US gov and OpenAI mislabel map of Africa at global conference
theguardian.comThe way a lot of people use AI is kind of proof the only thing they ever did or cared about was appearances and hype. There is not one iota of interest in reality. Every waking moment is a walking con and a sales pitch.
But one can say that previous technologies only magnify and exacerbate this lack of caring. The internet for example, makes most of our interactions casual and sometimes anonymous, and we get trained and even grow up on that model. And on the internet, what gets attention is what matters, not the consequences after because things move so fast.
AI is just the latest step in that. It is the apex of a phenomenon that dehumanizes people. Yes, there were people who just cared about appearances before the internet too, but there were greater consequences of that. Now the consequences have been reduced and it causes this behaviour to spread.
At no point has anyone in power ever operated otherwise. Frankly I'm surprised that you're surprised.
Remember when Musk in his capacity as DOGE boss cut the USAID funding for condoms in Gaza, because they didn't know Gaza is a highly HIV infected region in Mozambique? If they weren't endangering lives, it would be funny:
https://www.yahoo.com/news/truth-condoms-gaza-doge-canceled-...
The mislabeling is egregious but also the whole map is way out of sorts! UK, Greenland and Australia locations for starters my goodness...
can't blame OpenAI for hallucinating a map, that's what it's good for - can blame officials for not proofreading it though
I know ive been wanting chatGPT to create snazzy map routes to share with friends for our road trips and or vacations the past many years but it fails miserably! Which would show distance / drive time between each spot, interesting sites at each spot and along the way.
You can, but the trick is to ask it to write a python script to generate the "visualization". The added set of constraints imposed by library structure, internal data/shapefiles, etc keeps it from hallucinating
Ah thanks, yet tho not easily done. I think letting users easily generate and share AI-created snazzy map routes on social media would be a great marketing play for GPT. I think a lot of people need to be taught how to use it.
seems odd/inefficient that it would hallucinate a bad map instead of retrieving an actual map!
You actually can blame OpenAI for it. Because if a tool exists and people use it, that tool is an effective means of normalizing behaviour. We change our perception and behaviour through the use of tools, at least partially. So if someone invents a tool to do something poorly, and it becomes popular, then doing things poorly becomes more normalized.
It can work in a positive direction, too. Have you ever used Vim or Emacs? Haven't you noticed how dedicating some time to learning a complex system like this also makes you proud of mastering a precise set of tasks? It almost becomes religious. The tool promotes care.
We all have a societal responsibility to consider our inventions and what behaviour they'll encourage when people use them. OpenAI should also be held responsible, like all AI companies. The lack of care they encourage is enough for me to think that they should be all taken off-line and rendered bankrupt.
If I have to proofread a map you gave me, then your tool is a shit. And I can blame it.
yes but it seems likely that the user asked it to generate (hallucinate) a map, not annotate an actual map
one would be well-advised to proofread everything from stochastic generator tools
I agree but note that frontier companies are NOT selling stochastic generators you are supposed to check. They are selling supposedly trustworthy tool you should use for automation, for learning new stuff and use to replace your experts with.
yes they are while simultaneously warning that their models may be inaccurate - trying to have it both ways!
Landlocked Cote d'Ivoire is also interesting. Well, I guess it's a coast of some river or something facepalm
It's interesting that not even the Trump administration is using Grok for this kind of task.
Maximum effort going nowhere for 70 years.
No surprise, you hire idiots you get things like this making the US look totally incompetent.
We have seen things like this happen many times over the past 2 years.
It's difficult to make the government effective when a bit under half the country/voters thinks the government is evil and should be as ineffective as possible to reduce its ability to ruin their lives.
And this is only the stuff that's totally obvious. Imagine how much other stuff the kakistocracy is doing that's just as bad but not so easily visible?
What is the (unlabelled) left line even doing.
Is there anyone working for the Trump administration who is even borderline competent? Like, it is getting farcical.
I thought Rubio would be somewhat competent. I disagree with what he stands for, but he seemed smart.
But this proves to me he is as dumb as the rest of the administration.
the government's arsenal of slop cannons are just getting warmed up
Can we blame Rubio for this?
No, it was a plant by the radical-left communist antifa Soros-funded... ah, what's the rest of the formula? It keeps changing.
This administration is staffed by drunks and imbeciles. Their naked corruption is offensive but their stupidity possibly more so.
that's not mislabeling. that's pure AI slop
Could this not be intentional? Who wants to waste time getting a map of "shithole" countries right? /s