I guess my point is that you are knowledgeable in this particular area, so you can look at it and make a determination as to whether or not a particular response is accurate. Essentially fact checking the AI.
What hit rate of accurate answers do you need to feel comfortable? How would you feel if you were dealing with something in an area you were not familiar with and could not easily fact check on your own?
And is GPTChat’s method of taking a huge corpus of unrelated info in and then making up stuff based on patterns in that to answer questions the best way to do this? There has been a lot of work done on using machine learning to summarize information that might be a better way to deal with that sort of question.
I don’t know. I just feel like GPTchat is not great for that kind of application because fact checking it is more work for most people that looking up the info themselves
2 Likes
Absolutely. I would never trust it to give me a right answer to anything that matters.
3 Likes
Not being able to discuss your backer’s politics is an interesting blindspot…
(Also a reminder that this is another Elon thing, supposedly to save us from AI that will take over the world, but that will probably really ruin the world in way that cannot yet be predicted involving undermining basic concepts of truth, facts and reality)
4 Likes
Although, read through for a very good GPTChat bible verse
1 Like
Not so much pair programming, as going pear-shaped.
7 Likes
this seems strange to me. i have the J7 model and the pictures it takes to ask me “is this a temporary obstacle?” (usually a sock under the bed or some such) are terrible. i can’t enlarge them to make out some items, especially in dark areas. they never have any upward views, so no “high shelf, tv or ceiling light” just blurry pictures of the cat, my foot, or random object like a cat toy. the lens angle is not wide enough to capture anything above the lens.
i for one, have already welcomed my robot over(under) lord and while the folks at iRobot have a detailed map of my home in their data (problematic in other ways), they do not have pictures of any part of me from the ankle up. certainly no bathroom pics.
1 Like
The article mentions that these were testing units equipped with special cameras. The team that allegedly leaked the photos was tasked with manually categorizing objects an photos and supposedly shared images among themselves.
4 Likes
cool. i missed that specific note in the article. i did understand that these were hi-res cameras which made me snort, because i struggle to make out images mine sends me due to the decidedly low-res nature of them.
thanks for the - er - focus!
3 Likes
Yeah, cause it’s all teacher’s fault that far too often, the k-12 public education system grinds students down and makes them see education as a god damn chore, not the wonderful experience of learning how to expand your mind that it can be… Like teachers from pre-K all the way through the the university level (with a few, ivy league exceptions) aren’t over worked, underpaid, and deeply underappreciated and in some cases, hated and threatened in America. But yeah… WE are the ones who should “try harder”… 
8 Likes
Dear Brother (who is an Ivy League professor at Penn) was bemoaning the ease with which undergrads could try to pass off AI generated text as a college-level paper. he has been playing with ChatGPT and was pretty impressed with some of the stuff he got. of course, it helps to be able to give top-level prompts based on decades of his own research and writing.
he taught writing and economic anthropology before becoming a museum administrator and full time research fellow. he no longer teaches undergrads, but does fill in for other profs at times.
AI will create extra work for teachers at all levels, i fear.
4 Likes
Oh yes, especially those of us working at the bottom of the ladder in academia…
5 Likes
Couldn’t they use an AI to grade the papers? 
5 Likes
