Comment by mosura
11 hours ago
So let’s say OpenAI cure cancer and put every cancer researcher out of work depriving them of intellectual satisfaction, this would also be a problem? It would certainly impact the field.
The fact is these fields are supported by society because of the benefits to everyone else. Once the same results can be achieved in a cheaper and faster way that is what will be done. We should mourn this in the same way we do buggy whip manufacturers. Again people still ride horses.
> So let’s say OpenAI cure cancer and put every cancer researcher out of work depriving them of intellectual satisfaction, this would also be a problem? It would certainly impact the field.
Would you expect Fields medalists to cure cancer if you moved them from the math department to a medical research lab? This is precisely the fallacy that the frontier labs are counting on to inflate their valuation as their IPO approaches. They want to use headline-grabbing problems in pure maths to make their models look "smart" in the public eye. But what does "smartness" in mathematics really mean in terms of economic value? It is not at all obvious whether success in abstract mathematics should translate to successes and, more importantly, profitability, in more grounded endeavors.
Look at OpenAI's job postings (https://openai.com/careers/search/). Those roles involve far more pedestrian yet profitable duties than research mathematics. So why isn't OpenAI automating them with their vaunted models? Success in one field, no matter how "difficult", does not predict results in another field.
> The fact is these fields are supported by society because of the benefits to everyone else. Once the same results can be achieved in a cheaper and faster way that is what will be done.
Maybe it's worth double checking that you know how these fields benefit everyone else? Proving the blowup of the Navier Stokes equations in 3D isn't going to make your gas cheaper or make harvesting food easier or make drones easier to protect against. Maybe consider the deeper effects at work?
Math people absolutely are important to things like the SpaceX landing control systems, hypersonic gliders, 5G networks and so on.
If you can make breakthroughs on such areas as fluid dynamics, control theory or information theory with AI then that absolutely is a big deal with real technological implications.
A long complex proof to something we have been able to simulate in fidelity forever doesn't do any of that though.
Why don't we first assume OpenAI can create a cold fusion reactor and then give them a trillion dollars to get it done?