Hacker News Reader

Comment by babelfish | Hacker News Reader

doing math is not "misaligned"

Are they fully “doing math” in an aligned way though? They are finding mathematical proofs, which is important, but contextualizing results in the prior literature and clearly communicating the approach and implications is just as much part of mathematical research.

You might argue that these aspects of math are less important in the new AI accelerated math world, because agents will inevitably be smarter than humans, but I think clear framing and communication is even more important than before because with this technology we can choose to augment our intelligence instead of defer it

> Are they fully “doing math” in an aligned way though?

No matter what you think "aligned" means here, publicly releasing advancements in a scientific field shouldn't be gatekept or be perceived as misaligned in any possible way.

If anything, the true misalignment comes from people trying to prevent these advancements from happening or being disclosed, or putting research behind BS paywalls.

I don‘t think this is gatekeeping. OpenAI is filthy rich, they can afford to pay mathematicians to go over these papers and produce results consistent with the scientific method. The fact they don‘t is evidence of their misalignment. Their motives are obviously ulterior and have nothing to do with advancing knowledge. If they were they would adhere to the scientific method, hire experts, and produce reproducible results.

> Putting research behind BS paywalls.

Yes, that is gatekeeping. But there are more then one way to be misaligned. And the practice of publishers is not being discussed here. No need for whataboutism.

Wait, what has OpenAI done that is not “consistent with the scientific method”?

https://www.sciencebuddies.org/science-fair-projects/science...

Do other researchers or research institutions “pay mathematicians (other than those in their employ) to go over these papers” or do they simply publish their work for review as part of producing “reproducible results”?

AI may be taking over Mathematicians’ jobs, along with everyone else’s, and it’s OK to hate that, but the scientific method says nothing about that, or hiring experts, or ulterior motives, or “being filthy rich.”

OpenAI has not released the model for replication (not even to select few), they have not even shared the prompts. All we have is their results and their compute times (which may even be lies). This is the opposite of the scientific method. The scientific method asks for falsifiability and replication. If a hypothesis is not falsifiable, it is not scientific, if results cannot be replicated, they are not scientific. OpenAI fails on the second account here.

Not being scientific is excusable in some context. A high-school prodigy might stumble upon an unexpected result that they had a hunch about, but fail to make those results replicatable. Similarly a YouTuber with under a 1000 subscribers might propose an unfalsifiable hypothesis as entertainment. Being scientific takes skill and effort, and it is not free, so we can’t expect anybody to follow the standards. However OpenAI claims to be a trillion dollar company, and they do not have the same excuses as a small YouTuber or a high school student. OpenAI absolutely can afford to follow the standards, if they claim to be furthering the human understanding, then they should be scientific about it.

> advancements

They're not advancements yet. At this point they're just LLM hallucinations that may or may not be advancements.

write down "LLM's will cure cancer" on sticky note, put it on your desk and recite that every day, a hundred times and in the bathroom if you have to

you're really crashing out about this huh

I know I am, though maybe not as visibly so as your parent. Us anti-AI Luddites are arguing against a massive propaganda machine with trillions of dollars on the line. It is not easy, and it is mentally draining. Especially since AI-companies are behaving in an obviously anti-human, anti-science, and anti-industry ways for their own short term profits.

And worse yet, we have seen this before (albeit on a lesser scale). GMO was supposed to solve global hunger, Carbon capture was supposed to solve climate change, etc. etc. These were obvious lies and propaganda back then as much as the lies where AI are supposed to help humanity today. It gets tiring, and at some point we simply crash out.

unfortunately technology will never fix humanity's tendency to ruin itself (see: war, climate change, etc; all of which we know the solutions for ie 'stop killing people' 'stop using oil' etc) but that is not a fault of the technology itself (gmo, carbon capture, crispr, llms, etc)

This is simply not true. Plenty of technology exists which has helped humanity unambiguously. But there is a difference between good and helpful technology and a grift. GMO could have been helpful but it was mostly just a grift, GMO solving world hunger was always an obvious marketing stunt, and it worked, even though many of us (particularly on the political left) saw it as such and called it out as such. Ditto Carbon capture.

AI is currently the grift of choice.

Also I fundamentally disagree that there is any tendency for humanity to ruin it self. There are perverse intensives, which rewards undesirable behavior, there are unfair, undemocratic, or otherwise hostile powerstructures which allows one class of peoples to exploit another class of peoples. And some technology sometimes plays a pivotal role in maintaining or furthering these hostile power structures/perverse intensives.

Reply on news.ycombinator.com