To be fair my expectations is that those apps have done the prompt engineering, and schema, and tools (to query nutrition database), etc... and although they're not 100% consistent, the margin of errors should be narrow to the point that barely matter, and they should do a bit better than a random ChatGPT chat session.
the problem isn't one that can be solved with prompts. If I gave a panel of food and nutrition experts (human or machine) a bunch of pictures of food, they still wouldn't be able to tell if, e.g. a slice of cake was made with whole milk or skim.
The "pic of packaged food --> LLM --> nutrition DB call" pipeline is workable, but many users of these apps are using them for fresh prepared foods, which is just an unworkable problem without either an understanding of the preparation process or a bomb calorimeter.
Probably because the goal is to have more users, not necessarily profit per user. Netflix once had that "problem" and every lockdown increased the stock price.
I've tried and couldn't make it sound like Angine de Poitrine, it completely ignores the microtones. Sounds more like Polyphia. It does look like AdP is the answer to AI.... or we haven't trained the models with sufficient microtones, likely due to western music influence.
Now I'm only replying because I'll take any opportunity to prop King Gizzard and the Lizard Wizard, which is the third item of the original three dot points.
KG = KG and LW = KGatLW = King Gizzard and the Lizard Wizard.
I don't like all of KGatLW's music but, as someone who is also a big fan of Frank Zappa's extensive corpus of works, I love their versatility and their willingness to be versatile.
Not the one who posted it, but welcome to a decreasingly exclusive club!
When I first watched the AdP KEXP performance, it felt like my musical knowledge up to that point was just preparing me for AdP. I've been through micro-tonal (King Gizzard and the Lizard Wizard, specially their Flying Microtonal Banana album), polyrhythms and time changes (Tool, The Mars Volta, Metallica's St. Anger album, various other bits and pieces), looping (Party Dozen -ish-, Adam Page[0][1][2], various Math Rock bands, primarily Battles), and other general "out-there-ness" (Arthur Brown[3], Frank Zappa, Mr. Bungle[4]).
The Adam Page stuff is all, essentially, stream of consciousness, unique to the individual performance. He had a monthly residency at a pub in the city a long while back, and I reckon I went to see him ten times over the course of a year and each performance was unique. I think they're all recorded on Bandcamp somewhere, hey, yes, here: https://adampage.bandcamp.com/music. I even named one of his songs: "Preventing a future disaster" (from September 2014, although it's misspelled). I find part of the enjoyment of looping music is in the performance itself, the combination of conducting and choreography.
For bands less "out there" than AdeP it's also disappointing.
I'm struggling to get a JJJ style "Like a Version" cover of Sleaford Mods doing King Stingray or vice versa .. no joy in the sense of it's a struggle to get anything that sounds anything like either band.
The text it generates midway sounds promising .. and then it plays audio that has zero of the unique elements of either.
It does explicitly state that it can't (or won't) imitate the style of a band(?).
I imagine that AI can be trained on Angine de Poitrine, but it would never really be able to create something like that no matter how many text prompts it was given.
They have to be generic because it's a generic tool. If they write "this tool can arrange student field trips", people might ignore thinking it has a narrow purpose.
Yes, work is being trivialized, but the symptom here isn't caused by that.
The issue is not that they are generic. You could still be generic with phraseology that actually acknowledged the contributions and ownership involved in the jobs being done. For example, you could write e.g. "monitor for outages", "manage projects", "arrange community events", "handle logistics", and so on.
But the problem is LLMs can't do those things. All they can do is "edit files" and "send messages".
Like any interview, randomly. Some of them will think it's amazing, clever person, chose the best tool for the job. Other will think it's weird, person is too clever, chose the worst tool for the job.
It totally depends for WHAT you're interviewing, but unlikely the company will want Z3-backed code, so most reactions would be the later.
> Can you design text data specifically to mess with LLM training?
Maybe text that costs a LOT of tokens. Very, very verbose. I think if there are rules and on the internet, LLMs can eventually figure it out, so you have to make it expensive.
Another way would be to go offline. Never write it down, only talk about it at least 50 meters away from your phone. Transmitted through memory and whisper.
reply