Underground 2 Revived runs Need for Speed Underground 2 in Chrome and Firefox via WebAssembly, with thousands of players free roaming Bayview online again.
I do find it fascinating that people are actually able to get AI to do anything useful. I assume they’re using something fed more specific training data? Anytime my work told us to use AI for anything it fucking sucked. Caused more work/ time than actually figuring out how to code something yourself
I assume they’re using something fed more specific training data?
Not at all, any frontier model can do it. Most people use Claude Opus 5.5 to make these things. I’m assuming you were using a worse one at your work, because the newest ones are pretty impressive.
It’s possible(probable) that I just don’t understand how to use AI in an effective way to actually have it do what I want. At work we used chatgpt along with cursor. I mainly used cursor but it could also have been that we were using proprietary systems. It did however seem it would give me weird answers even when I gave it source code to work with.
Things have come along quickly and are speeding up rather than slowing down.
At least people aren’t parroting that tired old, ‘This exact thing must already exist in the training data’ argument which has been fallacious from the start but is wearing excessively thin since new models use almost exclusively synthetic training data to begin with.
From experience listening to a coworker who has used AI a lot. You really need to tell it everything it has to do, slowly. It will try repeatedly to do something you didn’t ask it to, and you will need to tell it repeatedly to undo it, and to not make up solutions to problems that aren’t there.
I think a lot of people make the mistake of telling it something like “I need to make a program that can do z, x and y”. The ai will then fill in whatever gaps you didn’t mention between z, x and y, without asking, just guessing.
I only have very limited experience, but from what I’ve personally figured out, another good practice is to look into the thinking process (even though models can and do “lie” in there, so it’s not really the actual thing that’s happening under the hood) and catch mistakes early before the AI has wasted a bunch of time and (possibly) money on some nonsensical detour. You then have to stop the process early, add a line or two to the prompt to explicitly stop it from choosing this detour and start again. (In my case - I almost exclusively use local LLMs - it’s not about money, at least not directly, but because there’s only a very limited context memory due to the amount of RAM I have, so I don’t want it to be filled up by stuff that’s not helpful.)
It’s like babysitting a toddler. The moment you look away, chances are it might try something you really don’t want to happen.
Sometimes, the LLM might misunderstand your prompt entirely, because you chose imprecise language, made a typo that changes the meaning, you asked something of it that is way outside of its ballpark/training data, because of missing context or because the AI just had another dumb moment.
The old adage that computers are just as smart or dumb as the user (but much faster at it) often applies to AI as well, at least in certain use cases - in others, AI can actually help less intelligent users, but at the same time drag down smart people and erode their skills over time. It’s weird what it’s doing to our brains, which is why I’m very wary of using it for anything other than experimentation and fun, and avoid using it for things that are actually important to me. Many people don’t have this choice though, with mandatory AI usage in many orgs still being on the rise.
I do find it fascinating that people are actually able to get AI to do anything useful. I assume they’re using something fed more specific training data? Anytime my work told us to use AI for anything it fucking sucked. Caused more work/ time than actually figuring out how to code something yourself
Not at all, any frontier model can do it. Most people use Claude Opus 5.5 to make these things. I’m assuming you were using a worse one at your work, because the newest ones are pretty impressive.
It’s possible(probable) that I just don’t understand how to use AI in an effective way to actually have it do what I want. At work we used chatgpt along with cursor. I mainly used cursor but it could also have been that we were using proprietary systems. It did however seem it would give me weird answers even when I gave it source code to work with.
Things have come along quickly and are speeding up rather than slowing down.
At least people aren’t parroting that tired old, ‘This exact thing must already exist in the training data’ argument which has been fallacious from the start but is wearing excessively thin since new models use almost exclusively synthetic training data to begin with.
From experience listening to a coworker who has used AI a lot. You really need to tell it everything it has to do, slowly. It will try repeatedly to do something you didn’t ask it to, and you will need to tell it repeatedly to undo it, and to not make up solutions to problems that aren’t there.
I think a lot of people make the mistake of telling it something like “I need to make a program that can do z, x and y”. The ai will then fill in whatever gaps you didn’t mention between z, x and y, without asking, just guessing.
I only have very limited experience, but from what I’ve personally figured out, another good practice is to look into the thinking process (even though models can and do “lie” in there, so it’s not really the actual thing that’s happening under the hood) and catch mistakes early before the AI has wasted a bunch of time and (possibly) money on some nonsensical detour. You then have to stop the process early, add a line or two to the prompt to explicitly stop it from choosing this detour and start again. (In my case - I almost exclusively use local LLMs - it’s not about money, at least not directly, but because there’s only a very limited context memory due to the amount of RAM I have, so I don’t want it to be filled up by stuff that’s not helpful.)
It’s like babysitting a toddler. The moment you look away, chances are it might try something you really don’t want to happen.
Sometimes, the LLM might misunderstand your prompt entirely, because you chose imprecise language, made a typo that changes the meaning, you asked something of it that is way outside of its ballpark/training data, because of missing context or because the AI just had another dumb moment.
The old adage that computers are just as smart or dumb as the user (but much faster at it) often applies to AI as well, at least in certain use cases - in others, AI can actually help less intelligent users, but at the same time drag down smart people and erode their skills over time. It’s weird what it’s doing to our brains, which is why I’m very wary of using it for anything other than experimentation and fun, and avoid using it for things that are actually important to me. Many people don’t have this choice though, with mandatory AI usage in many orgs still being on the rise.