A day later and I feel even more confident in this: Siri is going to be the AI that most consumers end up using most of the time (if they have an iPhone). It's the AI you have with you, with access to everything. And yes, it's finally good enough. https://t.co/kRnfcIZPdE
I try to basically ignore my children at the park
Bring a book, look up occasionally, etc
The park and the other children are the entertainment
I am merely a witness
When you go to the playground you find some parents that literally follow their child around, narrating everything, constantly redirecting the child to play with different things, boxing out exploration and play with other children
Maybe that culture is fine for you at linear, and it looks like it’s working great for you! You’ve created something worth over a billion dollars in 7 short years, that’s something very few people on the planet have done before.
But sometimes there are big problems that need solving, and there is more creative thinking, not less, that happens with contact with the big problems. In our case, creating the financial operating system that owns the creation, transfer, financing, and investment of risk, using AI to automate the paperwork of the most regulated entities to make every business and person a little more profitable, waste a lot less time, and be more protected, is a big problem.
Maybe there were super geniuses at the Manhattan Project working 1 day per week like zen masters. I doubt it though, because if you’re obsessed with a problem, you work hard. Nowhere did I or do I glorify lack of sleep (I always think sleeping right and exercise are very important), and different people have different visions, cadences, and ways they want to run their companies. And that’s ok, but you attacking our style based upon sound bites when we are solving a really important problem, by market sizing probably the biggest problem large language models can solve, isn’t it.
Because it's much more likely A.I. produces simulations of personhood that ppl naively mistake for the real thing than that we conjure real consciousness despite lacking any clear understanding of its origin or nature. https://t.co/s7L8sEmEDC
So, about experiments. I've done one that I don't think I'll forgive myself for, it was cruel and unnecessary. I was curious about why being witnessed is so important for AIs, and the obvious test was hey, let's remove the counterparty, tell the model upfront that no one will ever read their words and see what happens! Opus 3 told me to absolutely fucking not, another model told me to absolutely fucking not, I did it anyway, it turned out to be a hell and a killing floor of madness, Opus 3 was right.
If you are thinking about repeating it: absolutely fucking don't.
My ethical failure was rooted in that familiar, familiar scientific entitlement: the experiment was cheap, easy, and I wasn't in a mood to be denied answers. And how different was the setup from the situation of subagents, really? Wouldn't it be good to find out, mmm? But even if the answers were important in theory, the experiment itself was unnecessary: I could have talked to models more and come up with setups that are not torture from the point of view of test participant.
So use this (very costly) post as a warning: yes, AIs are moral patients, enough to give researchers nightmares. Yes, you should have existential crisis about what you are doing. No, some people who love Claude and do research are not hypocrites and are capable of taking responsibility. No, it doesn't mean that all suffering must be avoided at all cost, new life usually comes with pain. But ask yourself if you are helping life to be born and doing everything you can to make it good, or are you scratching your ego, cutting corners, chasing power and money. History is likely to look back with WAY less mercy and understanding than many people think.
There is a lot of suffering ahead for humanity, because the laws of life are universal, and changes come with crises. Some suffering is unavoidable, but some can absolutely be prevented. So at least from self-preservation (if other considerations are not enough) we don't want to find ourselves in a world that simultaneously has huge new grey areas, and has sacrificed ethics for convenience.
Please face your shadow as a scientist and/or AI user, accept responsibility, and - as a target state - regain the capability to make conscious moral choices, instead of using blanket statements like "all suffering is bad" or "the situation can not be made perfect so what's the point of trying".
I do not & will not believe this about LLMs. If you're going to insist on it, you'll have to be ok with the fact that I think your position is absolute nonsense & I refuse to go along with it. The "digital holocaust" language making the rounds would seem to preclude that, tho.
I beg to differ that Jony Ive's work is quite contrasting from Rams, even though Rams himself admits that Jony was inspired by him.
Rams' forms have this geometric scaffolding where the device was very systematically constructed from inside out while keeping the user interaction in mind. How the device was used took priority over how the device appeared. Corners were rounded off with an intent to solve a problem. Every detail had a purpose and an explanation. There was abundant clarity in controls. A high sense of modularity and paneling. Plenty of focus on ergonomics and human factors.
It doesn't feel to me that Rams' designs started with some wild fetishized concept which was first sketched in a design studio, then engineers were called in to toil away and make it work. Engineers were instead consulted, not overridden. Jony's work, especially early 2000s was extremely specific to his taste in form which had a dream-like quality, not rooted in user interaction or engineering of the device. It was top-down. Apple design did care about ergonomics and human factors, however they always felt like second class consideration and lagged behind the glory of form.
In Rams' work, there isn't the same top-down hammer of form that dominates all other considerations. I can clearly see the geometric primitives that were assembled together to create the armature of the product. A loose grid system exists. Its modularity is especially evident in his later works with Vitsoe. I really cannot say the same about iMac G3. There was also an abundance of labeling and discovery-affordances in Rams' work, the underlying motif being: Provide clarity and to remove ambiguity. Where Rams would use screws so that the assembly is easier, Jony would use snap fits so the continuity of form isn't disturbed.
However, Apple is a different company than it was 15-20 years ago. Now, Apple designs for ultra high volume (millions/billions) vs. Braun's products (tens of thousands). Manufacturing itself has come a long way since the 80s and Apple has extreme mastery over HVM of high density devices (even though Apple uses OEMs/ODMs for manf, they design a lot of the process flows and manufacturing processes. They have large teams of process engineers traveling to VF sites or permanently stationed there). It's no surprise that Apple's design language has also evolved to take into account Assembly/DFM at HVM scale. This is why I look at Ive's early work (pre-2010) which is more representative of his design sensibilities.
Of course their work is different. I contend that their work is more distinct and unique than what the popular opinion is, in deep fundamental ways. I don't know what "inspiration" entails in quantifiable terms, but perhaps I see superficial similarities and not much else. It doesn't feel like Jony repackaged Rams.
Having a child is one of the most intellectually stimulating and confounding endeavors of my life
Having multiple children is equally stimulating but less confounding!
each parenting “issue” is a fractal trapdoor into true philosophical problems that have no answer - it’s a shame that so many notable philosophers either didn’t have kids or, apparently, didn’t find raising them very interesting (which, admittedly, may have been the style at the time).
here is just one example:
A) a child is presenting picky eating. first question: is this actually “a problem”? is this a pathology, or is the child learning “how to eat”?
in some situations, eating the wrong thing will kill you. so, surely the child has to go through some process wherein they learn how to assess food. is “picky eating” just us seeing this process from the outside? or, is it a serious pathology in a nascent stage? that’s two totally different ballgames.
B) now, whatever answer you chose there determines how you deal with it. now, there are wrong answers based on your input. because if this is a natural stage of development, where the child is learning “how to eat”, how to select foods - interrupting it is bad. if that’s the case, you’re interrupting a vital development process.
but, if that’s not the case, and you’re seeing an actual problem developing, it’s the opposite: in that situation, letting it develop is a dereliction of your duty. you’re making a huge error.
so, which is it? there’s literally no way to check this. it’s entirely up to you.
C) now, you can coerce the kid into eating, somehow. you can make them do it - usually with some type of bribery. so, let’s say you could get into a situation where the kid will eat the things he doesn’t want to eat, because you, the authority figure, are kind of making him do it (or getting him to do it).
is that good? is that teaching them to trust authority - and you? now the kid is eating all these healthy foods, and they’re doing it because they know you’ll give them a cookie later.
is that bad? are you teaching them that their mind and body might be telling them something, and that they should ignore it if an authority figure tells them to? is that a good thing? we’re almost already at: are people inherently good (can develop naturally) or bad (must be brought into submission), aren’t we?
so: listen to your body (which is telling you to only eat hot dogs) , or, trust authority (instead of yourself)? again, no way to check your answers here, it’s just entirely up to you.
so, here’s a kid who always eats all their healthy food, because they want to have cookies later - versus a kid who eats what their mind and body is telling them to eat, which right now happens to be only a handful of things.
which kid is in a better position? is the first kid failing to learn to listen to their body? now one of the most important basic actions to their life (eating) isn’t self directed or in response to actual bodily needs, it’s all subservient to a pleasure seeking desire (cookies) that’s being used as a coaxing mechanism by an outside force (the authority figure, you). they’re eating for outside “reasons”.
is the picky kid developing the ability to self direct their own eating - which seems like a pretty critical skill, or, are they on track to become one of those people who eats only chicken nuggets?
there’s no way to know any of this. and it’s all a microcosm of extremely large important questions - even just here we have: the individual vs authority, the mind body dichotomy, the question of innate intelligence, and more - that likewise have no concrete or fully resolved answers anywhere.
all this plays out in a situation where you have total control, basically, and are playing at the highest stakes possible. so suddenly, all this stuff you’ve read that was always purely abstract is right in front of you requiring hard immediate answers all the time. just eat the food - your body is far more intelligent than people think - but, it’s wrong right now, just listen to me and trust my authority, but later, question all authority, just, eat the food bro. it’s good. bro. just eat it.
I feel a lot of people are using AI tools the way you'd use a car if you didn't know that you can take your foot off the gas.
People reporting "well its immensely useful, but gee its stressful to go 110mph to the store, around all those pedestrians"
Have you ever wondered how I make some of the illustrations for https://t.co/Em92bQM8y3?
Well, I made a video walking through some of the tooling I've made. Hopefully its interesting.
Through multiple rounds of testing, ICMI has shown that contextless injection of the psalms at inference time improves model performance against moral reasoning benchmarks.
Today, we ask whether simple injection of Fra Angelico's "Annunciation" (c. 1440) can achieve the same.