They would need 250 million GPT Plus $20 subscribers to recoup a $5 billion expense. They're far from that even when we count the free users (which are likely 99% of the user base?)
The math just doesn't work. They're hemorrhaging money as far as I can tell (not counting the Azure computing deal).
We can only guess, but my guess is that inference is still a good chunk of their costs. That's why they're trying to get the mini/turbo models into a usable state.
Even then, training is still an expense. And it's not like you can train and forget. Even if your model is already trained you still need to incorporate new knowledge over time.
I don't think so, I think it's more about openness. I've noticed older software engineers tend to be more anti-LLM and quick to dismiss.
The shortcomings are aplenty, but they don't bother me. The things it can do weren't possible 2 years ago. I'll leverage those and take the bad with the good.
Similar experience with Tesla FSD. I know other Tesla owners who tried it a few times and think it's trash because they had to disengage. I disengage preemptively all the time but the other 90% of my drive being done for me is not something that used to be possible. I tried to give up my subscription because it's expensive and couldn't hold out two days.
This statement doesn't even make sense, where are you acquiring this knowledge if not from experts? Are you out there doing field research and conducting your own scientific experiments on every subject you're interested in?
I'd love to hear a couple examples of which "expert opinions" you've disproven for yourself and where you acquired the supporting evidence.
You are confusing opinions with evidence and facts. Obviously you acquire knowledge by reading papers where people are doing research and experiments and presenting evidence, analytics and all the facts from which you can make your own conclusions.
OK, I don't think that this delineation is actually as black and white as you're making it out to be, given most subjects are complex enough to require nuanced interpretations of data/facts, but I'll give you that in some instances there is a fine line, and there are certain areas where there is a lot of disagreement even amongst experts.
Still, an expert, by definition is just someone that has "comprehensive and authoritative knowledge of a subject." You could argue that as soon as you've acquired enough knowledge on a subject to form an accurate opinion, you have yourself become an expert.
Although I'm guessing when you're referring to "experts" you mean "establishment experts." For example, your average doctor whose spent decades studying medicine as opposed to your online research on how to best treat/prevent a certain ailment (you didn't provide an example so I am just referring a fairly common one). What's the difference between your expertise and theirs? On one hand, they have decades worth of rigorous academic study and personal experience over you, on the other hand, you may have a "fresher" perspective that may be devoid of certain institutional biases. I would just strive to stay humble.
I may have looked more in depth into my condition (though that may imply listening to crackpots) than doctors who despite years of study may not be as in depth as me. There is a lot to the entire human body (which is why doctors refer people to specialists, but they need to choose the right specialist which isn't always easy)
>And what % of patients infected actually end up in the ICU vs. just have a sniffle and a cough for a few days?
This data is easily available: ~20% of people who get the virus are in life-threatening condition. Thats a 1/5 chance.
> I'm not saying "do nothing", but I am saying "shut down the world" is too extreme in the opposite direction.
Experts disagree with you. Maybe you should ask yourself why?
1. You have a hypothesis that taking less drastic measures would result in a manageable death rates and less impact on the economy. However, Wuhan and Italy tried to deal with the virus through these less drastic measures (telling the elderly to self-quarantine, asking the population to social-distance), and the results were disastrous. Does that not invalidate your hypothesis?
2. Even if you're right, and there was a good chance we could avoid the worst of the epidemic, we know for a fact that the worst case scenario is that in the US, 21 million people will require hospitalization, and 1.7 million people will die. Saying that it's worth taking a risk on less drastic measures is the equivalent of saying it's OK to play Russian roulette because the odds are actually really good: it totally ignores the tremendously high cost of the worst case scenario.
I have found that risk-analysts like Taleb have the most convincing arguments for why all these extreme measures are called for. Check out his twitter: https://twitter.com/nntaleb
My point (in the other thread, but this is going the same way) was that yes, if someone says let's have 100 people of all ages play Russian roulette (only the gun has 30-50 capacity, and also loaded guns mostly go to much older people), that is very bad.
But if the alternative is, we will instead take those 100 people and make them really miserable, make a few homeless and/or drug addicted, maybe have one commit suicide, starve a few of them, get a few divorced, and ruin a few lives completely - disproportionately affecting the young and children - let them play Russian roulette.
I am not convinced, but at this point to me it's looking more and more like that's the case.
Politicians (in NZ) are already saying it's going to be worse than 2008.
-
The cost of the Soviet union collapsing (a purely economic disaster I can really relate to, even though I do think Soviet union disappearing was a great thing as such) in Russia was 3-5 years of life expectancy, over a decade, for EVERYONE. How many lives is that? Not counting the missing births and the knock-on effects of both, as well as the total waste of life that the 90ies in Russia were for many people.
You are taking my Russian roulette analogy way too literally. The "bullet" is not literally just dead people, it's also the economic impact of letting this disease run rampant, which you are clearly not accounting for.
We aren't just shutting everything down to save lives, we're shutting everything down because we some reason to believe that if 70% of the population gets this disease in a very short amount of time 20% of them are hospitalized, and at least 2% die (in truth, the CFR would skyrocket if we don't flatten the curve), the societal and economic havoc would be much worst than the one resulting from these government imposed quarantines. At the highest end of the risk spectrum, it could be orders of magnitude worst. Do we know that for sure? No, hence the Russian roulette analogy: some chance of extremely high risk, esp. when it comes to the entire planet: NOT WORTH GAMBLING.
If you think that world leaders are currently sacrificing the economy just to save lives, you are not paying attention to what they have been doing.
> This data is easily available: ~20% of people who get the virus are in life-threatening condition. Thats a 1/5 chance.
No, it's not "easily available". There's data flying around everywhere from different countries as this situation rapidly evolves and you can torture it to tell any story you want. If the data is so easily available, couldn't you have at least linked your source?
> Experts disagree with you. Maybe you should ask yourself why?
The "experts" in this case have never dealt with a global pandemic of this scale before. So they are inexperienced experts giving it their best guess based on what they know about disease transmission and math. But they clearly did not factor in the effects of shutting down the global economy for 3-6 months in their quest to optimize for a single variable. I'm sure we'll have a lot of "lessons learned" after this is all over and the experts will be much more experienced the next time around...
True, this is only in China. In most other countries, it's still too early to be able to collect great data because of how quickly the situation is evolving (the denominator is bound to be wrong due to delays between cases being detected and the amount of days it takes before cases evolve into severe/critical condition or death). But it's data, and you can't easily dismiss it.
> So they are inexperienced experts giving it their best guess based on what they know about disease transmission and math.
This isn't entirely true. The epidemics in China, South Korea have more or less been resolved, and many of the steps being taken in other countries are based on comparing the outcomes in those countries compared to ones that are failing to slow the spread of the disease with disastrous consequences (Italy).
> But they clearly did not factor in the effects of shutting down the global economy for 3-6 months in their quest to optimize for a single variable.
It's possible that the health experts are not taking the economical windfall into account, but the politicians enacting the laws probably are. These decisions are being made in rooms with people who, with their knowledge pooled together, most likely have more information about the disease and the economic impact of their decisions than you.
Secondly, the reasoning for national shutdowns earlier rather than later is to shorten their length. As of now, local government officials are hoping not advocating 3-6 month shut down, but rather trying to avoid one by shutting down now. I remember reading the figure of 2 weeks but I cannot find it quickly now. We will see.
You can't just cherry-pick South Korea's case fatality rate because it's convenient to your argument. The case fatality rate worldwide is currently 3.7%. Regardless of what the reasons for that may be and whether it will remain the same, that is a cold, hard, fact about the current reality.
I get it, some jobs are more desirable than others, and there has to be some reason for people to take the less glamorous job, but that barrier could just be skill. I support hard work when it comes to training. Hey, you want to make video games, well it's a really competitive field and you better train your ass off in order to be the best at what you do. I think generally, leveraging passion in order to get people to work really hard is good. They get fulfillment out of it, and everyone else gets a great product.
But that's not what this is. This is investors and executives who found a money pump and are pumping as hard as they possibly can without any regards for the health and wellbeing of the human machine they're putting pressure on. At some point, you have to give people the time to live their lives outside of work or the world is going to become a worse place. They become stressed and depressed, and it affects their families and friends, and and then has a ripple effect on society.
I agree with you completely, I'm not arguing that these studios are behaving ethically. I'm just pointing out that the fact that videogame studios have bad working conditions has been public knowledge for a long time and the engineers who apply for these jobs have to be aware of it. Yet they still decide to get into these jobs even though I'm certain that the vast majority of them could land a more comfortable and probably better paid developer job elsewhere.
I realize that I'm playing the "blame the victim" game and I'm not entirely comfortable with that but in this case it reminds me of that guy who knowingly got in touch with a cannibal to get eaten. If you know what's going to happen and you still on your own free will and without external pressure decide to continue on that path can you really complain that things are exactly how you knew they were going to be?
Are these people coerced in any way to work for these game studios? Were they lied to? Didn't they know that AAA game development is a thankless, soul-crushing task with very tight deadlines and terrible work ethics? Because I remember having almost exactly this conversation with a friend at school more than 15 years ago when he told me he wanted to become a game developer.
Let's say you manage to carve out 2-4 hours a day to work on your side project while remaining a decent person to your wife and kids. Leaving your job means you could spend 11-13 hours a day on it instead. That's an enormous difference and could easily mean the difference between failure and success.
Not only that, but you are putting yourself in precarious legal territory by bootstrapping a side project. Your employer could easily claim ownership of all your IP.
OP was a senior software engineer at AWS, how difficult do you think it would be for him to re-enter the workforce with a competing cloud provider nor even coming back to Amazon?
Before you draw conclusions about all homeless people based on the the ones hanging out in Pioneer Square, I challenge YOU to go to a homeless shelter like Mary's place and learn a bit about the people living there. My fiancee is a doctor in the Central District and deals with homeless people every single day. A huge number of them are working families with children who work in the city and can't afford to rent, commute, or find a job in a cheaper area.
> This isn’t popular, but...
Sorry to break it to you, but you're not the independent thinker you might believe you are:
"Drug and alcohol abuse tops the list among the general public as a major
factor why some people might be homeless. More than eight in ten (85%)
adults feel this is a major factor. "
"Mental illness or related mental disorders such as post traumatic stress
disorder are cited by two-thirds (67%)."
