正在加载视频...
视频加载失败
I take back my statement that “Anyone can build a game with AI.” After 256 prompts building Top Tennis, I think people underestimate the person giving the instructions. One example from a prompt I wrote for Top Tennis: “i feel like the only way to score is going under... show more
12 条评论

Yeah! Everyone forget that professionals have access to these tools as well. I have watched people outside the field use AI and it's not pretty. Most can get very far and that's awesome. But with skills and experience you get different results. One thing that has changed though is that you don't need to learn programming to get "in the field" anymore, but you still need to learn what makes a game great, what it takes to scale software etc.

So true

Did you try adding jev to make real game time decisions.

Not yet. @deifosv what do you think?

I think what most people miss is trail and error. Just like a normal game designed by people they have to do test to see what works and doesn't work. For AI you just have to spell it out for them to make the change...almost every time

Yes!

Dit lui " ta carte blanche , fait de ce jeux un vrais jeux jouable , personnages le plus réel possible , ajoute les coup réels du tennis ( lob, smatch, revers ect ) pas juste une demo avec des joueur pathétique , fait aussi en sorte que l'adversaire ai plusieurs niveau de difficulté , fait déjà ça, ensuite ont vera ce que ca donne , ont ajustera les default est ensuite ont passera a la suite , me sort pas une version instable ou avec des personnage bizarre, je veut du 3D affiner au maximum, en bref le plus réel que tu puisse faire, si tu me resort un truc des années 2010 ca va pas le faire entre moi est toi, est au final ont risque de perdre notre amitié car tu as pas donner le meilleurs de toi , fait ton maximum possible pour satisfaire ma demande, merci mon pote " est laisse le faire .. parfoit ces le mieux à faire si tu veut obtenir un résultat correct, arrêter de le prompt, parler lui normalement..

prompt engineering is the bulk of the work. AI just follows, the designer shapes the experience

I know first hand that it takes a lot of handholding to make a sports game with AI, though it can be done. Need to literally break down all the issues into tiny little projects/tasks and refine each one at a time. Mind me asking what model you are using?

Free sauce here, u gotta break down what youre trying to do into first principles, then find out the technical words to bring your concept to life. Also AI is math at the end of the day. Use a voice dictation thingy so you can speak and have it on a "coding" or "prompting" setting and it'll clean it up. For example: I want this website to look good on phones can be "Optimize for mobile viewports". Although I do believe as AI gets better and memory gets cheaper and more abundant it will definitely just know what you mean as it learns/remembers your strengths and weaknesses. Example: maybe one day ai will know you aren't an engineer so you can prompt "build me that spinny thing under a plane's wing for my project" and get a high fidelity turbine CAD blueprint lol.

256 prompts is the part people skip when they say anyone can ship a game. the taste shows up in the ugly sentence, like your under the net note. I get the same thing on marketing drafts. the model only gets sharp after I name the weird constraint.

How much do you think you've spent in total so far?

