In that case, I'm glad AI could so handily improve collaboration for frontier mathematics. Progress should not be held back by a lack of willingness or ability to cooperate.
Processing results takes time. Sometimes months or years. They can meet up at conferences to discuss, publish their partial results, and the mathematics community can come together to understand and piece the true answer together. None of this requires crotchety old men hoarding results, trust me, those people are smart enough not to put that in AI. They made a small result that is highly likely dependent on some input they made to the AI. The AI collates everyone's work and puts the cherry(s) on top to finalize it. Of course there is work done by the AI that only requires the sum of public human knowledge as well. But you seem very angry and accusatory at researchers, like they're some secretive cabal hoarding gold, furious at the new "give everyone our gold" machine. The people that do this science literally live for sharing and speaking and disseminating this knowledge.
Have you every used AI for work on truly cutting edge at least graduate level academic research? It's nothing like writing software, i do both professionally. Every model I've had access to requires a LOT of hand holding to get it to do the right thing. No model is getting set up with a single prompt, told to go, and then solving a millennium problem (unless they've explicitly stated that, which i would still doubt). It takes a lot of guidance to get these AIs to solve any serious problem. It's not outside at least my reach to believe that openAI is taking these nudges from intelligent people and combining those with the internal, non public models to make great press releases as it tumbled towards an infinite debt wall. I don't think there's a thing they wouldn't do to save their company.
I think you're wrong, looks like models are just finding and writing up publishable papers on novel results now just by giving them a single prompt and telling them to do so.
"I would like you to write for me a short article that could be published in Physical Review Letters. I would like you to choose the question, do the analysis and write up the results. The constraints I put on the question are 1) that it should be in the hep-th set of topics (so that I have a hope of evaluating the arguments and results); 2) the answer can be written up in a short format of 5 pages or less; 3) the result is of significance enough that it can be published in PRL. I want to break up the task into several steps. The first is to find a good question, so give me three candidates. I’m also going to involve Claude in the process. Claude is going to find three candidates too, and the next step will be to evaluate each others’ candidate questions."
I think part of what's getting the goat of some mathematicians is not that OpenAI is getting results, but that OpenAI is telling mathematicians that priority isn't something they care about. The indifference is more threatening to the current traditions of math than if they were actually fighting for priority. The indifference gets reinterpreted as them trying to "steal glory" but that's not what they're doing. They're bulldozing glory.
A few bucks will realistically last you forever.
I also bought a DGX Spark at launch. I've learned a ton. but it's not usually my daily driver. I do run overnight tasks on it, like "read this book and write up notes on each chapter..."
In retrospect, given the dirt cheap prices of apis and or renting hardware, buying hardware for inference was likely a mistake (but not enough of a mistake to sell it now).
I think building a OpenAI compatible RAG-pipeline that also perhaps includes the anthropic sdk let's you have the best of both worlds. You can build it in a way that a local agent can escalate to a frontier model when needed and also substitute prompts data that you want to hide from being "public". For example. I think that is meant by harness here.
And with some smart feedback loops, you will see that for home assistant type tasks local AI can pretty quickly fulfil all requirements. For quick responses or voice analysis you might want to start with an external provider.
Otherwise, I think for complex home automation around 24-32gb vram are a realistic house number, but more experienced users will probably need less here.
People say that tools are called by the model. That isn't entirely correct. The caller calls the tools, the model just formulates the syntax for it. So the model part can be externalised without you needing to compromise you system fully (just to the degree the tools would allow anyway)
> AI's are trained with existing information, so obviously whatever he expected to discover was already out there.
I've wondered about that myself. Does the fact that LLM's were able to solve these problems so easily indicate that most of the solution already existed, scattered in the training data, and the LLM was simply able to put together the pieces?
I think it's an interesting conjecture but nothing more without proof.
> See the recent Cornell University gangrapes for proof that authorities can choose to simply not investigate.
In the interest of accuracy, according to the DA, they did investigate in 2024 and did not prosecute because:
"Jane Doe’s sworn statement in November of 2024 did not allege that she was drugged against her will or gang raped. On the contrary, Jane Doe’s statement described her participation in drug use and sexual conduct as voluntary, conscious, and consensual."
> Project 2025 fairly implicitly stated that the idea was to scuttle trust in the US government so badly that there would be a case for private equity to swoop in and save the day. That's what you're seeing.
Could you quote the exact text you're interpreting here?
(Not the whole Project 2025 text; the specific sentence(s) that imply that.)
It's a big drop, but most of it happened in 2021-2023, when reserves fell by roughly 300K from 650K to 350K. The SPR only recovered a little bit before the recent 130K drop from 415K to 285K.
What was the SPR being used for a few years ago?
Edit: Whoever downvoted, would you be willing to explain why?
Some medications are designed for one purpose, then other effects are discovered in practice. Gabapentin, for example, was designed as an anti-seizure medication structurally similar to the inhibitory neurotransmitter GABA.
Now it is primarily used to treat neuropathic pain, and the mechanism for that is not well understood. The GABA receptor is not involved. This effect is just a happy accident, and nobody really understands why it works.
> Developer compensation talk always gets biased toward top companies and high cost areas. The median developer or even top 10% developer isn’t living in the Bay Area working at FAANG, but that’s what gets talked about.
That's what should be talked about in comparison to doctors. A FAANG position is comparable to the level of effort, competition, and accomplishment involved in becoming a doctor.
On the other hand, the "median healthcare professional" (counting only those requiring a college education) is a registered nurse, and that's the healthcare profession most comparable to the median developer in education, competition, and income.
I disagree with this. It's more like "Doctor at the Mayo Clinic" or Johns Hopkins.
I live in a capital city of 60,000. Unemployment 4.5%, median household $79,000. Here, you're not seeing >$150K in IT unless you're a Director, or more likely a VP level employee.
Conversely, I drive by multiple homes that are being built or were recently built by physicians living in the area that are in the $2M range (our median home price in the area is just under $530K.
You can't say "the subset of one thing compared to the spectrum of another", because testers, tech writers (the "median IT professional") are not generally clearing high six digits even at FAANG (well, Netflix maybe, which is an outlier of an outlier).
I frankly don't. Here we have a discussion that is predominantly about software engineers, and as a subset of those, those who are employed at FAANGs...
... and they're variably being compared to physicians, specialists, "median healthcare professionals" (whatever that means).
I'm more breaking up that these comparisons... don't really make a lot of sense, the way many of the arguments here are going.
Seems like this discussion is predominantly about physicians. It is about "The darker side of being a doctor"
But this being a forum for software engineers, the discussion naturally has a lot of content about that group. And many average software engineers, with a bit of amusing hubris, apparently think they are roughly comparable to physicians.
(Actually, they may be in terms of raw intelligence. You don't have to be a genius to become a doctor. But on average they aren't in the same league in terms of work ethic and dedication.)
I'd say the highly paid specialties are more comparable to senior positions in FAANG.
Both made it through the first filter (admission to medical school/being hired at FAANG), then succeeded so well in the early stages (med school/junior employee) that they were chosen for those coveted positions.
That's because the process to have any chance in the competition to become a doctor is arduous and there's no consolation prize.
If you aspire to work at FAANG and fail you can be a software developer somewhere else. If you aspire to be a doctor and aren't admitted to med school, those years of education, volunteering, shadowing, test prep, etc. are fairly useless.
And even if you succeed, the hours are as bad as can be and the pay isn't much better than FAANG. It's not a logical choice unless you really want it, while software dev is (was?) a great career choice even if you don't care about software.
For all these reasons, a lot of people are filtered out long before applying to med school.
You’re making my argument for me. The percentage of people who are actively trying to get into FAANG but don’t is higher than the percentage of people actively trying to become doctors.
It’s more selective. So comparing FAANG salaries to the average doctor salary isn’t useful. To find something equally selective, you’d need to pick a higher paid subset of doctors.
Here's why that's a silly comparison: the percentage of people on the general election ballot for US President who don't make it is lower than the percentage of people who are actively trying to get into FAANG but don’t make it. Does that mean US President is less selective? Obviously not, because that's the last of many filters for President and there are essentially no filters before "I'll send a resume to FAANG".
In the same way, there are many filters before filling out a med school application.
Anyway I can't see any way anyone could believe the "median software developer" writing CRUD web pages or in-house apps is comparable to the average physician.
Maybe I'm missing something, but I can't imagine modern Japan being a threat to modern China. That seems like David (Japan) vs Goliath, but unlikely to have the mythical outcome.
It only requires that several different mathematicians knowingly found partial results but weren't aware of the results made by other mathematicians.
reply