I asked Google Bard what the Dinky programming language is? It correctly stated I created it then it went on that it was used in the LucasArts adventure games (clearly confusing it with SCUMM) and then said Grim Fandago was written in Dinky. It wasn't even written in SCUMM. Someone that didn't know the correct answer would be horribly mislead. #AISucks
@grumpygamer When comparing the success of Midjourney with the failures of ChatGPT it became clear to me that AI is great with fiction and poor with truth. The last thing we need right now is another source of false facts.
@grumpygamer I wish it would just say, "I'm sorry <insert name here>, I'm afraid I can't do that" instead of lying until its figurative pants catch fire.
At least then I'd just have to be worried about being locked outside of the space ship instead of whether it's lying to me.
@grumpygamer Just curious: what are key reasons to create Dinky in the first place or in other words, what features did you miss in other languages that you used? If you answered that in a blog post, I’ll be happy to read that instead ☺️
@grumpygamer This seems like a pretty good example of how LLM's are driven by statistics, not facts. If you start assembly the chain "X was a programming language invented by Ron Gilbert", a statistically common follow up would be falsely claiming that it was at Lucas Arts.
@grumpygamer@gbraad that sounds like a fabulous idea. Get the angry Twitter people on an AI run version (twAItter perhaps) and let them argue with an AI from the opposing political view.
Which just goes to prove that AI is a VERY dangerous tool in the wrong hands. How can anyone vote, when they can't even tell what is true or not in the "news" they read?
Not as if it has been easy with Fox and Co. up to now.. but it will become quite impossible if AI gets into the wrong hands... which it probably has already. 😢
@MaryPot absolutely. I use this as an example to show that (at least) the current generation of AI tools does not serve any purpose (and might indeed be harmful) because if you know the answer, you don't need them. And if you don't know the answer, you can't trust them.
There are very few places where I would comfortably say "yeah sure, go ahead and use it" at the moment.
I've seen a few people using ChatGPT (I think it was) to produce code.
I also know of delivery services that use AI to calculate the best unit to deliver, in this case, food to a certain address, taking traffic and speed of unit into account.
I can see that being useful. But letting it loose on things like history and (OMG!) psychology (yeah - it is on offer!) - that's just abuse and not worth the paper it is printed on.
AI has to have really clear paramters, and, even then, one of them managed to say 1+1 does not = 2 and that 2+2=3 !!!
It is not a good tool and should be withdrawn from public use until better parameters are set and laws are made to clearly mark AI generated everything.
@MaryPot producing code works ok for a few years now. The crucial difference is:
it is (currently, usually) used by a person who can understand the solution and tailor it to the needs
more often than not, that solution is somewhat verifyable
not saying generating code should get a free pass, but with that kind of scrutiny, it's ok. Once companies go "eh, let the cheap newbies do all coding and we'll fire all professionals" then this might become dangerous, though.
Exactly. AIs are so very fallible that they can not be used to replace people, only to do the drudge tasks and then the human expertise has to step in and verify/tailor.
Not a bad idea to let greedy businesses do what you describe, though. The nasties will crash and burn pretty fast! 😉
@claudius@grumpygamer I was watching one program on YT where Polish sport journalist is talking with different people. And that episode was about ChatGPT. And the journalist went with: "ha, chatGPT is great, I just asked it about last paragraph or <here title of classic Polish book> and it printed it for me, it knows everything!"
And the expert just responded: "you better check if it's true", what journalist mocked.
I played around a bit with Bing's chatbot. Sometimes it would admit to not knowing the answer. Other times, it would just make things up.
Like in this example, where it added a monkey wrench to The Secret of Monkey Island, and a leaky faucet to use it on. (When I asked why I needed to fix the leaky faucet, it said it wasn't sure.)
@et_andersson@grumpygamer while I'm very much a critic of this technology, I also absolutely find it amazing that we taught a piece of purified stone to think and then told it about Monkey Island.
Somehow, I hope the machine enjoyed Monkey Island as much as I did :-D
Add comment