I run an ML team in fintech, and am currently hiring. If a resumè came across my desk with this "skill set" I'd laugh my ass off. My job and my team's jobs are extremely stressful because we ship models that impact people's finances. If we mess up our customers lose their goddamn minds.
Most of the ML candidates I see now are all "working with LLMs". Most of the ML engineers I know in the industry who are actually shipping valuable models, are not.
Cool, you made a chatbot that annoys your users.
Let me know when you've shipped a fraud model that requires four 9's, 100ms latency, with 50,000 calls an hour, 80% recall and 50% precision.
What does 50% precision mean in this case? I know 50% accuracy might mean P(fraud_predicted | fraud) = 50%, but I don't understand what you mean by precision?
I'd hope you're shipping _anti_-fraud models. Fraud models are abundant in fintech. On a side note, anyone who _really_ knows LLMs will be able to do your primitive fintech models with their eyes closed. You don't need a ton of skill to build a logistic regression model that runs at 14 qps. :-)
I've played music my entire life (picked up a guitar at 6 years old and just never put it down). I actually just released a new record last Friday (https://open.spotify.com/album/6JU0jmz537a6r2xrTvCcmn?si=eg4...). I joined a band when I was 15 (~2004), and we had some long tail success. We were able to tour, play huge shows (the Gorge in Washington, sell out the Showbox in downtown Seattle, an arena here or there). After high school I went to school for audio production, and even then I knew it was going to be tough to make a living. I ended up pivoting, studying math, now I'm in machine learning.
Music is the thing I love more than anything. I love writing it, releasing records, playing shows, and connecting with people on an emotional level. Never once have I considered it possible to have a fruitful career as a musician, despite seeing more success as a musician than most can ever dream of. Additionally, the industry (like many others) has changed dramatically over the past 25 years. In many ways, it has put much more power back into the hands of artists: you don't need a huge studio/record label/promotion to release a record. You can just release records, and promote them yourself. The flip side of that is there are SO many more people releasing music these days, which makes it really difficult to cut through the noise if your music is halfway decent.
Finally, recommendation algorithms have truly transformed the landscape of content creation, likely irreversibly. I get messages _daily_ from people who have "hacked" the TikTok algorithm, and can get my bands plays. There is an entire cottage industry of algorithm "hackers", some of them actually have results too.
One odd anecdote: I love Alex G. I've been listening to him for over a decade, and have flown out to see him play in places like New york/Austin TX. A few years ago he played in Seattle, and the entire demographic of the audience seem to've changed overnight. Way younger, more "mainstream" looking kids, filled the Showbox in Seattle. The strangest part was that no one seemed to know the words to his songs anymore. I did some digging, and he'd gone viral on TikTok. A few of his songs went absolutely bananas on there, and it completely transformed his fanbase. They knew the words to those songs, but not his entire set. Is this bad? I have no idea, but the trimming down of content into bite sized morsels _feels_ bad to me, and I believe it will dramatically alter this next generation's baseline attention span. Again, not a moral judgement, just a factual claim.
>The flip side of that is there are SO many more people releasing music these days, which makes it really difficult to cut through the noise if your music is halfway decent.
I think one thing important to consider here is that part of the experience of enjoying music is not necessarily how good the song is, but how much, and how many, other people are enjoying it. People often listen to (mediocre) music simply to have a shared emotional experience with others.
Imagine bonding over gruel, because everyone else is eating it and you can’t connect with them unless you are able to discuss the consistency and mouthfeel of the gruel.
"Clichés like this are beautiful, because they reflect us and we are beautiful. Take, for example, this chord progression. It only became taboo because it was too powerful -- that's why you won't forget it." --Porter Robinson
Pop music isn't gruel. A lot of it may be slop, but it's deeply appealing. Somebody somewhere solved for what "works", and a million copycats cloned it with minimal effort because it works.
So don't think gruel. It's more along the line of... McDonald's. Bad food, but it's appealing. And people do bond over it, or at least they used to before people stopped caring and fast food places became utter hellscapes. You still see kids bonding over McD's in Japan.
In order to be heard, truly heard, you have to be able to be understood. Music is 80% familiarity. Rarely can you just add your 20% uniqueness and be understandable. All music starts with 80% gruel as the base recipe.
This is a good point. I'd argue well more than 80% though.
Time signatures, instrumentation, arrangements, chord progressions, etc are the base gruel that forms the core of almost all (western/popular) music.
The "new" contributions of most artists are more like flavorings or spices with the occasional unexpected twist on the base. And, critically, this is necessary to find an audience.
Even bands that "change music" are just permuting on the basic gruel. And, usually, just popularizing the permutations that other bands have tried first but were too early/didn't break out of their local audience/etc.
Often these permutations only get popular because they bring with them a new and appealing (or under-represented) aesthetic. It's not even really the music, necessarily.
There are exceptions, there are some real musical innovators. They rarely get popular though, no matter how much respect they earn from their peers.
Relatable. Some of my best friends were made in the heat of struggle, not in a fancy establishment. When you're happy and comfortable, people are a dime a dozen. When you're down on situation, any human contact is a luxury, and the experience embeds itself in your mind.
Music is 80% familiarity 20% novelty. Western scales are 100% gruel when you consider the available audio spectrum/combination. And yet I bet a large portion of music you enjoy is made up of 'gruel' made to be 'gruel' simply to have that common connection you deride. And even if not, do you not have genres that you enjoy? Each genres just being their own brand of gruel with whatever familiar makeup defines it?
> They knew the words to those songs, but not his entire set.
This has always been true for recorded music. Originally people would buy mostly singles after hearing a song on the radio, then maybe listen to the B-side too.
Listening to complete albums was only popular for a short while before streaming brought single songs back to prominence as the main way people consume music.
Don't disagree. I'm merely commenting on the dramatic change in his audience, which IMO opinion was driven by TikTok virality. Going from a crowd of people who were singing along to people standing around waiting for the "TikTok hits" was really strange.
I had a similar experience when I went to see James Blake; the audience was bimodal in age and there was a younger crowd that only knew a few of his singles that had gotten real big (collabs w/ Travis Scott and Rosalia)
So maybe this is normal as we get older? I didn't know this had happened with Alex G but I'm happy to hear about his success -- to me that's the main thing that matters, however an artist finds their audience.
Not for alex g. He has had a cult following as the best songwriter in rock music for a decade plus. Up until he took off on tiktok everyone at his shows knew almost all his songs. I guess really the complaint here is just that he went from cult musician to a having more pop appeal.
Man the tiktokification of alex g absolutely blows. Same with mitski, unbearable live shows now. It is a bit difficult for me to be mad about it though because at the end of the day the complaint just boils down to being mad that these artists have become more popular, pop sets have always been like this. More popular = more money for them which cheers me up a bit
I don't know if it's good financially, but do you have a bandcamp? I like getting cds / mp3s there usually and it doesn't need a sign in to listen to the song.
We don't. I probably should make one of those, but as a solo act, the number of platforms I need to keep up with is ridiculous. Reddit/Spotify/Instagram keep my time occupied, it's brutal honestly.
As a solo act, I also agree. There's 24 hours in the day and to do the metadata correctly for the releases themselves _just to send to the distributor_ takes like 1-2 hours. Then I need to make content, when I just wanted to make music, then upload the same release to Bandcamp. It's untenable
I think "get it to write functional code most of the time" severely discounts the value of the code produced by ChatGPT. Knowing zero Swift (but being an SDE), I was able to build an audio plugin for Logic Pro X in a day, by leveraging ChatGPT. I'd previously tried this twice, and gave up because of the learning curve and my lack of free time. It's the most insane 0% to 80% tool I've ever seen.
I’m personally looking forward to when I can use a ChatGPT yo build my own ChatGPT and make tons of money out of it too !
It’s pretty interesting because soon absolutely No business will be safe. All tech, most apps, everything can just be stolen as fast as ChatGPT can write it. I’d go as far to say as ChatGPT itself is vulnerable to this ? Am I wrong ?
I’ve been working on a startup and I’m actually really evaluating if it’s worth the time now. I mean it could be stolen pretty fast. Not sure how to reconcile it all yet.
Make my an app which is functionally the same as Netflix, then write the deployment code so it runs on AWS. Ha.
Microsoft and Open AI will may benefit a lot from ChatGPT, but it might eat itself too.
You can already ask it to put together various PyTorch scripts for ML stuff. Thinking about it as businesses or jobs "not being safe" is the wrong tack though. This technology is going to 10x everything it touches, yes it'll be messy while we figure things out but in the end everyone is going to be a lot more productive.
I think your worries are very valid, and I agree that this could be a level step change in the consolidation of power that we already see among a few extremely powerful entities. Another observation worth noting is that SDEs who embrace this technology immediately are going to absolutely smoke past those who ignore it, and the gap will continue to widen as time wears on. Imagine having a slightly dumbed down SDE 1 or 2 (eventually 3) slave that can work 24/7. It will completely redefine what "entry level" means for software engineers.
This is a nice sentiment, but I think the answer is way more nuanced. You can't just roll into a developing economy and pay way over market without also disrupting the local economy and the people that live there. Imagine some similar situation in America, where for some reason, an international business comes in and pays 10-30X the market rate as similar businesses in the area, for the same product. The new jobs become highly (and potentially dangerously) desirable, other similar business go under because they can't keep up with the wage growth, etc.
To remain stable, economic growth must be slow and steady. The alternative is you simply don't go to Kenya, rather, you go somewhere else, and Kenyans get $0/hour.
The concern about danger really does not follow your hypothetical situation. Anyway, our view of capitalism would suggest that the failure of companies that use labor less efficiently is a net good.
I think the question we are all considering is why OpenAI behaves differently overseas than they might when trying to poach a smart engineer from a competitor in the Us.
It’s not a sentiment. It’s the definition of taking advantage of. Of course there are various arguments for justifications, but it doesn’t change the situation.
I also didn’t suggest anything like paying way over market.
So then what? If paying $2/hour is taking advantage, and paying over market isn't part of your solution...what is the solution? Not hire people in Kenya?
Also...yes it is a sentiment (a view of or attitude toward a situation or event; an opinion). Our opinions about what constitutes "taking advantage" are different. Saying otherwise doesn't make your argument more compelling.
> The last step was a really strange and uncomfortable 1-1 chat with a "very senior employee"
I had one of these once. The chief product officer came into the room, stared at me for awhile, then asked "tell me something you've thought really hard about...". It was a classic case of a narcissist exec. Glad I didn't get the job, as they are struggggggggggglin' right now.
This makes it sound so much easier than it actually is. Tracking where customers from on iOS device is extremely difficult. Now spread that difficulty over 10 different advertising channels (youtube, Facebook, direct visits, google, etc), each with millions of dollars in budget, and the problem quickly becomes impossible to track yourself.
Most of the ML candidates I see now are all "working with LLMs". Most of the ML engineers I know in the industry who are actually shipping valuable models, are not.
Cool, you made a chatbot that annoys your users.
Let me know when you've shipped a fraud model that requires four 9's, 100ms latency, with 50,000 calls an hour, 80% recall and 50% precision.