Hey guys, I applied nonlinear regression to language data! I'm so smart. Please fuck my ass. I say "Machine Learning" even though there is no "machine" and "learning"! I used the Newton Raphson method and called it "gradient descent" and "backpropagation" because I wanted to look cool. By the way, I'm a computer science major. I have zero knowledge about modeling, optimization, and convergence of the approach. Moreover, I just installed tensorflow call myself a "Data Engineer". I just spam "black box" whenever a concept outscopes my comprehension.
Model development is democratized and that is a good thing. Nothing wrong with employing established models and tools in the real world. Do you think people working other jobs try to reinvent the wheel as well ever time? The brunt work has always been sourcing and cleansing anyway.The worst people are always Faka Basedsentists like you who have never worked with real data.
>>16539160Nobody hires someone to push a regression button. Get a real job.
>>16539145Seethe all you want but it's useful in real life no matter how much you strawman it.
>>16539160It's sheer hypocrisy to fool people by using existing models to pretend they've discovered something incredible.
>>16539196Literally no one does that, especially no Data Engineer. Meanwhile Faka Scientists that work in 'model development' have their models fail 95% of the time in the real world application.
>>16539181You never worked in Data engineering or Data Analysis and it shows. Pushing the Regression and the apply Model button is the easy part. For 99.9 of real use cases we already have optimized models. The work is in what comes before that.
>>16539145The AI is an abstraction class on computing. Iterative improvements in AI are now a kind of engineering feet, like improving processor chips. The everyday usage of an AI is completely encapsulated away from the underlying details of its implementation. You don't see it yet, but you will. It won't ever be a general intelligence, but it is a product that is here to stay.
it makes my games run at 200 fps instead of 15 and still look the sameit clearly isn't bullshit
>>16539145I did a ML project, shit is not labeled and the medical student just eye ball some of it. By the time the labeling cometh through I already had a basic eye balling programs to do the labeling with slightly more logical criteria and the results make logical sense, client satisfied and call it a year. I could've trained a model after its labeled but then I realize supervised machine learning is just asking the obvious.Mind you there are unsuperivsed and large language model but I am out of steam and ready to wage.
>>16539202it's complete marketing bullshit to call it "intelligence" anyway, it's linear algebra and statistics and optimization, not intelligence
>>16539160>>16539201Sourcing and cleaning is a laborwork that stimulates the brains of filtered midwits. It's like working in a coal mine. Giga chads like me who know proper math, prove theoretical convergence to the real values under specific curcimstances. (For your labor job to have a purpose.)
>>16539770>i'm too SMART to scrub the data i build my models on>the PROLES can decide whether or not my dogshit regressor will actually workinteresting approach.
>>16539770>>16539789too bad you are both machine learning cucks and are inferior to my superior human intelligence
>>16539837I'm evolutionary intelligence goyim.
>CS major checks out lmao
>>16539145the main part of data science is data cleaning nigga not using model.predict() lmfao thats easy
>>16539865what does that mean? you are retarded?
>>16539874>namefag>using a webshitter app to estimate IQ>namefag using a webshitter app to estimate IQYou are the blackest retard gorilla nigger I have ever seen.
>>16539865making machines smarter isn't a noble intellectual pursuit, you are glorified engineer
there's a few geniuses in the AI field and a whole bunch of wannabes who aint shit
AI is just a Kalman filter.
OpenAI CEO Sam Altman is negotiating major changes to the company's $14 billion partnership with Microsoft. The companies have defined artificial general intelligence (AGI) as systems generating $100 billion in profits [non-paywalled source] -- the point at which OpenAI could end certain Microsoft agreements, The Information reports.According to their contract, AGI means AI that surpasses humans at "most economically valuable work." The talks focus on Microsoft's equity stake, cloud exclusivity, and 20% revenue share as OpenAI aims to convert from nonprofit to for-profit status. The AI developer projects $4 billion in 2024 revenue.
>>16539963Cope harder
>>16540035Stop telling them
>>16539770LMAO
>>16539145Hello saar I will fuck you ass
>>16539160I work with real data. It's mixed. On one hand the ones who get a working product are indeed the ones who just deploy models with no real thought outside category, on the other... Most data once you dig into is actually very nuanced to the point that if you actually want rigor you have to almost build your own shit. It's mixed. The 80% corporate slop solution is just throwing best case algorithms at everything with no thought... But those workers are precisely why we're being led into a crisis of no one understanding why or how anything works. >>16539201Eh it really does depend. Like yes, but what do you consider easy? The work that I can get basic engineers who know SQL to do for X hours or the exquisite breakdowns that require prerequisite skill and knowledge to understand or know if you're doing things right? Like you have to balance your team and fundamentally, outside the ones who know networks, my data engineers are more expendable as 90% of their work is building databases and connecting them... Which isn't really difficult just long work. >>16539875Eh like in terms of time work yes... Maybe. But like the denigration of model.predict() is a problem. Model doesn't just go burr, you have to know the assumptions, axioms and risks. Data cleaning is basic, most of it is permissions then then the right joins and then checks to make sure the pipes don't fucking break. It's very mindless work. Figuring out what the data actually is, what distribution it falls within, what axioms and assumptions you are making, thinking about what to do with predictions or an understanding of the distribution of the data afterward, etc. is what I put my higher paid talent on. In my division we have tiers -webdev < journeyman data engineer < journeyman data scientist < senior data engineer < senior data scientist... The fundamental difference between them is basically just math backgrounds as well. Like I can't trust you unless I know you know the math...
>>16540947To add to this, this is also why you are seeing 'decision science' making a comeback. It is literally not just about getting the data or getting it clean or making predictions. It's understanding the data, understanding the predictions, figuring out what the optimization is, figuring out what utility is for everyone, constructing causality in order to determine optimal decisions, running simulations... Etc. These are things that have been forgotten and they really feed the rest of the loop... Because why are you making a prediction at the end of the day? Why are you making a visualization? Why are you pulling all this data together? To make decisions, to automate some as well... Are both those setups actually robust and rigorous?You realize how dire it is when you enter c-suite meetings...>>16540035No. It's just an SDE, kalman filters are just discrete SDE solvers.
>>16540947>journeyman data scientist < senior data engineer < senior data scientistApple/Face book have fires their entire Faka Scientist departments.Faka Scientist is a dead profession.
>>16540965A kalman filter is just a linear-Gaussian conditional mean approximation. Continuous-time KF's are also a thing (but they mostly are an academic exercise because your observations are almost always discrete-time in real radar/sonar contexts).
>>16540969SDEs exist both on discrete and continuous scale and you can go between them. And yes, I have done the continuous state KF and then I realized how many assumptions are built into the stochastic interpolation (i.e. taking discrete observations and going to continuous) and laughed at how engineers largely ignore the problem as is typically (i.e. the infinitely well defined reasonable dynamics that can define the in-between time steps). >>16540967Yes. They always do this. Analytics is the first thing cut because companies can get away with bad decisions. This is not new in the industries. >Faka scientistsI don't disagree, I hate the term. We used to just be called analysts. The government does it right and call it by the proper term, operations research analyst. Data scientist is a corporate Bojangle word but you have to use it for hiring purposes.
