I liked the monolith. If I didn't want any AI on my device, I liked having the ability to switch that off and be done with it like Firefox offers (and what Apple offered in the last major release).
While I broadly agree with the premise of re-directing the “purpose” of math, I quite detest the idea that judgement might be primarily based upon some in person discussion, or
oral presentation, and the claim that written mathematics that is not orally communicated might be less worthwhile in some sense (i know this isn’t the exact statement of the authors proposition).
There are a good deal of people, whom, falter much more in oral discussions, whether this be for a psychological thing, stage fright, or difficulty explaining things on the spot. There are also certainly brilliant people, who can’t give an informative, discussion inviting talk to save their lives, but given enough time, can formalize their thoughts in writing at the highest levels of their field, and that writing is likewise very enlightening (sometimes).
It’s not clear to me, that, AI as is, could not pose successfully in an oral discussion of a topic. I mention this because it seems that one implication of the article is that AI might write things that are logically correct, but devoid of understanding. I suggest rather that 1) it is not extremely improbably that AI is incapable of generating mathematics that furthers human understanding and if 2) it is indeed highly likely that they cannot generate mathematics that furthers human understanding in a textual format, then surely one could also differentiate between human and AI on a textual level, and judge the contribution of a human, without the need of oral discussion?
I suppose another aside is, one might claim that the existence of AI means people have much much more text to filter for, and so, it becomes difficult to find one person’s good writing amidst a sea of, logically correct, yet understanding devoid textual content. But by and large much or mathematical academia certainly operates off of some reputation/vouching system presently anyways, that already serves as a “filter” in some sense. Perhaps the existence of such a system/culture is not a good thing, but oral discussions/seminars certainly aren’t immune from such predilections.
Perhaps I’m babbling like an idiot, but the entire and sole purpose of this comment is just to say: for the love of god please don’t let the standard be judged by oral presentation
One thing I've always been curious about is, often times it seems that models don't seem to have these addendums like "dont do excessive/random shit" by default? Or I suppose, if it did have something like that, and still screws up like in your example, it clearly isn't working, so assume that it doesn't.
But the phrase "don't do random shit", semantically, from a monkey's paw perspective, could imply curtailing the model's creativity and 'thinking out of the box' capacity, that might have existed in its 'reasoning' process. So I'm always concerned if it's possible that, adding these phrases might be part of the reason why a model performs dumber than it should.
I don't have empirical evidence to support that supposition though.
I think of it as a tradeoff between creativity and specificity. Every instruction you give reduces creativity, and at best, increases specificity (I imagine a lot of prompts like "make no mistakes" do literally nothing but pollute context, but I haven't evaled them)
So if you're doing something very ordinary, fewer instructions result in better results. If you're doing something fairly off-piste, you have to give instructions to that effect and accept less creativity. For situations where you want it to do something extremely specific, tons of instructions and accept that you're going to get much closer output but much worse "intelligence"
Another way to think about it is Type 1 and Type 2 errors or sensitivity and specificity from statistical testing - do you want an agent that solves any problem but goes off the rails 10% of the time, or do you want an agent that can only solve 10% of the problems but nails them 100% of the time (sensitivity and specificity, respectively)
somewhat off topic but i’ve thought about tapering like this, but i’ve got wondered where can one get a calibrated milligram scale? or perhaps it doesn’t matter too much.
You can also taper off by reducing dosage one day of the week at a time. Thats how I tapered off a 10mg Escitalopram prescription: every 4 weeks, I added another day of the week where I reduced my dosage by 5mg. Took about a year to fully taper off and had to pause for a month here and there to stabilize, but it was worth it.
However, it was equally important that I had a support structure & tools to replace what the SSRI was doing. For me that was mindfulness and therapy.
You can also do a volumetric taper if you don’t have an accurate milligram scale (and none of the ones on Amazon are accurate). Dissolve a larger, known amount of medicine in a known amount of water or propylene glycol, depending upon solubility of the medicine. Then dose the liquid by volume. For example, 1 gram of medicine in 1 liter of water = 1 mg per ml. You can easily dose even microgram amounts this way.
This is completely incapable of measuring individual milligrams. Just because it claims to and shows a digit for it doesn’t mean it can.
Even $500 analytical balances like [0] can struggle with 1mg resolution, but I’d trust them a lot more. Generally you need to throw away the least significant digit of precision with digital scales.
The request may have been misguided. They don't need a milligram accurate scale. They're actually wondering how to taper a 10mg doses drug, which is actually comprised of a much higher than 10mg pill, more like 100 to 200mg. The rest is typically binder.
Also, note that while your $500 scale may be necessary for absolute accuracy, splitting 10mg is not the same as finding exactly 10mg.
There's a few modernist cooking tricks that require small amounts - tenths of a gram - for big impacts. Sodium citrate can make any cheese melt beautifully into a smooth sauce, using something like 1% of the weight of the cheese.
Any juice can take the roll of lime or lemon (in cocktails or baking) by adding a bit of citric or malic acid. Orange juice is 1% citric acid, so add 3g citric and 2g malic per 100g juice and now you can make an orange daiquiri or "key" orange pie.
Emulsifiers get goopy if you use a lot, but can make a smooth syrup in small amounts. You can make a nice syrup from any nut milk with a small amount of xanthan and gum arabic.
There's a wide array of ingredients including xanthan and gum arabic used in tiny amounts in gluten free baking
I sometimes wish my scale was precise to below 1g when measuring yeast for long-rising (bread) dough. A recipe may ask for 1g, and adding 2g will make a huge difference.
Mine was a $30 gold scale from Amazon. Probably not accurate but provably consistent which is what I needed. I'd compare each dose and used a calibration weight to verify.
It seems like you are mostly talking about cardinality, or proofs of problems pertaining to them, and the person you are responding to is saying they struggle with combinatorics
the moment you start talking about opportunity cost, I think you've already ducked the intent of the majority of startup-strategizers-for-exponential-gains-only, for which the carrot on the stick is "unbounded" gain
can you repeat that in a different way? opportunity cost is prioritizing gains - or at least a window of opportunity for outsized gains - I don't see how "ducked the intent" fits in
but yes I realize that any conversation with academics is going to be skewed from the market, I think its worth representing nonetheless since academics often interpret their privilege of choice as a nobler cause because they made that choice at all
reply