YouTube link is here. Like and subscribe.
This week, we discuss artificial intelligence, and also just regular poor intelligence. Also, Tony saw a lot of movies, and has a lot to say. Then we invent new familial names. Enjoy!
Podcast: Play in new window | Download
Hi MTH community,
I’m the slacker listener Tony mentioned who just takes takes takes and never contributes. Well, no longer (apparently). Perhaps after this post, you will prefer I go back to silent stalker status; if so, I completely understand.
Thoughts about the AI convvo/Blake Lemoine/LaMDA transcript —
First I would suggest interested listeners look into Large Language Models also known as Transformer models. These Natural Language Processing models have been insanely rigorously trained and use on the order of 530 billion parameters in order to execute actions like summarizing, predicting the next likely word in a sentence/question/dialogue. I work in this field every day, and let me just say, it can seem super freaky! However, rest assured the underlying mathematical models are probabilistic in nature and are just about putting words being based on TONS of training. If you want to read/listen to smart peeps about this, I spoke to Sergei Revzin at a conference this past spring and he is AWESOME to listen to (has some YouTube videos). For a deeper dive, I also listen to the Lex Fridman podcast. He taught at MIT and has some fascinating conversations about this stuff. Ok, back to the point…
What about Google firing Blaine? Sorry for the unsexy answer, but I predict Google took this action because he violated the NDA–they don’t want any information released on their models. To do so is a big deal.
Will Blaine suffer because of this? I have a feeling he will do JUST FINE in the long run.
Does this whole thing benefit media sources and Google? Yep. “He violated the NDA” doesn’t seem to make headlines or sell ads, yet look at all the folks talking about Google’s fancy AI.
The unfortunate thing is that anyone who chooses to anthropomorphize Large Language Models is not helping anyone, and doing so is out and out misleading. We would benefit from more careful use of terminology. Instead, these are at base probabilistic models that put together snippets to emulate conversation. That’s it. Now to be fair, these models can and do compose brand new stuff.. But still, it is a probabilistic assembly of snippets… A probalasnippetassembler …along the same lines as transmographier if you will… but definitely not even close to sentient.
As far as what will happen one day? All the things, I’m sure. But in the meantime, fear not and check it out!
As voters, we might want to request that if something is totally written by AI that we might require that to be made known, but that’s just a thought.
Super interesting! Thank you for coming out of lurk to give us this! I, knowing next to nothing about this field, felt this was probably the case. I just know that humans are meaning-finding machines and figured this was just a junction of really good programming and general pareidolia. If a pile of clothing can look like a witch, how easy would it be to believe the thing talking back to you is really sentient.
That… was awesome… thank you for sharing your insights!
I had heard Tony say “Niblets” before but in my defense… Tony makes up shit so there is no reason to assume anything he says is official.
Aunt and Uncle was actually the path that lead me back around to the niece and nephew thing, both seem like an oversite of language. Hopefully, we can come up with something better than “parent’s sibling.” Parsib? Still not necessarily right because that’s not always the exact relationship. I dunno, I’ll probably continue to think about it when my brain is idle.
Also, I was 100% sure Tony would continue with “Uncle” as he states often that he doesn’t like to do things… and staying with “Uncle” is the path of least resistance.
P.S. Week three, still enjoying my Bombas. They have been keeping my feet dry in this heat so that’s another pro. Still not sure they are worth full price but I am starting to think they are worth what I paid.
P.P.S.
I certainly had an A grandparent, and B grandparents. And we didn’t really use made-up words, it was just my maternal grandmother was Grandma Lou and my Paternal Grandparents were Grandma and Grandpa Little. My Mom’s dad died when she was young so I never really called him another other than “your dad” talking to Mom.
Modern-day, my Daughter just uses the “Grandma and Grandpa First” name convention for all her Grand and Great Grandparents. Also, she doesn’t seem to have an A and B grandparent. She sees my mom more days a week than not, but for shorter amounts of time. She sees my wife’s mom more infrequently but for longer stints and just them.
This probably has something to do with my wife, me, and our daughter all being only children.
Nib*ling*! Not niblet! That’s a piece of corn, I think.
Don’t try and make a name out of existing phrases. From now on, the sibling of a parent will be referred to as a “crunket,” regardless of sex or gender.
I had an Aunt George – although she was a woman. Her real name was Mary, but her brother had a pet rat during WWII (maybe while he was a POW?) that he said he named George after her, so a bunch of us called her (Aunt) George. She passed in 2020 but her memorial service as a couple weeks ago and I was so pleased I was able to go.
I’m sure my siblings and I have interesting words we say/use but what I find more odd about us is that we have some hand motions for things. Particularly for if someone is being difficult and the 80s group the Eurythmics. If you would like to see these motions, invite me to an IHOP breakfast and I’ll share.
I think I niblet is a piece of corn. I also think Tony is onto something with the no more than 3-4 descriptors before it gets a name of its own.
Prinsess Potayto, thanks for joining us. I hope you’ll be back!
Okay, having a hand motion specifically to signify the Eurythmics is really tempting….. but I think we gotta wait until William and I have a more regular IHOP schedule. Our off-mic time is too precious, currently. But soon…