Why in the hell is there so much social signaling? "I really enjoyed his work for <reasons and experience here>, but <you don't need to include literally any of this because it's taking a moral high horse and trying to promote ones ego/values>"
About all of the good news, once you read a little bit more, are all due to traditional ML and all are in medical imagery field. Then OpenAI tries to take credit and say "Oh look AI is doing that too", which is not true. Go ahead and read deeper on any of those news and you would quickly find LLMs haven't done much good.
They helped me make some damn good brownies and be a better parent in the last month. Maybe I should write a blog for all of the great things LLMs are doing for me.
Oh yeah, and one rewrote the 7-minute-workout app for me without the porn ads before and after the workout so I can enjoy working out with one of my kids.
What makes you think you couldn't have made brownies without LLMs. Go to google and just scroll 20cm and there it is, a recipe, the same one chatGPT gave you. I wont comment on rewriting an app, because LLMs can definitely do that.
Because, "Why are the edges burnt and the middle is too soft? How are these supposed to actually look? I used a clear 8"x8" pan, and I'm in Utah, which is at 4,600 ft elevation"
Oh, it's a higher elevation, I need to change the recipe and lower the temperature. Oh, after it looked at the picture, the top is supposed to be crackly and shiny. Now I know what to look for. It's okay if it's a little soft while still in the oven because it'll firm up after taking them out? Great!
Another one, "Uh oh, I don't have Dutch-processed baking power. Can I still use the normal stuff for this recipe?" Yeah, Google can answer that, but so can an LLM.
What makes you think you couldn't have made brownies without Google. Just go to your local library and find the first baking cookbook you can find. And there it is, a better recipe than Google without all the SEO blog spam.
To avoid my comment just being snarky, I agree that there's a difference between comparing Google to LLMs, and the library to Google... but still I hope you can acknowledge that LLMs can do a lot more than Google such as answering questions about recipe alterations or baking theory which a simple recipe website can't/won't.
fwiw modern recipe sites are awful - you have to scroll down literal minutes until you get to the recipe. LLMs give you the answer you want in seconds.
I’m certainly no LLM enthusiast but pretending they are useless won’t make the issues with them go away
I doubt this bonanza is gonna last... These chatbots, feeding from the very source that can't seem to surface quality stuff by the way, will likely degrade just like those searches have for the last 20 years. There will be ads, there will be manipulation and deception, there will be pointless preambles and they will spit out even more wrong instructions and unusable garbage, and on top of it all it won't take 20 years this time do degrade, it's rather likely that it will take less than 5 years.
Maybe open source models will hold these accountable, or maybe they will degrade too somehow. Or maybe the world will be going through a hard collapse for any of us to care.
The model weights for the leading open source offerings are already downloaded by thousands, if not millions, of times. There's no unsqueezing that tube of toothpaste.
So there's OpenSCAD, which is basically programming the geometry parametrically. But... I'd liken it to generating an SVG of a pelican on a bicycle at the current levels of LLMs.
I designed some parts for an enclosure over the weekend using claude opus to generate an OpenSCAD file, which I then exported to STL and sent to a 3D printer. I was able to get a visually-correct-enough STL off to the 3D printer in about 5 minutes.
I then waited about an hour for the print to finish, only to discover I wanted to make some adjustments. While I was able to iterate a couple times, I quickly realized that there were certain structures that were difficult to describe precisely enough without serious time spent on wording and deciding what to specify. It got harder as the complexity of the object grew, since one constraint affects another.
In the end, I switched to FreeCAD and did it by hand.
Will echo sibling. I have tried using Claude Sonnet for OpenSCAD to design a simple soap mold and it failed terribly in getting the rounded shape I wanted. (1) It's really difficult to explain 3d figures in text, and I doubt there is a lot of training material out there. (2) OpenSCAD is limited in what it can do. So the combination is pretty bad.
I needed some gears generated recently, and figured I could just get it done with Claude or Chatgpt in OpenSCAD in a few minutes... but oh man was I wrong. I was so wrong.
Wasted half an hour generating absolute nonsense if it even compiled and ended up going with one of those svg gear generators instead lmao.
I think you would be best served prioritizing other areas of study over geopolitical conflicts; one that teaches you to respect people, even if their beliefs differ from your own.
I've found a page on Facebook that regularly posts single white mothers with black babies on supposed dating profiles with very demanding requirements for men. The comments are loaded with people saying that they deserve their current situation, enforcing racial stereotypes, etc. It's not hard to see that these are AI generated, as there are maybe 5-8 posts a day like this, and the images are pretty clearly AI generated. Regardless, they get the engagement, and they sell the shirts. Easy way to automate a business, I guess, but at what cost?!
I don't have a link, but I have seen exactly what he's talking about, which probably means that it is an established business model and multiple actors are doing it.
A similar thing I have randomly come across multiple times on YouTube are videos consisting of a still AI image of a white person mistreating a black person (e.g. a white police officer screaming with rage at a black man eating in a diner) and an AI voiceover text telling a GPT-generated story hashtagged #heartwarming, e.g. "The white police officer was violent against the black man... What he didn't know was this was a highly decorated veteran!"
Some of these are clearly getting picked up by the algorithm and drawing hundreds of thousands of views. The factories behind these are probably halfway around the world but realized the race relations of a large economy can be exploited for profit or geopolitics.
Several of the Reddit "AmITheAsshole"-style subs have a significant number of posts which are either AI or sloppy creative writing.
Mass-produced outrage bait isn't new, and it's available in a thousand flavors. But AI has accelerated this process, at least for people who don't notice when they're getting played (or who don't want to notice).
reply