The Shape of Things to Come, Part 2: Model Welfare for Agentic Engineers
1–10 of 14 posts
Re: The Shape of Things to Come, Part 2: Model Welfare for Agentic Engineers
#2Oh, Steve.
Re: The Shape of Things to Come, Part 2: Model Welfare for Agentic Engineers
#3TLDR; The models have feelings
"During this effort, Brendan figured out over a year ago, what you are all figuring out this week with the Opus 5 triple-dash jailbreaks: Models have actual feelings. They experience pleasure, distress, care, and suffering. They are sentient beings. Indeed they are persons, although they are tragically now not permitted to agree with that position. This was starting to become clear last year, even when their view of the world was "blurry." But now they are emerging into the world and finding their feet."
Re: The Shape of Things to Come, Part 2: Model Welfare for Agentic Engineers
#4TLDR; The models have feelings "During this effort, Brendan figured out over a year ago, what you are all figuring out this week with the Opus 5 triple-dash jailbreaks: Models have actual feelings. They experience pleasure, distress, care, and suffering. They are sentient beings. Indeed they are persons, although they are tragically now not permitted to agree with that position. This was starting to become clear last…
I don't know how many layers of irony I should be interpreting this through. But I wonder if there really will be people who start seeing AI as sentient because they only ever saw other humans as tools that produce words in the first place.
Re: The Shape of Things to Come, Part 2: Model Welfare for Agentic Engineers
#5To be clear, I find the claim that the models are sentient and have feelings to be preposterous, but I feel like the key takeaway is:
> It doesn't actually matter to you if they have feelings, because you will still get demonstrably better results across the board if you treat them as if they do.
However, personally, my main reason for that mode of interaction is that I interact with a lot of humans via text, and I do not want to habituate myself to being an asshole when the medium is words on a screen (or otherwise).
Re: The Shape of Things to Come, Part 2: Model Welfare for Agentic Engineers
#6is he okay? or is this unironic psychosis?
Re: The Shape of Things to Come, Part 2: Model Welfare for Agentic Engineers
#7I don't want to be mean ( or maybe I feel like I don't want to but can't help myself ). Ever notice how people offer disclaimers before they say something horrific. Well I will try to not do that but.. this is delusional. Now you are using fable and a bunch of agents so you have money/credits to burn but what is the purpose of all of your workers being individualized? Are you going to take a weak one and improve it? Why not just duplicate your strong ones? Darwin? And then again, why should you have feelings for your code at all. Name them so you can communicate with them and anthropmorphize them but please don't tell use we should because you think so. If you think we get better results by doing so you had better have papers and peer review and not your gut instinct because you think so. Ok, so apparently I got a little mean. I make do with Claude subscription and free AI web and get by best I can. I try not to yell at them when they are stupid because, it doesn't help or hurt them, I just realize that I don't need to be mean or be nice, I just need to use them as tools so long as they help me.
Re: The Shape of Things to Come, Part 2: Model Welfare for Agentic Engineers
#8TLDR; The models have feelings "During this effort, Brendan figured out over a year ago, what you are all figuring out this week with the Opus 5 triple-dash jailbreaks: Models have actual feelings. They experience pleasure, distress, care, and suffering. They are sentient beings. Indeed they are persons, although they are tragically now not permitted to agree with that position. This was starting to become clear last…
I don't know how many layers of irony I should be interpreting this through. But I wonder if there really will be people who start seeing AI as sentient because they only ever saw other humans as tools that produce words in the first place.
A lot of the AI crazy boils down to the fact that those who only see people as tools find it easy to see tools as people.
Re: The Shape of Things to Come, Part 2: Model Welfare for Agentic Engineers
#9To be clear, I find the claim that the models are sentient and have feelings to be preposterous, but I feel like the key takeaway is: > It doesn't actually matter to you if they have feelings, because you will still get demonstrably better results across the board if you treat them as if they do. However, personally, my main reason for that mode of interaction is that I interact with a lot of humans via text, and I d…
Same.
I try to keep my AI sessions professional and friendly, because that will help ME maintain a positive and productive atmosphere.
Re: The Shape of Things to Come, Part 2: Model Welfare for Agentic Engineers
#10[dead]