(I'm just an idiot writing things at 1.30 AM on a Saturday, local time.)
I was about to reply to you with a ridiculously long message using the example from before about the lifecycle of a user story, but I bored myself halfway.
So I'll try to extract the gist:
Without human assisting, based on my humble experience, there is always something that wasn't accounted for.
Even if my company is integrated with all the tools and you have infinite tokens, there's always some microservice that wasn't considered or a feature that is completely incompatible.
Or there's a MUCH simpler way of doing it.
I'm also noticing that there are non-technical people who, with the help of AI, are sometimes able to reason better about the bigger picture than some engineers.
A colleague of mine was telling me the other day that the "functional" role might eventually become unnecessary. With all the time saved by writing code faster, whoever actually understands what's happening underneath could just do it themselves.
It sounds a bit extreme to me, but I understand what he means.
To get back to a slightly more serious tone, when it comes to programming, for me the huge thing is: how it should be taught nowadays ?
And I don't know the answer, because my experience started when I was 12, writing a little game in C, and it took me years before I could make something that compiled decently.
If I had instant help for every thought I had, would I have learned why memory is allocated in certain ways or idk what else?
I don't know, maybe learning faster is better because you learn more ?
Anyway fortunately it's not my job to figure out the answer :'D
In my case, I dev on macOS. The env the agent runs in is the same as my dev laptop, configured and in sync. Has access to all the same tools and environment as I would on my laptop.
From the examples given in this article it is absolutely impossible to tell whether the "incorrect beliefs" (be it racism, mushroom aliens, vitamins curing cancer, whatever) are in any way connected to the Nobel prize.
For all I know, they all could have been racists, psylocibe users, etc. long before they received any prizes and the associated attention.
In fact, it could even be possible that they had much worse and radical beliefs earlier in life which were tempered by the burden of being a thought leader.
I'd go as far as to say that this is a good example of an article that should be removed from Wikipedia.
Take any other award, list people who got it and had "incorrect beliefs", add an article about it and preamble it with "it is unclear whether it is an effect of the award". What purpose does it serve? Do you think there should be articles for all of those "diseases"?
The article mentions it later - journalists and laymen end up asking Nobel Prize winners their opinion on matters outside their area of expertise.
We don’t do that with Fields medal winners. Mainly because most people don’t even know what a Fields medal is. I can’t write a headline that relies on Elizabeth Moon’s or Stanislav Smirnov’s borrowed authority because the average joe doesn’t know who they are. But that definitely happened with multiple Nobel prize winners because the Nobel prize is that well known. A Nobel prize winners opinion on race science is inherently newsworthy.
You know how we know the Nobel is in a different league to these other awards? Because when you explain what a Fields Medal is to someone who hasn’t heard of it, you say “it’s the Nobel prize for math”.
And the fame of the prize leads to attention, which in the words of Milton Friedman, is “flattering but also corrupting”.
The difference is notability, which is the criteria used whether a distinct Wikipedia article is warranted. There are multiple cited sources that use the "Noble disease" term, there are possibly much fewer similar publications with "Fields medalist disease".
That is not to say that the theory holds water. Flat Earth theory is also notable and has a wiki page, even if blatantly wrong. There is no hyperboloid Earth theory wiki page.
> there are possibly much fewer similar publications with "Fields medalist disease"
Among mathematicians, I wouldn't claim that "Fields medalist disease" is an unusual term, though this term is typically meant to mean "some mathematician got a Fields medal, but did not prove any other deep mathematical result afterwards". There also exists the much more derogatory term "dead wood" for mathematicians who once proved highly celebrated results, but after they got tenured, no they didn't manage to prove anything deep anymore."
Concerning the "crackpot undertone" of "Nobel Disease": lots of mathematicians know the names of achieved mathematicians who got into some eccentric ideas ("eccentric" in the sense of "outside the Overton window") once they got older.
Famous examples for the latter point are Serge Lang
Ok, notability is a good argument for having such an article (although I'm not too sure about it really being notable), but the current shape is absolutely misleading.
No, I do. I truly believe - especially after seeing Mythos results at work - that the only way is to stop and destroy it all before it destroys us. In fact, it's already so bad that I'm moving completely offline all the important stuff that I care about, hoping that maybe we will turn back at some point. Otherwise we are doomed.
If someone makes a specialized hardware just for the unrestricted Mythos-class model to reduce the cost and increase the speed, we are going to be completely defenseless.
How is it bad faith? Because I don't believe in "we care about safety" words coming from people not just demonstrating that they don't care, but even bragging about it?
Anthropic deleting their models does nothing for AI safety. The rest of the industry will just fill the gap, probably with less consideration to ethics than Anthropic has today.
Or maybe researchers and engineers everywhere in the world, emboldened by such unprecedented move, will just refuse to participate in burning the world?
"If we don't destroy the world, someone else will" is such a weak defense I'm speechless.
Yeah no, most researchers and engineers do want to bring about the singularity, and they believe they can do it safely at companies like OpenAI and Anthropic.
reply