Live data from Hacker News

Goodbye, data science

ryxcommar.com

221–230 of 415 posts

Re: Goodbye, data science

#221

Calling it Data Science was a tell. Have you noticed how non-scientific things add "science" to the name to make it sound like it has scientific rigor? Data Science, Political Science, Social Science, Scientology Compare to Physics, Biology, Math.

Computer science...

Re: Goodbye, data science

#222

Earlier quoted context omitted.

Well good luck then, in my experience the most free time I've ever had in my life was during college. I squandered massive amounts of that time doing things completely unrelated to education, and I definitely don't regret doing that. College isn't just about book learning after all. But still, BY FAR, college is the time of my life when I had the most free time to do whatever I wanted.

Middle and high school is where a lot of students learn to stop being curious due to a lack of time. College demands far fewer hours per day, but it can be hard to forget what was taught previously.

If you are someone who is on the cusp of a better grade at university then any curiosity time is better invested in restudying the past exam papers. I think PhD has more of a curiosity culture at least in the first year but I never did one.

Also hard subjects at uni - there is only so much deep thinking you can do per day

Re: Goodbye, data science

#223

Earlier quoted context omitted.

I think you hit the nail on the head there with the survivorship bias and the raised in a bubble comments. Most people are raised in a bubble because children generally can't cope with how messy and complicated the world is. And systems and companies that last a long time can point to how successful they were because of their good decisions while ignoring their equally bad decisions that really should have undone the…

>The older I get, the more I realize how fragile a lot of human systems really are, but I suspect it has always been this way and it won't change significantly any time in my lifetime I agree that human systems have always been fragile, but have long been papered-over by things like "decency", "tradition" and "doing the right thing" and in extreme cases, mobs with pitch-forks. I disagree that it won't change in our l…

I took the prior post as in, "the fact that they are fragile won't change", not that the systems themselves won't change. And I would agree with that---I see it as yet another expression of the human condition. We may try to build order over chaos to make society, but we also keep loopholes and wiggle room for our psyches. I think the fragility of human systems emerges from that contradiction.

Students of history and the arts can get an earlier exposure to this worldview. I think we engineering types can get too focused on technology and imagine everything is innovation and progress. You have to work uphill against your default interests to expose yourself to a longer view and consider that fundamentally modern people with modern minds lived for (many) thousands of years doing almost all the same cognitive things as us, just with different physical props.

Our lungs are constantly in flux as we breathe. But at the same time, we're just breathing and that doesn't really change until our end. I'd say human social systems are much like that.

Re: Goodbye, data science

#224
post #59

I have never understood the what a good ML engineer couldn't do and a Data scientist could in _majority_ situations. When you need a decision to be made based on data its just common sense risk analysis added together with basic statistics. I feel some good field training in statistics(Look up Andrew Gelman) a couple of good courses on Linear, Bayesian Regression is all you need, rest is just engineering skill. The d…

I think the qualifying term here is "good". I've worked with a surprising number of MLEs that don't really understand gradient descent or how most models really work under the hood. They certainly couldn't implement most things from scratch if they needed to (neither could most data scientists). I used to think an MLE was a solid engineer who also had a strong quantitative and numerical computing background. The kind…

> I used to think an MLE was a solid engineer who also had a strong quantitative and numerical computing background.

Application of Computational Stats/ML Models are not all that hard to aquire but essential. I think we need a fundamental rethink of how applied stats/ML is taught to engineers to make them effective. Here are a few things I can think of:

1. Getting a solid understanding of actually coming up with a simple enough model to do the job 2. Do Power Analysis to figure out how many samples we need. Creating datasets with Hard Negatives and overcoming sampling bias. 3. Using things like Multiple Regression to do EDA. i.e. using models as a tool vs the end goal to understand a problem space.

Re: Goodbye, data science

#225

Earlier quoted context omitted.

Nope, nope and nope again. I refute this utterly, as a teaching academic. Contact hours at most universities are around 2-4 hours per week per 15-credit module. To gain a degree, you have to take 120 credits a year, typically two terms of 4 x 15 credit modules, or 8-16 hours of contact per week maximum with the entire summer off. You therefore have at least 24 hours a week to study on your own to bring your working w…

No personal attack taken but your experience and points fail to win me over. The difference probably belies in the rigor of the program. It sounds like you are working in a non-engineering based program. In our engineering programs we had 40 hours of class time + lab time per week. I had a concurrent arts degree at the same time which is was, in comparison, incredibly light workload - though concurrently it took time…

You posted this elsewhere in the thread, where I replied that this is not normal in the U.S. Can I ask what university and degree program it is where students have 40 hours of class and lab time per week?

Re: Goodbye, data science

#226
post #95

Earlier quoted context omitted.

>Meaning you could absolutely suck at your job or be incredible at it and you’d get nearly the same regards in either case. One of the things I don't like about statements like this said in a Data Science context, is that they are true outside of Data Science as well. Executives make big decisions, managers make smaller decisions, nobody can evaluate how good/bad they really were for months or years. Engineers build…

Not to get too off topic, but as a 35 year old engineer it seems the world in general has far fewer consequences than I was raised to expect. Everything from businesses with bullshit ideas flourishing at a loss, to January 6 even being possible (politics aside I expected the Capitol Police to crack a lot more skulls than they did once people started smashing windows), to the whole FTX situation and the tepid response…

I feel your pain. It often seems to me like Quality is on the decline, on many different fronts. Hard to say if it's just my perception. It does make me more fully appreciate it when I do encounter true craftsmanship or excellence -- which though it might be increasingly rare, is still relatively easily found.

Re: Goodbye, data science

#227

Earlier quoted context omitted.

Not to get too off topic, but as a 35 year old engineer it seems the world in general has far fewer consequences than I was raised to expect. Everything from businesses with bullshit ideas flourishing at a loss, to January 6 even being possible (politics aside I expected the Capitol Police to crack a lot more skulls than they did once people started smashing windows), to the whole FTX situation and the tepid response…

I think you hit the nail on the head there with the survivorship bias and the raised in a bubble comments. Most people are raised in a bubble because children generally can't cope with how messy and complicated the world is. And systems and companies that last a long time can point to how successful they were because of their good decisions while ignoring their equally bad decisions that really should have undone the…

Thanks for the concern, but I'm all right, I have the privilege of living near the top of Maslow's hierarchy and actually pondering these questions. :) If I'm a nihilist I'm at the "creating your own value system" part. The world is generally a giant blob of apathetic flavorless jello, I can at least inject some sugar and food coloring wherever I'm at. There's also some freedom in that, when people don't care they also tend to give way pretty easily. It's just disappointing, except for when you encounter that rare person that also gives a shit. Part of the reason I spend a lot more time on HN than reddit. :)

Re: Goodbye, data science

#228
post #132

> Nobody knew or even cared what the difference was between good and bad data science work. Meaning you could absolutely suck at your job or be incredible at it and you’d get nearly the same regards in either case. This seems to be a problem with the industry as a whole. I'm speaking as a SWE, but I've observed similar things with PMs. I don't think it's impossible or even very hard to appreciate the right things, it…

I think a big part of the problem is that most PMs are non-technical, and at some point up the chain so are most managers. Data science, when done thoughtfully, requires you to appreciate minute issues from data collection and management all the way through feature engineering, model selection/design, and validation.

The biggest, hardest bridge to cross was an appreciation of the importance of metrics. For some reason, getting a business person to grok something as simple as precision/recall/F-beta is a near-impossible task. You can do multiple presentations on it (after having honed those presentations over years with multiple manager audiences), and it never sticks. It's always "what's the accuracy?" It's impossible to do good work when your bosses insist on measuring and therefore optimizing for the wrong thing (which in my experience consulting for multiple Fortune 500 businesses, they always do).

Even worse, many organizations have such broken politics/cultures that the managers can't even tell you the big picture of what the project is trying to accomplish. Once you finally piece it together from the people who know their roles in-depth, it becomes clear that what they're trying to do is totally infeasible. At least that was my experience more than half the time.

Re: Goodbye, data science

#229
> 23 year-old data scientists should probably not work in start-ups, frankly; they should be working at companies that have actual capacity to on-board and delegate work to data folks fresh out of college.

If a 23 year old manages to get a data science job at a startup, and then actually delivers the results that the start-up expected of them, the learning experience there is infinitely more valuable than going to Google and using a bunch of tools that don't exist in the real world, on unrealistic timelines because you're not on the ads team and don't need to make money.

You can go learn "best practices" later, but working in a startup is an exercise in pragmatism. You deliver results, or you die.

Re: Goodbye, data science

#230
post #175
post #95

Earlier quoted context omitted.

>Meaning you could absolutely suck at your job or be incredible at it and you’d get nearly the same regards in either case. One of the things I don't like about statements like this said in a Data Science context, is that they are true outside of Data Science as well. Executives make big decisions, managers make smaller decisions, nobody can evaluate how good/bad they really were for months or years. Engineers build…

> One of the things I don't like about statements like this said in a Data Science context, is that they are true outside of Data Science as well. Executives make big decisions, managers make smaller decisions, nobody can evaluate how good/bad they really were for months or years. Engineers build something amazing, or build a house of cards, nobody cares as long as the money people are happy, even if the business use…

IMO the reason behind this is that a lot of "data science" driven decisions are short term decisions. So you can look at something on a PowerPoint, not really care if it's wrong unless you personally will get fired if it turns out to be wrong, and back out of it a quarter later when it turns out to be wrong. IME there's no shortage of justifications or pivoting when it comes to a decision you made a quarter ago. The consequences are relatively small, so the caring is only bravado, not really caring.

When it comes to disastrous long term decisions, there's plenty of time to get input from multiple stakeholders. I always remember the armies of companies who went chasing after Hadoop because Big Data was going to transform something or the other. All the stakeholders were on board, from the CEO and CTO to IT and Engineering management. How much money and time got flushed down the toilet trying to implement and extract value from data with Hadoop. They only people who paid the consequences were the employees at Hadoop companies who thought their stock options would be worth something.

Post reply on HN