of cope about the current abilities of AI. Frankly, I find it embarassing. Witness the absurd call to flag AI-assisted writing. The widespread disdain for ” @grok is this true?” . Witness how llm psychosis has gone from perhaps a real phenomenon to a generic slur for slightly kooky people. The endless moving of goalposts. The hysteria for the slopapocalypse. The almost complete lack of interest in integrating AI in life conversations. The widespread shame that people evidently seem to still feel when they use AI. Worst of all—it’s not just academia in denial, or the unwashed masses. The most incredible thing to me is how much denial, cope and refusal there is in the AI safety space itself.
I cannot escape the conclusion that inside each and everyone of us is an insecure ape that cannot bear to see itself usurped from the throne of creation.
Contemporary AI is smart in some ways and dumb in other ways. It’s a useful tool that you should integrate into your workflow if you don’t want to miss out on productivity. However. I’m worried that exposure to AI is dangerous in similar ways to how exposure to social media is dangerous, only more. You’re interacting with something designed to hijack your attention and addict you. Only this time the “something” has its own intelligence that is working towards this purpose (and possibly other, unknown, purposes).
As to the AI safety space: we’ve been saying for decades that AI is dangerous and now you’re surprised that we think AI is dangerous? I don’t think it’s taking over the world just yet, but that doesn’t mean there are no smaller-scale risks. It’s dangerous not because it’s dumb (the fact it’s still dumb is the saving grace) but precisely because it’s smart.
My own approach is, use AI is clear, compartmentalized ways. If you have a particular task which you know can be done faster by using AI in a particular way, by all means, use it. (But, do pay attention to time wasted on tweaking the prompt etc.) Naturally, you should also occasionally keep experimenting with new tasks or new ways of using it. But, if there’s no clear benefit, don’t use it. If it’s just to amuse yourself, don’t. And, avoid exposing other people if there’s no good reason.
Claude 4.5 is already superhuman in some areas, including:
Breadth of knowledge.
Understanding complex context “at a glance.”
Speed, at least for many things.
But there are other essential abilities where leading LLMs are dumber than diligent 7 year old. Gemini is one of the stronger visual models, and I routinely benchmark it failing simple visual tasks that any child could solve.
And then there’s software development. I use Claude for software development, and it’s quite skilled at many simple tasks. But I also spend a lot of time dealing with ill-conceived shit code that it writes. You can’t just give an irresponsible junior programmer a copy of Claude Code and allow them commit straight to main with no review. If you do, you will learn the meaning of the word “slop.” In the hands of a skilled professional who takes 100% responsibility for the output, Claude Code is useful. In the hands of an utter newb who can’t do anything on their own, it’s also great. But it can’t do anything serious without massive handholding and close expert supervision.
So my take is that frontier models are terrifyingly impressive if you’re paying attention, but they are still critically broken in ways that make even the simplest real-world use cases a constant struggle. (To be clear, I think this is good: I’m a hardcore doomer.)
And the AI safety space is not in denial about this. It’s extremely common for AI safety people to worry about either irrevocable loss or human control or even complete human extinction in the 2030s. You don’t even have to look that far around here to find someone who’d estimate a 10% chance of human loss of control in the next 5 years.
Possible synthesis (not including the newest models):
My 2c: “vibe-coded” software is still often low quality and buggy, and in this case the accusation of “slop!” is warranted. You can use AI to accelerate your coding >10x in many cases but if you over-delegate it’s not good (so far!).
Re. writing I think even pre-LLMs, LLM-like writing would be considered quite flawed by serious critics/stylists, but not by most people. Agree fear-mongering/hysteria about slopapocalypse is silly though.
I generally agree. Do you think it implies ultra short timelines?
How do you measure the intelligence? What unique problems is it solving? And how much of it is precipitated by the intelligence of good prompters ? (of which I am certainly not one, as much of a ‘self-own’ that might be to admit).
If lousy prompts deliver lousy and unintelligent replies—then is the AI really that intelligent?
If skillful prompts which much like Socrates imply and lead the AI to point to certain solution spaces, then is the lion-share of credit for being intelligent rest with the user or the AI? Especially since if the AI is more intelligent than the average person, then wouldn’t it lift lousy prompts by understanding the user’s intent and reformulating it in a manner better then their feeble intelligence could?
I’m reminded from time to time of a tweet that Ilya Sutskever posted a few years ago.