I tried... it started with the idea was that log loss might not be the best option for training, and maybe it should be a loss related to how wrong the predicted word was. Predicting "dog" instead of "cat" should be less penalised than predicting "running".
That turns out to be an ultrametric loss, and the derivative of an ultrametric loss is zero in a large region around any local minimum, so it can't be trained by gradient descent -- it has to be trained by search.
When I last looked up the literature, Keto diet was one of the least effective interventions.
That is, if you follow it, I'm sure it works.
But the vast majority of people drop out of keto diets very quickly. So it's lousy advice and an unsuccessful intervention.
It's a bit like saying to a patient "you gotta sacrifice -- you should doing 3 hours a day of cardio". If they do follow through with it, it will work. But the vast majority of people won't be able to maintain doing that.
I started keto in June of 24, lost 50lbs and added a compounded version of Ozempic in November to get through the holiday season with a little extra help. I'm on a fairly low dose, 50mg/week, and it's working tremendously. I've lost another 25lbs up to now and it's about 10x times easier to stick with keto, macro logging, and calorie tracking.
I feel like even with keeping my calories to about 1500/day I'm just fine, and the cravings for sweets and over indulging just aren't in my head.
I'm happy you have found something that works for you but the diet tribalism on this site is getting old. At least it's good to see the initial Keto comment getting downvoted to oblivion.
The epilepsy version is indeed hard to maintain, but can be life changing (increase life quality in epilepsy, bipolar, schizophrenia etc)
The T2D version is way easier. If you studdy it or get a coach, you will know all the pitfalls. But its like therapy, you need to want it yourself. Cant be forced into it.
There's a problem though: it's nice to be paid to consume, but it's not so nice to pay to produce. The return on investment calculus for power generation, especially solar and wind becomes trickier, which means the project financing becomes more expensive. Then fewer projects get greenlighted, despite solar panels being cheaper and cheaper.
Pretty much. The Greek pantheists weren't that different to Marvel or Star Wars fans. The point of the mythology was to have a common identity. If you asked the priest at the template of Demeter "Next spring, I'd like to go and meet Persephone on her journey to home to Demeter, which road does she normally take?" -- they would think you were some kind of fool getting reality and myth confused, while thinking up some mythology about your journey that would turn into a nice play next year.
The modern-day equivalent would be meeting a travel agent at a DC Comics convention and asking them to book flights for you to Gotham City. The best-case outcome is that they write some fan-fiction about you.
This was one of the reasons that Christianity was very disruptive, and exploded across the Graeco-Roman world over the following centuries. It provided a common identity with historical grounding -- if you wanted to go to the temple in Jerusalem where Jesus had kicked the merchants out, you could, and there was no ambiguity or vagueness about which one this happened in even after it was destroyed.
Graeco roman beliefs had historical grounding too. Mt. Olympus is a real place. The pillars of Hades were real. Cities were founded by gods and they existed right in front of people. Offerings would be made and outcomes would happen.
What was really so disruptive about christianity was the aspect of proselytization. That was new with christianity that wasn't really an aspect of judaism. And with proselytization came a need for formal organization of the faith, which served as a useful tool for government to maintain a mandate of power and quell divergent beliefs as heathen or even worthy of crusade, in contrast to synecratic greco-roman paganism.
I think this is not really true. They did all sorts of things… like, these fairly poor (by modern standards) people sacrificed valuable resources to their gods. There is no particular reason to think they believed in their gods any less than current religious people.
I know people with tens of thousands of dollars of Marvel paraphernalia. They spend thousands a year on tickets, events, comics. These people are not well off, it's money they otherwise would do well to have in retirement savings. Humans are not always rational.
They had big rituals that cost them a lot. I could see these as being performative. But then, for something to be performative, the people it is being performed to need to believe in it, right? Like modern generals don’t perform a sacrifice to Iron Man because modern soldiers don’t believe it is necessary.
They also had boring little rituals that weren’t really very effective performative signals.
What reason is there to think they didn’t believe in their gods? It is hard to query what’s going on in the heads even of living people, let alone long-dead ones. But I think the null hypothesis should be that people in the past at least believe their religion as much as modern ones do.
The comparison I would make is Santa Claus. He is not all powerful, but he has a lot of supernatural powers. He makes demands of your behavior (but you don't have to align your whole life around him) that comes with a tangible reward (presents). There are big, expensive and complicated rituals relating to him.
The Christian Scriptures actually include something of a counter-example to this, in Acts 19:35 and following, where an angry crowd is settled by being reminded that the statue in their temple _fell from the sky_:
"When the city clerk had quieted the crowd, he said: 'Men of Ephesus, what man is there who does not know that the city of the Ephesians is temple guardian of the great goddess Diana, and of the image which fell down from (Zeus/Jupiter)? Therefore, since these things cannot be denied, you ought to be quiet and do nothing rashly. For you have brought these men here who are neither robbers of temples nor blasphemers of (y)our goddess. Therefore, if Demetrius and his fellow craftsmen have a case against anyone, the courts are open and there are proconsuls. Let them bring charges against one another. But if you have any other inquiry to make, it shall be determined in the lawful assembly. For we are in danger of being called in question for today’s uproar, there being no reason which we may give to account for this disorderly gathering.' And when he had said these things, he dismissed the assembly."
Sure, but for some reason we are assuming that people don’t believe their religions… if we apply that logic to Christians as well, I guess an excerpt from their book won’t be very compelling.
> The Greek pantheists weren't that different to Marvel or Star Wars fans.
What are you basing this on? (And note, I think you mean "polytheists", as pantheists are people who don't typically believe in gods, but in the idea that everything in the universe is divine).
Polytheistic people pray to the gods just like monotheistic people do. Some believe in more concrete notions of their gods, some in more poetic ones, and both co-exist in the same societies. Just like many Christians believe Jesus existed, lived, died, and became physically resurrected, so do many modern-day polytheists, and many ancient ones as well.
And beyond the specifics of the stories, people most of all believed and believe that performing or not performing certain rituals will attract the benevolence or ire of their gods. They perform rituals to attract the rain, or to bring good luck in battle, or to bless their crops. They try to put curses on their enemies or competitors. These are all real beliefs that exist today, in both polytheistic and monotheistic religions, and that have existed since the dawn of humanity based on everything we know.
And one clear proof that people truly believed and still believe in the importance of these things is the significant resources they are willing to invest in them. Sometimes they directly perform sacrifices, sometimes they give money for the building of altars, sometimes they sacrifice their time or enjoyment towards these goals.
> i think plausibly being able to use youtube video as training data was the major reason for google to buy youtube in the first place
When I asked Eric Schmidt the "why did we spend so much money on buying youtube?" question his answer was "if it's the future of television, it was a bargain; if not, we overpaid."
There didn't seem to be any expectation among senior management at the time that it was anything other than "televisions carry advertisements, we want in on that market."
What's your sales experience? If the answer is "none", then you'll discover that bootstrapping is not nice at all.
---
1. Services business, had a few clients that I knew would say yes.
2. None. But trying to pivot away from services to products -- dozens, none successful.
3. When you have a good pipeline of clients, it's great. Then it can all come crashing down to zero in a matter of weeks.
4. I want to, but now I'm essentially unemployable.
5. Solo.
6. Found myself back in academia of all places, and simultaneously advising large corporates on how to make use of language models.
I've got a list of 200+ things that I've accumulated since 2019 that the world seems to need, or that someone has said that they needed. It's a very raw list with ideas that range from the trivial and unresearched "why doesn't (thing) exist?" through to ideas that I've worked on for some period of time before I've realised that I need to hand it over to someone else and haven't yet met that someone. (My email is in my profile if anyone wants a copy.)
One of the things that has frustrated me is that there are many ideas that are worth pursuing that could make a big difference to the world, but can't be done in a way that is likely to be commercial viable -- and so can't be solved by starting a business.
If I had to provide guidance, I'd say "pursue a niche". Find something small that's obviously wrong and do something about that. You're more likely to have an impact than if you try to tackle some big headline problem. There will always be someone else more qualified and capable of handling that big problem: but there will always be too few people looking at all the small problems of the world -- and in aggregate, the small problems add up to more than any big problem.
That turns out to be an ultrametric loss, and the derivative of an ultrametric loss is zero in a large region around any local minimum, so it can't be trained by gradient descent -- it has to be trained by search.
Punchline: it's about one million times less effective than a more traditional architecture. https://github.com/solresol/ultratree-results