I think this move by OpenAI is crazy. At best, if all unconfirmed accusations are unfounded, they still heard a rumour that someone had solved a huge million dollar problem and was about to make a name for themselves. Then, they decided this was a good opportunity to pour millions of dollars into trying to snag the glory while the researchers were busy cleaning up their notes and polishing the announcement.
Would this be unethical if it was a human who heard rumors about a solution then attacked the problem, solved it and published first? Often knowing of the mere existence of a solution carries a lot of information--you would know the problem is accessible, you would expect clues in recent progress (the two Spanish researchers in this case), you would probably have a sense if the solution is a counterexample or positive proof, and so on. I think there are similar examples where we think of them as maybe unsporting but not quite unethical. Does it change if it's openAI and not a human?
According to Buckmaster, the prompt used on the AIs was based on his approach and solution that was unpublished. So they were starting from 90% of the way there.
The problem with your counter-hypothetical is that not only is it unrealistic, it's utterly impossible. No human would be able to do in such a short timeframe what the LLM did. Part of what makes the OpenAI move so egregious is how bullying it was. It was the big guy coming along with their nearly infinite resources and squashing the little guy who's devoted a good chunk of his career to the problem.
Actually my hypothetical is completely realistic as I've been involved in such scenarios. It's unrealistic maybe for a millennium problem to come in on a rumor and still front-run but not at all for the many other problems we work on and which manifest our ethical code. If you're saying ethical rules change depending on the prize be clear about it, because I can see arguments that they change to favor either side.
In my work as a graphics programmer I often find that I look at a problem and will immediately see how to solve it, more or less. But the devil is in the details and often nothing works unless you get every detail right. So you spend a lot of time coming up with complex solutions, then boiling them down to simpler versions. In the end you often end up with a fix which is short, simple, and seems obvious. But it gets a lot of subtle details just right and avoids countless potential issues you wouldn't know if you hadn't failed a lot getting there.
And that is actually how you learn and master the craft.
Now, imagine you describe how you sort of solve it to a machine and it spits out the simple, correct implementation and you nod approvingly, never knowing all the ways it could have gone wrong. If this is how mathematics - or programming - is done from now on, no one will actually master their craft. I definitely see why this would worry someone whose career is built on mastery of the craft and a legacy meant to teach the next generation.
The (new) lack of need to redo solutions to already solved things opens up new craft we've not yet conceived.
We shouldn't reimplement solutions and call it self-education; humans have always stood on the shoulders of giants. These tools contain those giants (copyright notwithstanding) and let you receive solutions to solved things; if your mindset is in order you can leverage that to ignore the solved bits and do even more ambitious stuff.
Everything is an apple, it seems. You also have earth apples (pomme de terre in French, aardappel in Dutch) for potatoes, grenade apples (pomegranate in English, granatäpple in Swedish, ...), pineapple, ...
It took me some time to remap the muscle memory but it felt so worth it. The Steam version is not just a new AI, it introduced a lot of gameplay changes which really improved the underlying systems a lot.
The version available for free on their website is the same exact program as the one you purchase on Steam. It simply includes a different set of graphics for the UI. These graphics are designed to look as if they are made up of text characters, but in reality they are still just sprites. The sprites that come with the version you buy from Steam are designed to look like a normal but stylized gui.
Is that insane? Every game[0][1] I ever helped ship has had fully localized UI. In game development that seems to be the standard way of doing things. And it requires a lot of testing and sometimes comes with workflows to automatically test that all strings fit their intended box.
[0] Before I worked on game development I actually worked in virtual surgery simulation, and there we also had a localized UI with a very large number of supported languages and tooling to detect issues.
[1] Need For Speed (2015), NFS: Payback, NFS: Heat, Lost in Random, Wavetale
I had a friend who was really into EVE online and showed me about 20 years ago. It felt too much like waiting and spreadsheets for me.
I got into Dwarf Fortress around 2015 and I was hooked. These days it's something I often play together with my son. I love how dynamic and deep it feels. It was such a privilege to get to buy a copy when it finally came to Steam.
One of my proudest professional moments was when me and three others managed to reduce memory load of the game Wavetale from 20+GiB to under 3GiB so we could port it to Nintendo Switch.
The 100 TiB number almost gives me vertigo. Though in this context it was "just" 50%
It's an indie open world game originally released for Stadia and in its original form simply loaded the whole world into memory at boot.
We had to implement a streaming system and figure out a good way of chunking the world. This was a challenge because everything was on water and you could see nearby islands quite far away. An intern called Tommi did a great job identifying a good strategy and writing the system.
We also had to reduce the density and model complexity of a lot of environmental details such as rocks, vegetation, and stuff like pots and clotheslines. This was done largely by sorting things by memory size and frequency of use and identifying outliers.
One of the biggest issues was actually really silly: the journal fetched Portrait images and names of characters by referencing the actual NPC and having them embedded there. Meaning the journal, which was always loaded, would pull every NPC involved in a quest into memory including their behaviors, textures, and models.
On my end I also found a lot of silly details wasting hundreds of megabytes. Special render passes using huge textures and render targets, poor structuring of the render pipeline caused memory increases, several key shaders referenced huge textures which weren't necessary, ...
My blog posts about the project[0] mainly focus on rendering performance because that's where I spent more time and it contains more interesting content for discussion. But reducing memory was an ongoing concern with countless little improvements over the 1.5 year porting process
I think using AI often feels faster than it is because you put less thought and effort into the problem yourself.
Another possibility is that the people who experience these 100x productivity increases are honest, correct, and simply had abysmal productivity which has now been increased to near-average junior levels thanks to AI.
That still sounds highly unethical.
reply