Got an email today from the Wikimedia Foundation because I donated like twice in 20 years and they've "invited me to beta test" their new "AI" Chrome extension if anyone's interested. I tried it...
Got an email today from the Wikimedia Foundation because I donated like twice in 20 years and they've "invited me to beta test" their new "AI" Chrome extension if anyone's interested.
I tried it on a few things that were actually not really great tests in hindsight (like the signature in the email) but when I went to this article and copied the line "Low-Earth orbit is becoming increasingly crowded with active and defunct satellites." and hit the extention, it came back with the following:
Low-Earth orbit is becoming increasingly crowded with active and defunct …
Show more
The quote directly supports the claim that Low-Earth orbit is becoming increasingly crowded, mentioning the congestion caused by both active launches and, implicitly, defunct satellites contributing to space debris.
With links and green background indicating wikipedia agrees. At first try it looks pretty cool.
Disclaimer - it's only at v.0.0.21 with 104 users, so... at your own risk, maybe don't expect too much, and YMMV... but it's also only 1.7MB.
Enjoy if you're into this kind of actual real possibly decent use of "AI".
That's an interesting and clever use of LLMs. I'm always disappointed to see many people treat ChatGPT and similar services as something like "truth engines", seemingly or even openly believing...
That's an interesting and clever use of LLMs. I'm always disappointed to see many people treat ChatGPT and similar services as something like "truth engines", seemingly or even openly believing that if the model says it, it must/should be true, when every model that I know of has no concept of truth, facts or basic logic.
But by using Wikipedia as a basis for a "truth evaluation", you could potentially sidestep that issue and treat those assessments almost as reliable as you see Wikipedia itself to be. Of course that depends on how accurate this project will be.
I'll see if I can find some time tomorrow to try it out and test its limits.
Got an email today from the Wikimedia Foundation because I donated like twice in 20 years and they've "invited me to beta test" their new "AI" Chrome extension if anyone's interested.
I tried it on a few things that were actually not really great tests in hindsight (like the signature in the email) but when I went to this article and copied the line "Low-Earth orbit is becoming increasingly crowded with active and defunct satellites." and hit the extention, it came back with the following:
With links and green background indicating wikipedia agrees. At first try it looks pretty cool.
Disclaimer - it's only at v.0.0.21 with 104 users, so... at your own risk, maybe don't expect too much, and YMMV... but it's also only 1.7MB.
Enjoy if you're into this kind of actual real possibly decent use of "AI".
That's an interesting and clever use of LLMs. I'm always disappointed to see many people treat ChatGPT and similar services as something like "truth engines", seemingly or even openly believing that if the model says it, it must/should be true, when every model that I know of has no concept of truth, facts or basic logic.
But by using Wikipedia as a basis for a "truth evaluation", you could potentially sidestep that issue and treat those assessments almost as reliable as you see Wikipedia itself to be. Of course that depends on how accurate this project will be.
I'll see if I can find some time tomorrow to try it out and test its limits.