I just asked MetaAI to help me with Wordle and it understood and gave me a sane answer, so…
Edit: ah, I spoke too soon. My first question was too "easy" but I asked a few more, and sure enough... it can understand what I'm asking and it can write an answer that's well formed, but it's fundamentally not understanding the rules of the game or giving me valid guesses. Cute!
It's actually that it has no idea how words are spelled because they're blind to letters. This is because they operate instead on tokens instead.
They do seem to know the first letter of each word pretty well (they love to alliterate and can also produce a list of things starting with A then B then C etc) but they are all very terrible at spelling beyond that. I presume they could be trained to spell even while retaining tokens but I guess I don't know for certain.
With tasks like these it helps if you break your words up with some kind of separator so that it all tokenizes to one character per token. They do know what individual letters are conceptually.
It helps a little. ChatGPT4 still fails the following prompt fairly often (maybe 40% of the time):
which of these 5-letter sequences a has P in position 4?
P U P I L
S T A P H
C R E E P
P O O P Y
L I P I D
It usually gets it if it reiterates each sequence before deciding (but not always) and it almost always fails when it just answers right away.
And this doesn't actually help at all with the actual related Wordle problem - "think of a 5-letter word with P in the fourth position".
You could try filling up its context with a list of a thousand of the most common 5 letter words all spelled out (and I've done so, even with additional hints like the positions of the letters in parens). But it really didn't help (as might be expected since it can't even reliably do it with the above list of 5 words.)
Usage will continue to grow as it gets better. And it got a lot better over the last months.
And then there is Waymo which proves that with geo-fencing, high resolution maps, additional sensors and remote operators on standby, robotaxis are already possible today. Tesla could add some of these too, if it wants to accelerate meeting regulatory requirements.
> Tesla FSD is already being used for on a billion miles per year. Search for "Cumulative miles driven with FSD" [...] Usage will continue to grow as it gets better. And it got a lot better over the last months.
There is a big, hell a gigantic, difference between that and robotaxi. You need the car to be able to handle from start to finish, the entire trip, not parts of it, not even just 99% of it.
Also, the car cannot stop or be unable to deal with some situation and require help. If owners randomly have to go help their car miles away because the "taxi" got into a unknown situation, they're quickly going to get out of that deal.
And then you have all the regulatory and insurance issues. Is Tesla or the owner insuring the commercial part ? Are they required to get commercial license ? What happens if the car hit and kill someone during a robotaxi trip, who is responsable ?
Waymo has employees watching their cars and resolving issues when they come up. So in order for Tesla to do robotaxis like that, you will now need to pay Tesla do to the monitoring and issue resolving as well. And then of course you will need your own special business insurance without the bulk purchasing power of Waymo/Cruise. When your car gets into an accident, you don't get to replace it with one from Tesla's fleet like Waymo does, you are just SOL.
1: Tesla can use their own cars. They dont't need to base the service on the cars owned by individuals.
2: If they want to rent cars from individual owners, they can make the conditions as comfortable as needed. Monitoring and insurance for example would most likely handled by Tesla and not be paid by the individual owner. The owner would just get a fee for every hour Tesla uses their car.
Classic symptoms of anxiety from the sleep deprivation, no?
I've also found that there's a lag on recovering from sleep deprivation, where the first night of good sleep often leaves me feeling foggy and lethargic the next day. After 2 or 3, that energy and positivity returns - not sure if that's idiosyncratic or not.
I feel exactly the same way. The first night of real rest feels as if my body “caught up” to how sleep deprived I was and I often don’t accomplish much that day.
And the gradient of sentencing for the convictions -- defrauding people of $1 gets you less jail time that defrauding people out of $100k, which gets you less jail time than defrauding people out of $100M.
Fake numbers but say you shot and killed someone -- the law (and related sentences) can accomodate a wide variety of scenarios which will lead to a wide variety of punishments. E.g. if you were cleaning a gun and it went off, you could be charged with involuntary manslaughter and face 5 years, or if you were in a fight and pulled a gun you could be charged with voluntary manslaughter and face 10 years, or if the fight seemed avoidable it could be a 3rd degree homicide charge facing 15 years, or if you shot them will robbing a bank, it could be a 2nd degree homicide facing 20 years, or if you planned an assassination and shot someone, it could be 1st degree homicide where you would face life.
There's a binary "guilty / not guilty" but on the guilty branch, each of those sentences has enhancements that will determine your length of incarceration. Was this your first crime? Were you on drugs? Was the robbery over $5 or over $50k? and so on and so on.
What? VLSI, VHDL, Multics, Hypertext, GaAs semiconductors, autonomous navigation, GPS, and packet switched networking were a few things they funded. Totally absurd, right? For every successful tech, they probably explored scores that didn't pan out. Basic, high risk research has huge payoffs across society even if you don't like who's funding it or why.
So true. I read once that fundamental physics research on quantum tunnelling in Japan resulted in antennas that immediately reduced satellite dish sizes from meters across to the size of a dinner plate. Now you can connect with a modern smart phone.
"Guns and abortion and corporate speech" aren't even the cool cases. These are. You just don't hear about cases like this because you're supposed to be angry at the world, instead of appreciate how cool it is.
Edit: ah, I spoke too soon. My first question was too "easy" but I asked a few more, and sure enough... it can understand what I'm asking and it can write an answer that's well formed, but it's fundamentally not understanding the rules of the game or giving me valid guesses. Cute!