that’s exactly where Inaturalist (and you) come into the process, to validate and correct what chatgpt produces.
do you expect this game to make money? if so, then I suggest not relying on unpaid labor to fix what chatgpt can’t get right in the first place.
I promise you, it can be so much more fulfilling to take the time to learn all these taxa yourself. you clearly have the passion, and that’s great! I also understand wanting to take advantage of modern tools to try to get something produced ASAP. however, I really think you would be better off taking the time to educate yourself, with real-world experience and resources authored by real-life human beings, giving yourself a solid foundation in biodiversity knowledge. once you have that confidence, once you can correct the hallucinations an AI feeds you, I think you will be much better equipped to work on such a project!
A friendly reminder of the importance to listen to your target audience. You have a lot of people telling you that using ChatGPT is not the way to go, and these are the people you will want to be interested in your game (those with an interest in bugs). It’s a good idea to listen thoughtfully to your target audience. You have the possibility to use thousands of iNaturalist images with the appropriate licences and which are correctly identified – at this early stage in development, do consider that as an alternate path.
Yea, but not for something that’s supposed to be fun. Latin names are impossible to remember or pronounce - definitely not in the “fun” sphere needed for a game.
I think common names have their place, but I don’t think primarily using Latin names goes against “fun”. especially considering a potential target audience for a game like this - bug nerds! - Latin names might go over a lot better than you’d think.
Well, I can think of two conflicting explanations.
They are not as cheap as AI
They are not as good as AI
The first is a valid complaint. There is always a high proportion of the potential client base who are so cost-conscious that they will always go with whatever is cheapest.
But here is my experience on the Freelancer platform. There is a group there (which I have since left) for people to come for advice on becoming more successful on the platform. Many are given good, detailed advice, don’t follow it, then come back a few weeks leter and ask again. Anyway, apparently – this surprised me – there are some who try to market themselves as translators, and put on their profile that they “can translate between any two languages.” Yeah, well, so can anyone who uses Google Translate. Nobody is going to pay you to use Google Translate for them; if someone is going to pay a translator, it is because they expect that translator to do a better job than Google Translate.
Look at the comments earlier in the thread:
And skilled, talented creatives can’t compete with this? Yes, there is always a large portion of the potential client base who just go for whatever is cheapest; but, as the adage goes, “You get what you pay for.” A creative who can do a better job than an AI should be able to command a higher price than an AI and find their niche among clints who care more about quality than about whatever is cheapest.
I’ve found that I have to apply the same critical facilities to AI and to
human authorities. It’s certainly a problem that chatgtp always has
supreme confidence in its own accuracy.
The world is changing. Just as skilled photographers and graphic artists traded in their
darkrooms, inks and pastels for photoshop, they’re going to have to adopt AI image
generators as tools of their trade. I’m sure talented people in the field can get much
better results than I can, and worth paying for their time. But if I’m paying I expect
the talent to use the best tools for the task.
And if I’m working alone, having generative AI at my disposal
makes 100x better art that I could ever do. Just a brief example,
a friend suggested using cartoon images instead of realistic images
to denote categories. With AI I took a set of 40 realistic images and
transformed them in to a set of 40 cartoon images in an hour. Instead
of it being impossibly beyond my ability.
Chat GPT (and other A.I. models that rely on theft), while they were never great at that, are literally getting WORSE at that. The reason is because people use ChatGTP in this way to make articles, books, or, I don’t know, games, and it makes stuff up. After the media is published ChatGPT steals it and takes it’s own nonsense as fact while adding NEW nonsense.
It’s a vicious cycle that is killing A.I…
..Which is exactly why this is not in the foreseeable future for both art and scientific data.
I’m not big on the “AI is bad because it takes jobs” because I don’t think we should do something just because we want to pay someone to do it. However, if you’re making a game based in science, you need something that can produce images and data that is true and accurate.
For example, this mayfly doesn’t even have connected wing veins, so how could anyone tout that it’s any real species at all? It’s a combination of several species and families. This failure is caused by trashy AI images of other mayflies as well as mislabeled stock footage flooding the internet, literally poisoning ChatGPT and other models that rely on theft for their data.
Personally, I’d only trust ChatGPT with coding, and even that needs to be human verified. To get accurate images, you’d need to train your own AI model with non-stolen, verified data (perhaps the CC0 images on iNat?) And to get your scientific facts, unfortunately, there’s no current substitute besides searching for and reading it for yourself.
So true! - I would have loved to play (or follow on here how it grows) a game about insects collecting cards and learn a little more about them.
But a game with chat gpt, which is destroying more and more of our world, creativity knowledge about real/not real and provides false
information, just makes me sad…
