There was literally a vaporware "AI" hype cycle in 1960. Propositional logic programming was poison to investors for 50 years because of that one, just like LLMs will be poison to investors for 50 years because of this one.
Why do articles like this always say things like "I've used LLMs to get some stuff done faster" and then go on to describe how LLMs get them to spend more time and money to do a worse job? You don't need LLMs to frustrate you into lowering your standards, the power to do that was within you all along.
Much of this feels like when they did studies on people who take mushrooms for example feel like they are more productive, but when you actually measure it they aren't. It's just their perception.
To me the biggest issue is that search has been gutted out and so for many questions the best results come from asking an LLM.
But this is far different from using it to generate entire codebases.
The problem with this kind of plagiarism isn't that it violates someone's specific copyright.
But the discussion around plagiarism calls attention to the deeper issue: "generative" AI does not have emergent thinking or reasoning capabilities. It is just very good at obfuscating the sources of its information.
And that can cause much bigger problems than just IP infringement. You could make a strategic decision based on information that was deliberately published by an adversary.
The official definition of AGI is a system that can generate at least $100 billion in profits. For comparison, this would be like if perceptrons in 1968 could generate $10 billion in profits, or if LISP machines in 1986 could generate $35 billion in profits, or if expert systems in 1995 could generate $50 billion in profits.
This sounded a strange abstract way to define it, but you're right, in as much as Open AI and MS deciding this between them. I don't think they mean it in a general sense though, it's framed to me as a way of deciding if OAI have been successful enough or not to MS on their investment.
> Microsoft and OpenAI have a very specific, internal definition of artificial general intelligence (AGI) based on the startup’s profits, according to a new report from The Information. And by this definition, OpenAI is many years away from reaching it.
Norwegian citizens found that they could vote themselves money and it worked out fine. It's almost always a better strategy than voting to give money to oligarchs.
I'd argue that rather than vote themselves money, they voted for sharing the exceedingly plentiful (relative to population) natural resources of their country.
But yes, that does seem to be a version of "vote yourself money" that is far less problematic than other instantiations.
Ironically the result of all the cuts will likely be a reduction in the size of the economy - which can actually lower the tax take too (a portion of the money flowing around the economy comes back in taxes) so the combination of that will likely be an increase in the US's debt to GDP ratio and reduced living standards for most people.
That is not to say that Governments can spend to excess with no consequences of course, but that these indiscriminate cuts are likely going to have the opposite effect as intended.
There may not be a correlation, but you can clearly see that the bottom-right quadrant of the plot is basically empty, which is an important insight.
A more accurate aphorism would be "You can sell good kebabs anywhere, but you can only sell bad kebabs near a train station."
And if you look at the "minimum viable quality" instead of the overall quality, there does seem to be a linear correlation with the distance. You can use a 5% quantile regressor to easily find the lower edge of the distribution.
> but you can clearly see that the bottom-right quadrant of the plot is basically empty, which is an important insight.
I don't think so? It's mostly a result of the fact that (obviously) the best place to sell food is where there are people, which is also the best place to put a metro station. So on average the kebabs are pretty good and on average they're near a station. In Figure 9 one of the worst reviewed restaurants is over 3km from a metro.
You're likely seeing a pattern where there isn't one, which is normal for humans.
There's one obvious place to go around here for a good kebab, it's a few minutes walk to the station, but the way you can tell it's the best place for a kebab is how late it's open every night. Long after other kebab places are dark they're still doing enough business to justify remaining open.
The best place for pizza in my city is very close to a train station but that's a total accident, they park (it's a van, no really, best pizza in the city but they hated owning a restaurant so they put their oven in a van instead) in the car park of a railway station's pub about five minutes walk from me.
> Long after other kebab places are dark they're still doing enough business to justify remaining open.
Is that a quality signal, or just a sign they don't mind selling to people going back from parties, in various stages of being drunk? I always assumed the latter. Few restaurants (McDonald's and KFC aside) want to work those hours, so whichever does is almost guaranteed a steady trickle of customers who literally have nowhere else to eat (other than home). There isn't much pressure for quality in this situation.
The KFC next to them shuts long before they do. The nearest McDonalds (a drive through) is 24/7 but I've been there late at night and it's extremely quiet. Moreover neither sells kebabs, whereas plenty of places which do sell kebabs in this part of the city close earlier.
However, thinking about it more carefully, while I've never bought a kebab from them technically the Chaiiwala which is 24/7 does sell kebabs. They're a bit fancier (and of course, more Indian) than the kebab you'd get from the kebab shop but that's definitely a chicken kebab. Their clientèle in the middle of the night are a mix of "gig workers" and people either going to or coming back from prayers (for whichever of the religions is into praying when other people are in bed - Islam and maybe others?). I have never seen drunk young people in there, but it is open 24/7 so that must happen once in a while.
> but the way you can tell it's the best place for a kebab is how late it's open every night
That logic definitely does not apply to SF Bay Area. Most of the places that stay open are all pretty meh. A few pizza by the slice places, Dennys, Grubstake, Orphan Andy's, Mel's. Oakland and LA (SoCal) are no better.
AFAIK most places have no interest in staying open late. Maybe they don't want to stay up. Maybe they don't want to deal with drunks. So, given there are so few, the few that are open have no competition.
I been to / lived in cities that actually have good late night options. Tokyo, Hong Kong, Kuala Lumpur, Singapore. California has a curfew which doesn't help.
Late is a tradeoff. On resteraunt can make a killing serving anyone out but there are not many so two staying open late both go bankrupt from lack of business. Or maybe the area can support two (3? 10?) I don't know the real number but not as many as the daytime lunch crowd.
> "You can sell good kebabs anywhere, but you can only sell bad kebabs near a train station."
Insightful!
97.38% of bad studies measure the wrong variable.
Do drunk french people buy kebabs? In my city one central late night kebab place has great kebabs. Anecdotally I remember one great kebab cart serving at least one drunken customer in Nice (France) - not near a station and a long way from the Paris metro!
I think there's some population selection flaws. Drunk people don't leave reviews. In foreign countries it is difficult to know the correct search term.
I suggest an alternative study: how much lager does it need to make a train station kebab taste great?
Source: lived in France for two years. Bought a lot of kebabs. Drank with French people a lot.
Also, I really miss French kebabs. They use the thick pide bread, and harissa sauce is always available. Also, if you order an "American" one, they put fries in it.
Wonder whether that's an instance of Berkson's paradox.
Basically nobody bothering to report about Kebab shops that are neither good nor conveniently located. Probably not, but "one quadrant empty" is a handy red flag that you might encounter Berkson's.
The best way to get the right answer from an LLM is not to ask it the right question; it's to post online that it got the wrong answer.