From the post he's referring to text input as well:
> Maybe it makes more sense that all inputs to LLMs should only ever be images. Even if you happen to have pure text input, maybe you'd prefer to render it and then feed that in:
Italicized emphasis mine.
So he's suggesting that/wondering if the vision model should be the only input to the LLM and have that read the text. So there would be a rasterization step on the text input to generate the image.
Thus, you don't need to draw a picture but generate a raster of the text to feed it to the vision model.
I’m normally skeptical of claims like this, but looking at the examples it seems that Sora is reproducing some of its training data verbatim. I guess it’s a case of overfitting? In particular the Civ example seems like it must have been copied almost verbatim.
I agree with the sentiment but I want to point out that a car is not essential for most people living in SF, although many people outside the city think this. Around 35% of households don’t have a car: https://www.sfmta.com/sites/default/files/reports-and-docume...
Today I learned! I remember it being sold to a Chinese company a few years ago, didn’t know it went back to US ownership. Ironically just deleted it because even with an Adblocker it’s still unusable with all the bots
The median household net worth in the US is $193,000. The rate of home ownership is 65%. I don’t think the median American is at risk of becoming homeless during normal unemployment. Maybe you mean that the article isn’t as relevant for a global audience which is fine, but I would think that the median American lives in “the real world”.
It doesn't say liquid assets though, it mentions multiple options which vary in how easily you can/should use one to pay an unexpected bill or period of unemployment.
For example, if one is using their life-insurance payouts to pay their rent... well, something has gone very wrong somewhere.
Specifically, this part:
> financial asset—which includes transaction accounts, certificates of deposit, savings bonds, other bonds, stocks, pooled investment funds, retirement accounts, cash value life insurance, and other managed assets
For the highly-liquid "transaction accounts" (checking, savings, money-market) the conditional [0] median is just $8k.
[0] AFAICT "conditional" here means "we don't include $0 data points in the median." That explains why the subcategory of "stocks" has a higher conditional median value than the more-general category of financial assets.
Sure, I couldn’t find a better data source than that. If you find a better source that includes liquid assets specifically that would be helpful. I am skeptical that of the 39k in assets listed there, there isn’t a substantial amount that can be used to pay the bills (i.e. who has $39k in their 401k but $0 anywhere else?).
It seems some "Survey Data" files are publicly available [0], but without investing a bit more effort in parsing them (which is definitely not happening on my phone) I'm not sure if they have the kind of household-by-household data that could be used.
Again I don’t think that means what you think it means. It says “value of all financial assets” which includes banking accounts, cds, stocks, bonds etc. that is not liquid.
The next exact line it calls out those bank accounts and the mean at $8,000. That is not a lot of liquid emergency savings for a household let’s say of 3.
Again that number is inclusive of retirement accounts, hard to say much without knowing the breakdown but we do know $8k is the closet cash number. I would suspect the majority of the rest are in retirement accounts. Sure bonds and stocks are liquid but not like cash is. Using Fred is not that useful for these type of exercises as I don’t think those datapoints do a great job on topics like this.
So I don’t see what you are trying to point out. Most surveys I have seen put something like a quarter of Americans without any emergency savings.
Good luck accessing any of your home equity if you don’t have a job. I guess you could just sell your house of many years and move your family into an apartment.
Until you factor in maintenance, utilities, taxes, etc…
Depending on where you live, renting is often cheaper than owning, but real estate agents who make money off selling houses will never tell you the full cost of owning a home.
reply